Home

Desarrollo Software de librerias para el control y la adquisicion de

1. 2400 MAX_PIXELS_X 2240 MAX_PIXELS_Y 2344 PIXEL_SIZE 0 050000 FILTER THICKNESS 1 000000 FILTER MATERIAL AL ANGLE_OFFSET 0 000000 ROT_DIRECTION 1 CONTROLLER UNIT ID FOR RING y CONTROLLER UNIT ID FOR BED z CONTROLLER UNIT ID FOR DETECTOR w MINIMUM Dsd 205 000000 DEFAULT Dsd 205 000000 DET_TRAVEL 95 000000 DET_STEP 5 000000 Frame grabber configuration files FORMAT_BIN_1 video config FORMAT BIN 2 fconfig bin2 FORMAT BIN 4 fconfig bin4 System correction and configuration files D CORRECTIONS FILE d corrections CT txt MOTOR CONFIG FMT CT txt DARK FILE BASE dark FLOOD FILE BASE flood Swept parametres FORMAT SWEPT fconfig bin4 VOLT SWEPT 40 000000 AMP SWEPT 20 000000 OVERLAP SWEPT 0 Connection ports MOTOR PORT dev ttySNX1 configureTECHNO Initializing TECHNOSOFT System on port dev ttySNX1 Baud Rate of 9600 HOST ID 0 station config Station ready TUBE PORT dev ttySO Calibration Args GAIN_CAL_PATH gain_cal_data NUM_STEP_GAIN 8 NUM_IMG_B4 40 NUM_IMG_B2 10 73 Annex A System initialization log when simulating step and shoot acquisition NUM_IMG BI 10 GEOM_CAL_PATH geom_cal_data X ray tube maximum ratings MAX_CURRENT 800 MAX_VOLT 110 OOO Eng of configuration summary 78 Aedo le ee ell ella le le ll ll ell ej e parse_args X ray source disabled parse_args Correction of projection images disabled init_tube X ray source disabled
2. 7 storage capacitor TFT pixe electrode Figure 11 Direct left vs indirect right flat panel detector 21 19 20 Chapter 2 Principles of CT imaging Frame Grabber The device in charge of generating the synchronization pulse that starts up the pixel acquisition line and afterwards of transmitting what the sensor reads to the hard disc allocated in the working station The user can define the exposition time or the photodiodes integration interval acting directly over Signal to Noise Ratio and deciding the acquisition protocol 11 Working Station A master CPU hosts the calibration register and control software as well as the user interface It is also where image reconstruction from the acquired data takes place 2 4 Market research study The increasing demand of this sort of imaging tools has contributed to settle micro CT systems as part of the health instrumentation market We present in table 2 a comparison among the best rated models which yields a good picture of the state of the art of the hardware Device Company Specifications SuperArgus SEDECAL eXplore Locus RS GE Healthcare Inveon Siemens Xtream CT Scanco SkyScan 1176 Bruker TomoScopeSynergy CT Imaging Ey p Uno I f ae Resolution um 20 50 100 45 90 10 MTF 20 40 104 10 MTF 15 10 MTF 80 10 MTF FOV diameterxlength cm 12x45 4 5x8
3. Dell OptiPlex 745 1 500 00 100 8 60 200 00 Matlab License 6 000 00 100 2 60 200 00 Total 400 00 9 F rmula de c lculo de la Amortizaci n A A n de meses desde la fecha de facturaci n en que el equipo es utilizado XCxD B periodo de depreciaci n 60 meses B C coste del equipo sin IVA D del uso que se dedica al proyecto habitualmente 100 MEA Costes Descripci n Empresa E imputable Internet Telef nica 960 00 Fedora 9 Red Hat Inc 0 00 Windows XP Professional x64 Microsoft 130 00 Office 2003 Microsoft 100 00 Putty Simon Tatham 0 00 Image J NIH Image 0 00 VNC Real VNC 0 00 Material de oficina 150 00 1 340 00 9 Este cap tulo de gastos incluye todos los gastos no contemplados en los conceptos anteriores por ejemplo fungible viajes y dietas otros Sobre coste directo de personal 1 724 41 16 66 Presupuesto Costes Totales es Personal 27 936 Amortizaci n 400 Costes de funcionamiento 1 340 Costes Indirectos 5 935 Total sin IVA 35 611 IVA 21 7478 Total estimado 43 089 Discussion and future work The work presented in this project contributed to the development of the new version of the libraries used for acquisition and control of cone beam micro CT systems and to the implementation of a transparent and usable interface for generic micro CT devices This software is implemented in C which is upwards compatible with C and one of the most freque
4. Ritter and E Lang Calibration model of a dual gain flat panel detector for 2D and 3D x ray imaging Med Phys 2007 34 3649 64 21 J A Seibert J M Boone and K K Lindfors Flat field correction technique for digital detectors Proc SPIE 1998 3336 348 22 A Sisniega Contributions to the improvement of image quality in CBCT and CBuCT and application in the development of a CBuCT system in Bioingenier a e ingenier a aeroespacial 2013 Universidad Carlos III Madrid 23 A Sisniega J J Vaquero and M Desco Design and Assessment Principles of Semiconductor Flat Panel Detector Based X Ray Micro CT Systems for Small Animal Imaging in Integrated Microsystems Electronics Photonics and Biotechnology E K Iniewski ed New York CRC Press Editor 2012 p 309 335 79 80 References 24 Micro Photonics Inc SkyScan 1176 Micro CT 07 25 2013 http www microphotonics com in vivo skyscan 1176 micro ct 25 CT Imaging CT Imaging TomoScope R Synergy Specifications 07 25 2013 http www ct imaging de en ct systeme e mikro ct e html 26 A Sisniega et al Comparative study of two flat panel X ray detectors applied to small animal imaging cone beam micro CT in IEEE Nuclear Science Symposium and Medical Imaging Conference NSS MIC 2008 p 3836 3840 27 H Mori et al High resolution and high sensitivity CMOS panel sensors for x ray in IEEE Nuclear Science Symposium and Medical Imaging Conferenc
5. step by step description of the general interface execution and the performed assessments It includes the middle tests which were performed whenever a low level library was finished and also the final acquisition program together with the image reconstruction procedure In chapter 6 the budget of the project is presented Chapter 1 Motivation and Objectives Chapter 2 Principles of CT imaging 2 1 Introduction Biomedical imaging is the discipline referring to several different techniques that are used to observe a certain property of the body and represent it using visual information in order to diagnose or monitor medical conditions It has been the breakthrough with highest impact on the clinical and preclinical environments of the last century becoming essential in daily clinical practice Indeed it highly increases the accuracy of diagnostics the effectiveness of treatment and the chances of abnormality detection When using medical imaging techniques the spatial distribution of a particular property of a biological tissue is measured and represented by the difference in light intensity between the image areas 1 e by image contrast According to the source of contrast two different modalities can be defined structural imaging referring to those techniques that provide morphological information on subject anatomy and functional which on the other hand put up information about functional features such as metabolism or bloo
6. 14 Opencom Initialization command Opens communication serial port configapisystem configfile Configures the motion system with the config configfile configuration file path passed in the first argument home axis id Homes the axis whose char identifier corresponds with the first argument get axis id Prints information about position speed and acceleration relmove axis id distance Moves the axis whose char identifier corresponds with the first argument to the relative position indicated by the distance and direction of the second argument assuming it is expressed in millimeters degrees absmove axis id distance Moves the axis whose char identifier corresponds with the first argument to the absolute position indicated by the distance and direction of the second argument Exit Exit the program close Before leaving motor shell the communication closeAPIsystem with controllers must be closed It also turns off standing power of the motors kill System emergency stop k Requires system reconfiguration Table 14 Motor shell commands With these commands we could verify the behavior of the motion libraries bed and ring displaced the specified distance home position was reached accurately hysteresis was minimized and motion could be precisely repeated 57 58 Chapter 5 Assessment 5 3 X ray tube module We designed a set of tests for the evaluation of the individual features of the library mana
7. Relative signal compared to an ideal flat energy response detector as a function of x ray tube working voltage and xray beam alumninum filtration We used 66 kVp 0 2 mm Cu as optimal x ray beam inferred from experimental results on optimization of the acquisition protocol described in 22 The anode current needed to get a suitable signal level approx 90 of pixel saturation level was 90 uA yielding an output power of 6 W Results With the acquisition protocol described in the previous section we were able to acquire a complete angular span of 360 of raw images However raw data required several correction steps before being suitable for tomographic image reconstruction In this section the resulting images are presented as raw data corrected data and after tomographic reconstruction 59 Chapter 5 Assessment Raw data The frames from the step and shoot acquisition for the assessment of the system interface were stored as 16 bit unsigned images of size 600x600 pixels with little endian byte order One of the acquired raw projection images can be seen in figure 28 Figure 28 Projection acquired at 225 66 kV 90 uA As stated above this raw image contains several non idealities that must be addressed prior to tomographic reconstruction By design the active area of the detector is smaller than the physical area The gain non uniformity of the pixel response creates structured noise that shows
8. adjustment Image reconstruction For the reconstruction of tomographic data we used a filtered back projection like approach implemented as part of 22 35 This approach is based on the well known FDK algorithm 15 A frequency ramp like filter yielding the high frequency version of the original projection data filters every row of the pre processed projection The filtered projection is then backprojected along its angular direction onto the voxel grid of the reconstructed data These three steps must be carried out sequentially since the output of each one of them is the input for the next but every pixel can be processed independently within each step This is the reason why the software takes profit of the highly parallelizable nature of the reconstruction process and the massive parallel computation capabilities offered by Graphic Processing Units GPUs For the present acquisition we performed two reconstructions of 560x560x586 voxels one using a Shepp Logan filter and another one using a Hann filter For both images the cut off frequency was set to the Nyquist frequency u mm 0 1 Figure 31 Reconstructed image with a Sheep Logan left and a Hann right filter The color bar on the right of both images represents the attenuation values The degree of cupping in the reconstructed images gives a hint about the uniformity of attenuation values across the volume For its estimation we measured the mean value of an 18x18
9. allocated memory Now the system is ready to exit the acquisition and data to disk processes 51 52 Chapter 4 Software interface design Chapter 5 Assessment Each coded module has its individual test bench from the lowest level of the interface to the highest The final data acquisition executable directly interacts with the acquisition interface unaware of the underlying components integrating this platform In this chapter every stage of the followed testing protocol is described The micro CT models used for the assessment of flat panel sensor and controllers were the Argus system and the fDOT CT system The x ray source and acquisition interfaces were tested on the DOT CT system 93 54 Chapter 5 Assessment 5 1 Detector module Test bench With the implementation of three main global functions we were able to run firstly a program that acquired a single snapshot and a sequence of 20 images see figure 24 and secondly to test the continuous acquisition approaches we implemented a test program with two processes one for image acquisition and a second one for image processing as shown in figures 25 and 26 Call detector constructor Open detector Print on screen image information Allocate memory for image storing Call ACQ_snap Launch free run video capture in pingpong style Call ACO_seq for 20 images averaging in blocks of 16 images Call ACQ_seq for 20 images without averaging End c
10. called CT number or Hounsfield Unit HU which is the most common unit for CT data representation CT value HU 2250 HU 3000 bone window C W 1000 2500 250 HU 2000 150 HU 1000 mediastinal window Y AA C W 50 400 OS 250 HU 250 HU AR 1450 HU Figure 2 CT number windows 4 Towards multimodal imaging systems Nowadays fields as genomics and molecular biology are on the cutting edge regarding both technology and diagnosis The first term encompasses the study of all the genes of a tissue at genotype DNA transcriptome mRNA or proteome levels sketching the fine scale genetic mapping of the entire genome s networks 3 The second field which at some points overlaps with the area of genomics is focused on the interactions between DNA RNA and protein biosynthesis 4 Thus both disciplines share a common scope the investigation of the roles and functions of single genes using three dimensional functional and structural studies preferably obtaining complementary information With modern medical and biological research focused on this common topic it is latent the urge of non invasive systems for experiments requiring longitudinal studies over long periods since disease follow ups are of high interest The vast majority of the aforesaid experiments use animals of reduced size such as rodents as subjects under study 95 of laboratory animals are mouse and rat models of human disease due to the r
11. implants e or patient exceeding the field of ineasu rement i ida heh ee he RI et beter Le tate eco das 15 Figure 9 Overall view of the DOT micro CT system with the shielding elements in Pe A Steer MER oe ia le ace ivy A LEE 17 Figure 10 Block diagram of a micro CT system sese 17 Figure 11 Direct left vs indirect right flat panel detector 21 19 Figure 12 Graph relating the components of CT systems with the target systems of this project and the hardware manufacturers ai ns 24 Figure 13 Hamamatsu C7940DK 02 X ray flat panel detector A and detail of the needle like CsI Tl crystal structures on the scintillator layer B Images courtesy of Hamamatsu Photonics K K u em es ie 25 Figure 14 Dexela X ray flat panel detector left and dark image sample right Dexela Ttt 29 E torte tee E eq purse UO Evae reped quu 26 Figur 15 Interface architecture i aic ers oa odo bid stent nd celo 30 Figure 16 Wiring and communication diagram of the flat panel sensor and frame grabber Bow aa 3l Figure 17 Detail of figure 15 isolating the motion module structure 35 Figure 18 Schematic illustration of the homing process for linear axes The blue arrows illustrate the triple movement to verify it accurately reached home position 39 IMAGE INDEX Figure 19 Ring home function for the DOT CT system Crosses indicate the starting point s
12. in preclinical translational research has become commonplace Nowadays it is widely recognized that non invasive imaging techniques can provide a window into the biology of the living laboratory animals especially when combining functional and anatomical information The use of imaging techniques opened up the unique possibility of making repeated measurements of biologic function and specific molecular or genetic events within a single animal over time In the aforementioned preclinical imaging environments anatomical information is usually provided by x ray micro CT For this reason cone beam CB CT systems have been developed using a variety of embodiments and geometrical configurations adapted to high resolution small animal imaging CB micro CT The proliferation of these systems is due to the degree of development experienced by the technique the relatively low cost of the technology and the ease of integration in combined designs The goal of this project is to develop a set of software libraries for the management of hardware components in generic micro CT systems and to integrate them into a high level application for the acquisition of micro CT data With all the presented considerations in mind the specific objectives of this project are Chapter 1 Motivation and Objectives To develop low level C libraries for controlling hardware elements in micro CT systems based on a previously existent set of libraries in C that targeted spe
13. offers two different working modes one for high sensitivity protocols where the x ray flux is minimal and one for high dynamic range protocols The main features of the detector are shown in table 4 3 3 3 4 Parameter Value Pixel size 0 075 0 15 0 3 mm Binning 1x1 2x2 4x4 Photodiode area 145 x 115 mm Number of pixels 1944 x 1536 1934 x 1536 Frame rate 26 86 fr sec Binning 1x1 4x4 Ni 309 electrons HFW 126 electrons LFW Sto dcha 1 4 M electrons HFW 0 4 M electrons LFW 73 1 dB HFW 70 1 dB LFW Resolution MTF 0 2 6 3 Ipmm Dynamic range Table 4 Dexela x ray flat panel detector characteristics X ray source The micro CT system makes use of a tungsten anode microfocus x ray source L9631 MOD2 Hamamatsu Photonics K K Hamamatsu Japan The source has 40 to 110 kVp energy range and maximum anode current of 1 mA limited as a function of kVp to deliver a maximum power of 50 W The focal spot size of the x ray beam varies linearly between 15 and 80 um with power for power values larger than 6 W A fixed collimator made of a tungsten bismuth alloy shapes the x ray beam to cover the detector area and a filter holder is included to place filter materials for the x ray beam 25 This applies for all target system except for the Argus system which uses a microfocus x ray tube with a tungsten anode the Apogee XTG5011 Oxford Instruments Oxforshire UK In this case the nominal focal s
14. the raw signal is contaminated by two types of correlated noise namely dark signal and gain heterogeneity of the detector 16 The dark signal is the image that results only from random generation of electrons and holes in the reverse bias photo detectors in the sensor area that generates a small current even when no radiation is reaching the sensor 17 Since this signal is always positive it has a non zero mean with a random component superimposed This non zero average signal forms an offset image that depends on the image integration time and temperature of the sensor and that must be subtracted from the acquired data The dependence of the offset on temperature makes it advisable to acquire the correction data just prior to the micro CT data acquisition The second component of correlated noise arises from the small differences in gain of the different individual pixels that generate slightly different signal values for a uniform flux of radiation The gain map of the sensor is usually estimated by averaging a set of flood images for which the detector is irradiated with the beam set at the acquisition protocol configuration in terms of kVp and filtration and for a value of exposure sufficiently separated from the saturation point of any pixel in the detector The straightforward way to correct the acquired data for dark and gain is the so called flat field correction 18 19 which consists in the subtraction of the dark offset and the subsequen
15. up as a bumper like pattern There are some faulty rows columns and individual pixels The methods used to compensate these non idealities are described in the next section 60 Processing Inactive detector areas were removed from the processed projection data by cropping the original images After reducing the image area to the active part defective elements were corrected by substituting their original value with the interpolation of the values of the surrounding elements There are several approaches to compensate the remaining non uniformity in pixel gain among which we have selected the classic flat field and dark current removal for the correction of raw data in the presented scanner It has been shown to largely compensate the non uniformity of pixel response and it is easy to implement as a Matlab script For these corrections we acquired the flood and dark images shown in figure 29 I At N Ii Figure 29 Dark map left and flood map right for the same exposure settings 66 kV 90 uA Acquired prior step and shoot projections The pixel value after subtracting the dark offset is divided by the result of subtracting the dark offset to the flood data The results of this procedure can be assessed by the comparison shown below in figure 30 61 62 Chapter 5 Assessment Figure 30 Raw data top row and processed data bottom row with two different contrast brightness
16. voe EVER esse e tru 28 4 1 Software architecture iii idad isis 30 4 2 Detector mode AA A ae 31 4 3 Motion mol a anni buntes 35 AAs O E A A E 43 4 5 System ME AC 45 5 1 Detector mod le a te 54 5 2 Gantry module e 57 5 3 X ray tube Mol uan innsin inansa iia el 58 DA Compleie systern reino a E E iia 58 5 5 RESIS I EEn E E E AE Seat eg E Dto aln 59 PROJECT BUDGET isciesccccesccestosceoctccsscesovcsseocsecescocbeoscsoodsscscosovosSeoesecedcosbtoscsoecsscavesseosseoeee OD SYSTEM INITIALIZATION LOG WHEN SIMULATING STEP AND SHOOT ACQUISITION 71 SYSTEM PROCEDURE LOG WHEN USER ABORTS ACQUISITION eeeeeeeeeeeseseseseseseseeees 15 INDEX xil Image Index Ligure Project workllows u a na ea oles ied aie ea Seale aa 3 Figure 2 CT number windows 4 ue areas 6 Figure 3 Human mouse size comparison 5 esses enne 7 Figure 4 Photoelectric interaction a and Compton interaction b 10 10 Figure 5 CT geometries parallel beam a fan beam b and cone beam c 10 12 Figure 6 Schematic illustration of the acquisition of a projection dataset 12 13 Figure 7 Sampling patterns of the frequency space ideal situation A real situation for backprojection approach B and the compensated situation C 10 13 Figure 8 Artifacts due to failure of electronics a patient motion b beam hardening c partial volume effects d metallic
17. voxel ROI at the centre and close to the edge of the cylinder defining Ucenter and Hedge respectively and computed cupping using the following expression Le ge Lente f 100 E Hedge Results yielded a cupping level of 12 5 in agreement with results in the literature 22 for the same system proving the correctness of the data provided by our new software implementation 63 64 Chapter 5 Assessment Chapter 6 Project Budget Chapter 6 Project Budget UNIVERSIDAD CARLOS Ill DE MADRID Escuela Polit cnica Superior PRESUPUESTO DE PROYECTO Ana Ortega Gil Bioingenier a e Ingenier a Aeroespacial Titulo Desarrollo software de librerias y aplicaciones para el control micro CT Duraci n meses 8 Tasa de costes Indirectos 20 44 000 00 Euros N I F no rellenar Dedicaci n Coste Firma de Apellidos y nombre solo a titulo Categor a hombres mes hombre Coste Euro S informativo a mes conformidad Alejandro Sisniega Ingeniero Senior 4 4 289 54 17 158 16 Ana Ortega Ingeniero 4 2 694 39 10 777 56 Hombres mes 8 Total 2 1 Hombre mes 131 25 horas M ximo anual de dedicaci n de 12 hombres mes 1575 horas M ximo anual para PDI de la Universidad Carlos Ill de Madrid de 8 8 hombres mes 1 155 horas ERO Periodo de MESE 26 Uso dedicado Dedicaci n HO Coste Descripci n Coste Euro depreciaci d proyecto meses E imputable
18. 0 8 4x5 5 10x10 12 6x24 6 8 x 20 6 5x15 X rayenergy kV 20 110 35 80 35 80 60 20 90 20 65 Focal spot um 15 80 10 6 50 80 50 Dector panel CMOS CCD FP 125 165 mm FP 170 mm CCD FP Min acq time s 15 15 0 15 80 lt 60 15 Max power W 50 40 40 60 25 Mid power Radiation safety vSv h lt 1 at any point on the instrument surface during scanning 7 8 9 6 24 25 Table 2 Comparison of current micro CT system models Chapter 3 Hardware The x ray system is a simple PC based setup with an x ray tube and digital detector controlled by data acquisition cards in the PC To generate CT data the object to be scanned is either placed on a steady stage around which a mechanical system with detector and x ray source mounted rotates or on a rotating stage between the x ray tube and the detector Three out of the four target micro CT systems are built upon this first named design The test bench design for Carlos III University implements an in vitro configuration where the sample instead of the x ray tube detector assembly rotates Such system designs have maximum flexibility to allow to easily vary both geometry and acquisition parameters in order to suit user s demands There are small differences among models of components Hardware specifications are described in the current section 23 Chapter 3 Hardware 3 1 Overview of the equipment used in this project The different target systems were listed at the beginning of
19. 344 Frame rate 2 4 9 fr sec Binning 1x1 2x2 4x4 Noise rms 1100 electrons Saturation charge 2 2 M electrons Dynamic range 66 dB Resolution MTF 0 2 5 8 Ipmm Table 3 X ray flat panel detector characteristics 25 Chapter 3 Hardware Dexela detector The second x ray flat panel detector contemplated for the software design is the 2315 model figure 14 Dexela Limited London UK A complete description and an evaluation of some of the performance parameters can be found in 28 It is an indirect flat panel detector based on a 150 um structured CsI TI scintillator screen and a high fill factor CMOS image sensor to detect the optical photons Figure 14 Dexela X ray flat panel detector left and dark image sample right Dexela Lmt 29 The x ray photons are stopped in a 67 6 mg cm2 0 15 mm thick CsI Tl scintillator and the resulting visible light is then detected by a 3072 x 1944 CMOS active pixel matrix with a pixel size of 75 um Full well capacity is the term used to defines the amount of charge an individual pixel can hold before saturating Each pixel contains an option for switching between two separate levels high full well HFW and low full well LFW modes The typical fill factor of this detector is around 84 The detector offers the possibility of extending the area of a pixel using pixel binning 1x1 1x2 1x4 2x1 2x2 2x4 4x1 4x2 and 4x4 pixels This design reduces the image integration time and
20. J Hsieh Computed Tomography 2009 Bellingham Washington SPIE 11 A Molins Aquisici n y Procesado de Datos en un Sistema de Imagen Animal Multimodal 2004 Madrid HGGM 12 L Juang X ray chest image reconstruction by Radon transform simulation with fan beam geometry Measurement 2010 43 3 p 447 453 13 J A Fessler Statistical image reconstruction methods for transmission tomography in Handbook of Medical Imaging M S a J M Fitzpatrick Editor 2000 ed Bellingham SPIE 14 X Jia et al GPU based iterative cone beam CT reconstruction using tight frame regularization Phys Med Biol 2011 56 3787 15 L A Feldkamp L C Davis and J W Kress Practical Cone Beam Algorithm Journal of the Optical Society of America A Optics Image Science and Vision 1984 1612 619 16 H Illers D Vandenbroucke and E Buhr Measurement of correlated noise in images of computed radiography systems and its influence on the detective quantum efficiency Proc SPIE 2004 5368 639 17 S M Sze and M K Lee Semiconductor Devices Physics and Technology 2012 Hoboken N J Wiley 18 Hamamatsu Photonics K K Hamamatsu Application Manual X ray flat panel sensor C7912 amp C7942 amp C7943 rev 2 10 kr1 150006 2003 19 A C Konstantinidis Evaluation of digital x ray detectors for medical imaging applications in PhD Department Medical Physics and Bioengineering 2011 University College London London 20 C Schmidgunst D
21. Nothing to do init_det Initializing values for X ray detector init_det Opening frame grabber device Configuration file for frame grabber fconfig_bin4 Opening EPIX R PIXCI R Imaging Board Using configuration param Library version PIXCI R 32 Bit Library 3 07 00 09 11 04 Driver version PIXCI R 32 Bit Driver V3 7 002 6 27 25 78 2 56 fc9 1686 Image frame buffer memory size 356352KBytes Image frame buffers 493 Number of boards jd Image resolution Xdim 600 Ydim 600 Colors 1 Bits per pixel 12 Annex B System procedure log when user aborts acquisition The user may decide to abort the acquisition for many reasons an unexpected event occurred the user forgot to set some parameters something happened with the patient Whenever something goes wrong the command contol c will abort the current acquisition The following log report illustrates the steps for terminating the acquisition without leaving zombie processes or blocks of allocated memory not released controlc The user has aborted the acq process exiting controlc Shutting down x ray source xray_turnoff X ray source disabled Nothing to do controlc Stopping ring quit_power Switching off power Power turned 0 controlc Stopping detector quit_power Switching off power Power turned 0 controlc Stopping bed 75 76 Annex B System procedure log when user aborts acquisition quit_power Switching o
22. Query about triggering mode set_trigger Sets triggering mode Table 5 Functions to access parameters The library can be extended by including additional functions which are only available for a particular hardware These functions are only accessible if the appropriate hardware is present One example of this kind of functions is the setting of the waiting time between the acquisitions of two consecutive frames in a sequence that is available only for the Dexela detector Once the attributes of the class are initialized the interface is ready to interact with the hardware device and perform data acquisition routines Control functions By opening the imaging board after the system initialization but before programming any data acquisition the connection to the frame grabber card must be started Once the required tasks are finished this connection has to be closed In order to perform these opening and closing of the imaging board two initialization functions have been developed do_open and do_close The abovementioned format file is specified at the time of the detector object creation and accessed within do_open For the flat panel detector control our library provides the functions described in table 6 Function Description do_open Opens the imaging board communication channel do_close Closes the imaging board communication channel Initiates continuous capture into buffer 1 and buffer do_pingpong 2 alternately do
23. Table 9 summarized the main methods used when programming system movements Function Description Home Homes the selected axis homeAIl Homes all the axes simultaneously Programs and performs selected axis motion for a specific distance Moveaxis wait move Waits until the selected axis stops moving wait move all Waits until all station axes stop moving stopaxisImm Aborts motion of the selected axis Power Turns on off the standing power of the selected motor Table 9 Motion functions 4 4 X ray source interface Managing the x ray source from the acquisition computer requires transmission of specific definitions and methods included in tubemanger through the serial port This instructions fall into one of the following categories Control Monitoring Status notification Error notification The control commands are not only the activate deactivate pair but also the functions for setting up emitting flux voltage and current performing self test resetting from an overload status or starting the warm up process Unless the warm up protocol has successfully finished 1 e the tube is ready these methods cannot be called The x ray source has an internal watchdog timeout If it goes off the radiation is stopped since no command has arrived for a guard interval Monitoring consists on checking the watchdog timeout value either periodically or on demand and also on sending a command for resetting the w
24. The motion library is formed by a set of C classes divided in two levels The upper level station provides a hardware independent interface while the lower level classes provide access to the hardware adapting the functionality offered by station to the underlying hardware Figure 17 illustrates the structure of the motion libraries station Motion module api func techno functions api axis techno axis Figure 17 Detail of figure 15 isolating the motion module structure Depending on the manufacturer properties of the motor controllers are managed through the corresponding low level class techno axis or api axis With respect to the kind of axis the functions perform differently depending on whether the particular instance to an object of the low level class refers to the bed the gantry or the detector The reason for this is that their movements required specific unit conversions and limits The ring rotates whilst the bed and the detector linearly displace for example their speed and distance are independently selected etc Therefore the object storing the parameters of the axis contains the basic information about the hardware it represents as listed in table 8 35 36 Chapter 4 Software interface design Technosoft Identification integer Displacement units counts radians degrees or distance units Acceleration conversion value Initial position Velocity Velocity range max min Acceleration Homing t
25. _unlive Stops the acquisition after the current field Table 6 Detector control functions Acquisition routines Three different approaches for image acquisition were implemented namely single frame acquisition snapshot multiple frame acquisition in sequence or multiple frame acquisition in continuous mode 33 34 Chapter 4 Software interface design Single frame Acquiring a snapshot is a simple process that calls do_snap The process actively waits for the last field acquired to change its value This means that a frame has been acquired and temporally saved in the block memory indicated by current buffer field The method concludes by dumping the image into a user buffer previously allocated in the high level application Multiple frames In general a single image does not provide enough data for imaging purposes The user requires larger number of images at different angular and or linear positions Launching free run video capture takes advantage of the maximum data rate provided by the detector 18 thus reducing acquisition time and delivered dose These processes initiate a continuous capture of images into a couple of buffers from the grabber alternately i e ping pong style In multiple frame modes the value of the averaging block size variable is the number of images to be averaged in each angular position Both sequential and continuous acquisitions serve this purpose Frame sequence This method captures a s
26. ain inequalities and the polyenergetic nature of the x ray beam reinforcing the energy dependency of the attenuation factor as shown in the equation below 10 I I Si pte exp He x dx de 0 2 3 Where p represents the incident x ray spectrum and J represent the incident and transmitted x ray intensities with energy E and u x is the linear attenuation coefficient of the object at the same energy This energy dependence is usually ignored by later image processing and reconstruction stages to a large extent due to the lack of data to incorporate it in the processing chain giving way to the generation of quality degradation effects such as beam hardening It is desirable that the beam approaches a monochromatic i e mono energetic beam for minimizing the number of issues affecting image generation Other advisable characteristics of the x ray source are the followings 11 It must be able to produce enough rays within a short period of time The user shall have the chance to choose the energy of the rays beforehand Rays should be reproducible Beam coherence and radiation losses have to suit security standards We have already outlined some compromises and trade offs that shall be sought in compliance with image quality for diagnosis Pushing too far any design parameter will lead to image quality degradation Moreover the reconstruction technique assumes the consistency of all acquired values reflecti
27. an A detailed description and the evaluation of some of the performance parameters can be found in 26 27 It is an indirect flat panel 24 detector based on a CsETI scintillator screen as it can be seen in figure 13 B and a high fill factor CMOS image sensor Figure 13 Hamamatsu C7940DK 02 X ray flat panel detector A and detail of the needle like CsI Tl crystal structures on the scintillator layer B Images courtesy of Hamamatsu Photonics K K The x ray photons are stopped on a 67 6 mg cm 0 15 mm thick microcolumnar CsI Tl layer directly deposited over the CMOS photodiode matrix surface figure 13 A The active pixel elements of such matrix are connected to CMOS transistor switches that enable detection and read out and offer a high fill factor 79 The CMOS sensor makes use of on chip signal amplification channels with low noise level and an offset suppression circuit based on correlated double sampling CDS assigned to each of these channels This design achieves a high degree of image uniformity and a low noise level However correction tasks must be performed after the acquisition of an image in order to obtain the best achievable quality getting rid of residual non idealities The features of the flat panel detector as reported by the manufacturer are summarized in the table 3 Parameter Value Pixel size 0 05 0 1 0 2 mm Binning 1x1 2x2 4x4 Photodiode area 120 x 120 mm a 2400 x 2400 2240 x 2
28. apture Save data to disc Free resources Figure 24 Flow diagram for snapshot and sequence acquisition Continous acq Call detector constructor Open detector Print on screen image information Create and map shared memory object for image buffering Construct the shared structure in memory ACQ_cont While Istopping Acquire averaged image Lock semaphore Write data to shared buffer Buffer index Read stopping flag Release semaphore Figure 25 Flow diagram of the process managing the detector and frame grabber 55 56 Chapter 5 Assessment Image processing main Initialize the number of image noi Open and map shared memory object created by continuos acq program Obtain the shared structure Print out the pointers to check the memory blocks are correct i 0 Buffer_index 0 Lock semaphore Read data in shared buffer Buffer_index Free semaphore Alternate buffer_index Notify through STOP_CONTACQ Save data to disc Free resources Figure 26 Flow diagram of the process dumping the acquired datasets 5 2 Gantry module Motor shell To perform customized movements and have full control of the tests we implement a shell The motor shell is a program that allows us to type in commands and run them in Unix system controllers accomplish the request and the motor shell shows on the screen the generated reply The supported input commands are enumerated in table
29. ariables belonging to this class are detector station and tubemanager objects that have their own constructors as it has been shown in previous sections of this chapter It is worth mentioning that daq sys has a default set for the initialization of input arguments in case they are not specified at runtime The assigned values are shown in table 13 Number of angular positions Detector binning 1 2 or 4 Frames per position 1 Inf Image axial size px Detector s maximum pixels in Y Image radial size px Detector s maximum pixels in X Acquisition mode 10 normal step shoot 10 11 normal continuous Anode current mA 0 0 float X ray source peak energy kVp 0 0 float Bed positions to scan 1 Overlap between two consecutive 20 bed positions mm Initial bed position mm Scan angular span to perform rotation projection per scan angle noi 360 0 float Source to detector distance mm Detector dsd_default value External trigger through frame grabber O NO 1 YES f 0 No external trigger no need Exposure time ms i to select exposure time Scan files base name unknown patient Table 13 Acquisition rutine default values Acquisition procedure The scanner is ready to perform the acquisition tasks after each component has been initialized Once the acquisition scheme is determined the system starts scanning and collects the images that are stored for us
30. at panel digital detectors are substituting CCDs connected optically to a scintillation screen in despite of their wide spread use The so called flat panels are semiconductor based light detector matrices which compete in resolution and image quality with CCDs but with a more compact and lighter design Their weakness is the speed being relatively slower From the point of view of detection we can find two different approaches direct or indirect conversion In any case the shared fundament is that flat panel devices convert x ray photons i e primary quanta to electric charges that are gathered and analogue to digital converted 23 We talk about indirect conversion if the imager has an x ray phosphor or scintillator screen to convert absorbed energy into visible light The photodiode layer on the surface of the array produces photo induced charge within each detector element and the aforementioned charge is stored in a local capacitor Opposed to indirect detectors direct flat panels use a semi conductor material layered between two electrodes and electron hole pairs are directly produced as a result of x ray energy absorption A high voltage bias placed between electrodes separate the charge pairs with negligible side spread While both procedures allow high intrinsic spatial resolution indirect conversion is preferred for small animal 21 x rays t phosphor HV bias N Screen electrode A gt 2 JZ pp gt PA selenium 2
31. atasets one after the other Shoot state The machine returns to the shoot state but this time accessing a different shared area The behaviour of the processes is exactly the same as described in the first shoot state the acquisition process acquires the image in the current gantry position while the data to disk process waits for the dataset to be ready Step state Analogue behaviour but processing the dataset from buffer 2 and moving the gantry forward to the next angular position Once the data Acquisition process A AN source y f b f Y object i 7 beam j NL J detector Data to disk process to disk process releases the shared area the system switches to shoot state Figure 23 Diagrams of step and shoot acquisition schema The acquisition loop continues so on and so for until the system has gone through all the defined angular positions After acquiring all the projection data the data acquisition process turns the x ray source off and homes the mechanical system and the data to disk process dumps the images to a file and clears its buffer to be ready for new datasets System Finalization The acquisition process closes both the motor system and serial communication link Next call is the detector stopping function which can be seen as part of the releasing memory procedure together with deallocating the global variables Meanwhile the data to disk process releases its
32. atchdog The status of the source may be notified at three different points in time before warming up for retrieving hardware errors interlock status and the pre heat monitor during the warming up process to obtain data about the heating status and once the system is ready determining which tube is interacting tube status or checking updating the x ray tube real output voltage and current Emission time and internal battery requests are also available The error notification function allows interpreting the different error codes that could arise during any command execution 43 Chapter 4 Software interface design Attributes of the tubemanager class store the configuration information of the source and they are accessible through get set functions Table 10 shows the tube structure struct SXTubeInfo X ray source model Actual voltage Actual current Actual status Warm up status Warm up pattern Warm up step Interlock status Hardware error Watchdog timeout value Power time Emission time Table 10 X ray source information structure within tubemanager Tubemanager instances can be built using void or one parameter constructor The second option assumes the communication channel is already setup and ready to transmit and it receives as input parameter the file descriptor for the port The control and status methods available in the tubemanager library are described in table 11 The structure handlers are not in
33. ating the detector object sets the default parameters which may be modified according to the configuration file within the opening function 47 48 Chapter 4 Software interface design Once created enough memory must be allocated for the buffer used to store images in disc The shared memory structure must be defined here as needs to allocate buffers and semaphores Initializing the motion system The motion controller system station requires in its creation two axis objects one for the bed and another one for the ring and format file Once station is instantiated a call to station configuration method configures the motion parameters of the axes In order to unequivocally know the location of every mechanical component axes should be homed Thus subsequent moves can be performed referred to that initial point Acquisition loop The user can select the acquisition scheme together with the rest of acquisition parameters by setting the proper flags and operands when launching the application System fetches dark data or in other words checks what is detected without exposure just before turning the x ray source on Dark frames are used in correction stages later on Once both the gantry and the tube are ready the acquisition protocol is loaded step and shoot or continuous in order to make it possible to obtain the projection data for the different angular positions Most of the current systems use a continuous acqui
34. by the manufacturer TML lib 33 On the other hand when using API controllers a dedicated software interface between the serial port and the user calls was to be implemented the class serial io Motion types At first sight the clearest classification of motion tasks differentiates between axis homing and single movements Home functions return the selected axis or all at once to the position of reference while movement functions performs a concrete movement of the distance passed as input parameter Home routines vary depending on the type of axis For the case of the linear stages e g bed in figure 18 the task begins by stopping the axis wherever at the current position and reconfiguring the clockwise limit input as standard input Continuous displacement starts until the limit switch is activated For accuracy the bed steps back at a lower speed until the limit switch is deactivated and steps forward again at an even lower speed until it is activated again This is called a hysteresis cycle Limit switch Figure 18 Schematic illustration of the homing process for linear axes The blue arrows illustrate the triple movement to verify it accurately reached home position The algorithm for homing the ring varies from one system to another Nevertheless we focus the explanation on the home algorithm for the fDOT CT system since it is the system where our software was assessed The homing process starts by moving the rin
35. chapter describes the architecture of the developed software how the modules are divided into classes and how they are integrated in an interface for the acquisition of CT data The development presented here was based on two sets of existent software designed for two of the target systems the Argus CT and the fDOT CT Each one of these two systems ran dedicated software specifically targeted to the particular hardware of the system The existent software tools were developed in C language As was previously mentioned the software developments included in this project had the main goal of providing a set of software libraries offering an interface independent of the underlying hardware allowing the possibility of modifying hardware components without any change in the upper level acquisition routines For the last reason we have designed a set of C libraries one for each of the main hardware components with a two layer architecture Each library consists of a set of C classes managing the low level hardware and an intermediate level layer consisting on a single C class interfacing with the high level acquisition routines The low level objects are the units on which the system interface will relay since they are the ones directly interacting with manufacturer s hardware libraries There is an evident analogy Chapter 4 Software interface design between hardware and software modules Above them three classes were design for each
36. cific hardware The libraries must provide a generic interface allowing the substitution of hardware elements with minor modification to the code and maintaining the same functionality To design and implement a high level application for the acquisition of micro CT data transparent to the underlying hardware To integrate the software package in different platforms and operating systems To evaluate the performance of the software The work included in this project is framed on one of the lines of research carried out at the Biomedical Imaging and Instrumentation Group at the Departamento de Bioingeieria e Ingenier a Aeroespacial of the Universidad Carlos III de Madrid A significant part of the developments presented here has been or will be integrated into the preclinical high resolution scanners manufactured and distributed worldwide by SEDECAL Madrid Spain Key milestones This project has a clear task division since the software modules are directly related with the hardware components of the imaging systems We first developed the libraries for the flat panel sensor data collection Then proceed with the mechanical module and the last interface the x ray source manager At the end of every stage the interfaces were assessed as standalone hardware blocks While working on the tube interface we received the manufacturer specifications and drivers for a new flat panel sensor working on Windows operating system Therefore we d
37. class The configuration file starts with the identifier of the serial port for communication followed by the number of axes in the system and the sequence of axis identifiers The parameters for each axis can be for example arranged in columns tab indent The kind of configuration file we used for this project has the following design ok ok ok ok ok Serial port ok ok ok ok ok ok Ioport dev ttySO Number of controllers and their identifiers api_num 4 Joapis XYWZ Pactivate 0 1 1 1 Home type 1 switches 2 ring 3 hall 4 hall 2 tipo_home 3 2 4 3 Motion type O rotatory linear 1 rotatory rotatory jotipo mov 0 1 0 0 U Api managing CT relay orele_api rele_output Api used for connecting handset handset_I_api 1 Api moving with handset orders Zohandset M api 2 KKK KK K K K Groups id eee Zocoldet group 2 fEasySetUp File motor conf files owis t zip motor conf files newport openloop n2012 t zip motor conf files owis t zip motor conf files owis t zip die ie is AS 2 2 CoOnversions 7 7 2 AE AE AEH 2 2 oleo he le le kk kkk k 1 pos_unit equals mm 0 1deg mm mm rev_conversion 51200 2560 51200 51200 1 speed_unit equals mm s deg s mm s mm s vel_conversion 40 96 20 48 40 96 40 96 accel_conversion 0 03277 0 016384 0 03277 0 03277 RRE Motion Variables axisposition 0 0 0 0 max_desp 295 3600 95 295 minvelocity 3 0 5 0 3 0 3 0 maxv
38. cluded as they are just accessories with get set prototypes Function Description x_on Turns on x ray emission x_off Turns off x ray emission Sets user defined voltage and anode Control see current methods x chWD Checks the Watchdog timeout value x tsf Starts the self test X rst Resets from over load status X wup Start the warm up protocol Status methods x_checkConn x_typ x_sts X swe X SWS X snr X svc x update set X check times X check battery X err Checks the communication link Discovers with which tube the system is dealing Queries the tube about its current status Warm up process status check Warm up process step check Checks the status when the tube is not ready Checks the real output voltage and current Checks the preset voltage and current Checks the soure power on and emission times Checks system internal battery status Interprets error codes Table 11 Tubemanager methods 4 5 System interface The acquisition interface acts as the coordinator of the three modules detector controller and x ray source Among other tasks it has to manage memory allocations parse command line parameters and initialize each device with the correspondent configuration data A single method from this class is able to perform complex tasks that require multiple calls to the underlying libraries The collection of methods included in the class and a brief d
39. cro CT system 17 Chapter 2 Principles of CT imaging Gantry This term refers to the mechanical part of the system Specifically it encloses the ring and the chassis X ray Source This is the element in charge of generating the x ray radiation The x rays are created by the collision of electrons which have been accelerated using high voltages with a metal target tungsten for the case of study the anode It is this transference of energy whether partial or total what generates the x rays There are two main situations prone to generate radiation namely High speed electrons are forced to decelerate by the electric field of the target nuclei generating what is called Bremsstrahlung radiation In the worst case scenario high speed electrons hit the nucleus directly Although it is the most unlikely mechanism it is also the one which generates higher amount of energy since the entire kinetic energy is converted to x ray energy High speed electrons interact with target electrons with enough energy to eject them from their electronic shell When outer shell electrons fill in the vacancy characteristic x rays are emitted Diagnostic x rays are on the range of mid soft x rays due to their moderate penetrating capability On the other hand they provide remarkable high amount of low contrast information This is the reason why x rays with wavelengths ranging between 0 1nm and 0 01 nm energy range 12 4 124 keV are useful
40. d perfusion X ray Computed Tomography CT can be defined as the structural medical imaging modality that provides a volumetric representation of the attenuation coefficient shown by the different body tissues to the incoming x ray radiation Chapter 2 Principles of CT imaging This non invasive imaging technique consists on the acquisition of multiple projection images from different angular positions around the subject These projection images record the attenuation undergone by the x ray beam when traversing the subject under study Thus the projection images provide line integrals of the physical quantity under evaluation the x ray attenuation coefficient The acquired projections relate to the distribution of the attenuation coefficient of the different tissues inside the subject by the Radon transform Radon s theory 1 states the complete characterization of the two dimensional distribution of a property inside an object by an infinite number of line integrals According to this for the formation of the CT image X ray attenuation measurements are carried out in all angular directions around the subject forming a projection dataset for each angular view Many narrowly spaced data points are measured for each projection for the reconstruction of accurate anatomical images 2 Usually the attenuation coefficient of a tissue is referred to that of water to minimize the impact of non idealities of the acquisition process yielding the so
41. e e Universidad a 5 AMBOS de Madrid Departamento de Bioingenier a e Ingenier a Aeroespacial PROYECTO FIN DE CARRERA DESARROLLO DE SOFIWARE DE LIBRERIAS Y APLICACIONES PARA EL CONTROL DE SISTEMAS MICRO CT Autor Ana Ortega Gil Tutor Alejandro Sisniega Crespo Legan s 18 de julio de 2013 11 Titulo DESARROLLO DE SOFTWARE DE LIBRERIAS Y APLICACIONES PARA EL CONTROL DE SISTEMAS MICRO CT Autor Ana Ortega Gil Director Alejandro Sisniega Crespo EL TRIBUNAL Presidente Manuel Desco Men ndez Vocal Julio Villena Roman Secretario Javier Pascau Gonz lez Garz n Realizado el acto de defensa y lectura del Proyecto Fin de Carrera el d a 18 de julio de 2013 en Legan s en la Escuela Polit cnica Superior de la Universidad Carlos III de Madrid acuerda otorgarle la CALIFICACI N de VOCAL SECRETARIO PRESIDENTE 111 iv Agradecimientos El proyecto fin de carrera es el Ultimo proyecto que se elige exclusivamente por y para uno mismo Es el momento de disfrutar con lo que has aprendido Agradezco al Laboratorio de Imagen M dica del Hospital Gregorio Mara n en especial a M Desco y a J J Vaquero que me brindasen la oportunidad y la idea para realizar este proyecto A mi guia y modelo A Sisniega por estar siempre disponible para resolver mis dudas corregir mis errores y compartir sus conocimientos Ha conseguido contagiar a sus proyectandos su motivaci n y dedicaci n por la inves
42. e NSS MIC 2002 p 29 33 28 A L Goertzen Development of a Combined microPET and microCT System for Mouse Imaging 2003 University of California Los Angeles 29 Dexela Lmt 2315 CMOS x ray detector specifications 2010 30 Technosoft S A ISCM4805 8005 v1 3 Technical Reference 2010 31 API DRVKERNEL Serial Communications Kernel for the API Intelligent Microstep Motor Drives DM 224i and DM 22051 Vol Version 2 1 1999 32 PIXCI Reference manual XCLIBtm Imaging Board C Library VERSION 3 7 2011 Epix Inc 33 Technosoft S A TML_LIB v2 0 User Manual P091 040 v20 UM 0609 2009 34 I Gaztanaga Chapter 13 Boost Interprocess 2012 07 25 2013 http www boost org doc libs 1 53 O doc html interprocess html 35 J Aguirre et al Design and development of a co planar fluorescence and X ray tomograph IEEE Nuclear Science Symposium and Medical Imaging Conference NSS MIC 2008 p 5412 5413 81
43. ecide to carry out the necessary modifications to port our interfaces Once every block was finished we integrated them into the acquisition interface and finally assessed the complete software The above described work flow is illustrated in the chart below New detector acquisitio nassesme 2nd motion assesment 1st motion assesment Source setup assesment Figure 1 Project workflow Target systems The department is equipped with two CB micro CT systems and two more on the way namely FDOT CT Argus PET CT Test bench for Carlos III University SuperArgus Outline of the document The present document is organized in 6 chapters Chapter 2 presents the state of the art of the CT technique and an introduction to image formation in x ray imaging in general and in CT in particular including common quality degradation sources and image reconstruction Chapter 2 finishes with a brief introduction to the components of micro CT systems Chapter 3 presents a detailed description of the hardware employed for the project In chapter 4 the author presents an overview of the implemented software tools Following a down top approach each module is described in terms of configuration attributes and function prototypes Finally the software architecture integrating the low level modules is presented The validation of the performance of the software is presented in chapter 5 that consists on a
44. elocity 4 5 160 5 0 5 0 axisacceleration 30 100 0 300 300 37 Chapter 4 Software interface design KKK KK ee KK K Inputs of interest eee Isn 24 0 33 33 lsp 2 0 35 35 fhhome input 34 38 34 34 Get and set functions are the body of these classes and allow easily handling and fetching of attributes of the particular axis As stated above the functionality for performing motion commands is provided by another class named techno function and api func respectively The controller module calls multiple methods from this class to program and execute any kind of displacement Homing one or all axes Moving an axis with specific speed acceleration and or distance Waiting for one or all axes to complete the command Interrupt a motion task in progress As in many other cases the communication link needs to be opened when the system is initialized Consequently it must be closed when the system is shut down Commands are passed on a write read like message interchange station writes the order techno functions and api func build and transmit the message and lastly the controller executes it and sends back an answer notifying the operation status The techno functions and api func classes translate the requests by station into the commands understood by the particular hardware and forward them using the serial RS232 channel For the case of the Technosoft controllers they are directly accessed through a library provided
45. equence of n images with m averaged frames for each one of them The total number of desired images the size of the averaging block and the pointer to user allocated buffer are the input parameters of the function The frame grabber must be already running in video capture Every image is stored one after the other in user buffer until the number of acquired images equals the desired number of images The capture won t be stopped inside this method although no other image will be dumped from the board s buffers Continuous acquisition A call to this mode initiates continuous capture of images with m averaged frames in each one of them Images are stored in user dedicated memory The user is responsible of memory allocation for the storage Unless a second process toggles the end of acquisition flag images will be acquired non stop As it can be inferred two processes running simultaneously are needed for this scheme as it will be explained in detail in section 5 1 As the capture was started by this function it must also be stopped in this scope terminating after the current field or frame To sum up the acquisition functions are shown in table 7 Function Description ACQ snap Captures a single image ACQ sequence Provides a sequence of n images averaged in blocks of m frames ACQ cont Initiates PERDE capture averaging in blocks of m frames each image Table 7 Acquisition mode functions 4 3 Motion module
46. er processing To shut down the system memory has to be released and communication channels have to be closed The complete acquisition requires of two execution processes One of the processes is in charge of carrying out the movement of the system while synchronizing it with the flat panel detector In the meantime the other process transfers the data provided by the detector from the shared memory buffers to user defined memory disk In the following sections we explain in detail the tasks involved in each of the stages Initializing x ray source Before starting the initialization process it is necessary to check whether the x ray tube is enabled The first step is to setup the communication port and perform some preheating tests communication link availability source model and status batch check time properties Then assuming every returned value is correct set the periodic Watchdog check At this point the status of the x ray source will correspond to one of these cases Needs to be warmed up before performing any operation so the warm up process wup function has to be called next Already performing wup status that will cause the acquisition to abort Already emitting status that also aborts the acquisition Overload protection activated System error or interlock is activated i e the source is not ready Itis performing the self test process Initializing the frame grabber and detector Instanti
47. escription of their functionality is provided in table 12 Parse args Init CT Init det Init tube Launch WUP Alloc shmem Check interlock Contolc Read and store in a structure the values given as a result of the input parameters System initialization with the values read from the system configuration file Detector frame grabber initialization X ray source initialization Launch x ray source warm up protocol Some acquisition strategies require two simultaneous processes running and accessing frame buffers Therefore the shared resources must be concurrently handled Check system interlock status Called whenever the system is going to move To abort execution saving all the necessary data and closing connections 45 Chapter 4 Software interface design X ray_turnon and x Radiation control ray_turnoff Set the coach to the desired position It receives as input Set_bed and arameter an integer indicating whether the watchdog must set_bed_home P 8 5 e be reset Dark acq and flood acq Acquire data with correction purpose Step shoot Step amp shoot thread function Cont shoot Continuous shoot thread function Data process Data processing thread function Table 12 Description of daq sys methods Constructor The daq sys class does not have any other constructor apart from the void constructor mainly because any design parameter will be parsed from the program input arguments Global v
48. esemblance of their physiology and genetic with that of the human body 5 This fact has direct implications on instrumentation requirements namely the radiation dose shall be carefully reduced and limited to permit the performance of several scans on a single subject and electronic instruments shall be scaled to retrieve images with a higher resolution while maintaining a low noise level to achieve comparable image quality to that of human body scans Figure 3 Human mouse size comparison 5 Chapter 2 Principles of CT imaging In order to suit the resolution and radiation minimization constraints imposed by the size of rodents new medical imaging instruments shall be designed At this point CT comes to the rescue offering structural data that in combination with functional imaging techniques such as Positron Emission Tomography PET Single Photon Emission Computed Tomography SPECT or Fluorescence Diffuse Optical Tomography FDOT provides the accurate anatomic location of functional data in living animals From this alliance arises a brand new key interest increase the compatibility between devices and modalities i e create multimodal imaging systems Considerations for small animal imaging As previously mentioned there are some remarkable differences between human and small animal scanners which impose critical constraints over instrumentation namely Laboratory animals are a thousand times smaller than humans requir
49. ff power Power turned 0 close El puerto de comunicaciones esta cerrado controlc Comunication link closed Stopping acquisition PIXCI R imaging board closed 77 References References 1 J Radon Uber die Bestimmung von Funktionen durch ihre Integralwerte l ngs gewisser Mannigfaltigkeiten in Ber Verh Sachs Akad Wiss Leipzig Math 1917 2 W A Kalender Computed Tomography 2005 Erlangen Publicis Corporate Publishing 3 Interim Policy on Genomics E E P Agency Editor 2002 US Science Policy Council 4 W T Astbury Molecular Biology or Ultrastructural Biology Nature 1961 190 5 S C Cauter Now you can image a mouse like a man Laboratory News M Sc 2005 6 SCANCO Specimen microCT Brochures SCANCO Medical AG 07 25 2013 http www scanco ch fileadmin webmaster_img Brochures XtremeCT_v3 pdf 7 SEDECAL Super Argus PET CT SEDECAL 07 25 2013 http pmod jp pdf O8sedecal supera rgus datasheet pdf 8 GE Healthcare Technical Pubication GE Healthcare eXplore Locus User Guide Vers 2 07 25 2013 http www oucom ohiou edu ou microct Downloads eXplore 20Locus 20User 20Guide 20Rev 202 pdf 9 Siemens Inveon PET SPECT CT Siemens Healthcare 07 25 2013 https http www medical siemens com webapp wcs stores servlet PSOptionProductDisplay View catalogld 11 amp catIree 100010 1007660 1011525 1029715 1029721 amp langId 11 amp productld 200641 amp storeld 10001 J 10
50. for diagnosis As abovementioned this scanner is meant for small animal imaging Therefore to not degrade image resolution it shall make use of an x ray tube whose focal spot diameter remains as reduced as possible i e microfocus x ray sources are the most appropriate Filters of different materials are incorporated into the beam to improve beam mono chromaticity and reduce both radiation exposure and image artifacts Stability of the x ray source is also highly appreciated to maintain the noise and signal characteristics of the projection data across different angular views High Voltage Power Supply This component fixes the x ray source working settings in terms of acceleration voltage and anode current The current supplied shall keep the x ray sensor from saturation whilst getting good sensing performance at maximum acquisition speed The combination of these two parameters shall never exceed the power limit imposed by the x ray source constraints Thanks to an interface between the working station and the control unit the user can set the working point in the defined ranges X ray detector Detectors are meant to be fast and highly sensitive and at the same time stable If these characteristics are achieved the number of required correction datasets is drastically reduced and so does the acquisition time Detectors in small animal micro CT are integrated in the gantry whose motion demand light components For this reason fl
51. g in the clockwise direction until one switch mech placed at a fixed position changes its state After detecting the change in the switch and depending on its original state and the distance traveled several scenarios are possible 1 Ifthe state of the switch changes from 0 to 1 and the distance traveled was larger or equal than the distance separating the home position and the switch denoted as D from now on the ring starts moving in the counterclockwise direction a total distance of D 2 If the state of the switch changes from 0 to 1 and the distance traveled was lower than D the ring continues moving in the clockwise direction a total distance of 360 minus D 3 If the state of the switch changes from 1 to 0 the ring continues moving in the clockwise direction a distance of 180 minus D In order to avoid hysteresis effects on the switch detection the ring is moved back and forward more slowly to deactivate activate mech a second time The whole process is schematically depicted in figure 19 39 40 Chapter 4 Software interface design Figure 19 Ring home function for the fDOT CT system Crosses indicate the starting point solid arrows the direction of the first displacement and dashed arrows the direction of the second one Grey arrows illustrate the triple movement improve method accuracy The functions moveaxis or for simultaneous motion of two axes moveaxes can perform absolute relative or continuous movemen
52. gement of this sort of systems The aforementioned issues were partially overcome in the next generation when equally spaced rays arranged on an enveloping detection arc minimized translation and therefore the acquisition time This configuration was known as fan beam figure 5 B and it increased the sample area covered by the source detector system at each position However scanning with this geometry is restricted to a few slices For even quicker scanning and to minimize detector issues they resorted to fan beam with detection ring Nevertheless the economic investment required for building the stationary ring did not make this design viable Brand new detectors such as Charge Coupled Devices CCDs with flat crystals or flat panel detectors lead to cone beam projection shown in figure 5 C Samples still focus on a single point with a slight difference multiple fan beam planes are collected simultaneously to cover a volume This current trend answers high resolution demands for reasonably small active areas without a strict time constraint 2 11 Chapter 2 Principles of CT imaging detector te detector detector ZZ pp A Kun In c source source source a b c Figure 5 CT geometries parallel beam a fan beam b and cone beam c 10 The multi slice nature of cone beam CT allows getting the maximum benefit from every acquisition since the detected information can be processed with different methodolog
53. ght gray for existent but modified developments and dark gray for native code 4 2 Detector module The sensor makes use of a frame grabber card to interface the detector from the controlling computer A schematic depiction of the connections between detector frame grabber and computer is shown in figure 16 Both detectors contemplated in the presented development connect to an Epix frame grabber card Epix Inc Buffalo Groove IL and make use of the functions of XCLIB library 32 When first accessing the frame grabber it has to be configured according to detector specifications For this purpose when the system is initialized a configuration file created with the XCap software Epix Inc Buffalo Groove IL has to be loaded The library has been designed to work in Windows and Linux environments However the Dexela detector is accessible only in Windows due to restrictions of the manufacturer To achieve this duality we implemented a function that retrieves a handler to the appropriate library as a function of the operative system OS in the target computer using the dynamic library handlers provided by the particular OS REAR VIEW OS ACQUISITION SOFTWARE PC AT LIZ VIDEO OUTPUT 12 BIT DIGITAL Vsync Hsync GRABBER Pcik 00000 Oo X RAY SOURCE f MONITOR I i BINNING EY d binO bin1 1 IntExt y t ExtTrgGrb f 1 1 VOLTAGE h SOURCE 1 A Vdd D Vdd V 7 5 Ex
54. ging the x ray tube In order to minimize radiation and detector aging functions devoted to turn on and off x ray flux were not included in the assessment of the individual library and their validity was proved using the system acquisition interface Test bench 5 4 A testing program was designed to query and configure the x ray control unit despite the restriction of not using XON XOFF commands The goal was to check that communication channel transmitted the desired requests by printing out the received responses With this in mind we first setup the port and created a tubemanger object using the corresponding port descriptor We proceeded to configure voltage and current and query about the x ray source features and status fulfilling the information sheet shown in table 15 Values for tube structure Model L9631MOD2 Preset voltage value 60 kV Preset current value 90 uA Real voltage value 0 kV Real current value 0 pA Current status value NOT READY HW Error 0 Interlock status 1 Power time 962 h X ray emission total time 11 h Table 15 Tube information sheet Complete system The previous sections introduced the dedicated tests aimed at the evaluation of individual libraries of the software architecture However they are meant to work together providing full functionality to the micro CT system Since the acquisition interface integrates all components under a single library there were only m
55. idad de modificaci n en las aplicaciones de alto nivel Las librer as dise adas consisten en un conjunto de m dulos individuales para el control de detectores de rayos X fuentes de rayos X microfoco y controladores de motores proporcionando una interfaz transparente a la rutina de alto nivel Las librer as de manejo de hardware han sido integradas en una aplicaci n de alto nivel para la adquisici n de datos en sistemas micro TAC La funcionalidad de los componentes individuales fue evaluada en dos sistemas micro TAC desarrollados previamente Por su parte el conjunto completo fue probado en uno de los sistemas obteniendo un volumen reconstruido que prueba la validez de los componentes software para la obtenci n de imagen micro TAC Palabras clave rayos X TAC micro TAC C proyecciones adquisici n 1X Index N MOTIVATION AND OBJECTIVES ssesoossessossocssessoesocssessoesoossessoseoossossossoossosssesoossosssessossose PRINCIPLES OF CT IMAGING eseessessossoossessossocssessoesooesessossoossessossoossoesoesoossosssessossossseseossose D 2 1 Introduction tedesca e petebat step rc a ASEAS adds 5 22 Hundamens or AA RAN 9 2 3 Micro E components iso etal A A RSs 16 24 Market research Udri ae 21 3 1 Overview of the equipment used in this project 24 2c Blatpane detec tdi 24 3 3 X fay SOULCES oi e a a E e A cookie 27 34 Motionsystem nsessiensisninsknsaliaibsnntlateiiesnien AE 27 Aie Working SEA OM oie hese
56. ies and can be reformatted into sagittal coronal oblique or curving planes or even be displayed as surfaces and volumes Up to this point we have seen how and what does the electronic instrument measure and with what purpose reconstructing the structural image But how do we jump from attenuation values to image Image reconstruction Every point that gathers the emergent x rays is an element of the virtual matrix on which the computer will work Each one of the elements is called pixel when considering 2D and voxel when represented in 3D In other words from the point of view of the computer any sample can be considered as a voxelized volume crossed by the x ray beam Image reconstruction is the process of obtaining the spatial distribution 4 x y of the attenuation coefficient through the combination of the projection data from each angular view p 1 0 The sampling sequence starts with the set of measurements along x ray paths that are uniformly spaced forming a projection as shown in figure 6 Subsequent projections are acquired at different angles maintaining a constant increment between adjacent views until the 360 degrees are covered Figure 6 Schematic illustration of the acquisition of a projection dataset 12 Once the sampling data is available reconstruction of the attenuation function can be estimated through different approaches The most intuitive and extended approach is the inverse Radon transform The i
57. ing a consequent adjustment of image resolution only achievable by increasing radiation Laboratory animals are short lived which appeal for minimally invasive in vivo processes in order to reduce tissue damage during the scan and also to give the possibility of carrying out follow up studies Since the animal will be radiated quite often for this kind of studies dose must be minimized As a consequence of the two previous points a compromise between image quality and radiation dose must be sought In addition to the considerations presented above The animal must remain steady during acquisition In human imaging a single apnoea is enough for most of the acquisitions whereas in the case of animal models they are anesthetized creating the need of monitoring of biological signals during acquisition The facts exposed above can be translated into numbers characterizing micro CT scanners and distinguishing them from in vitro and clinical systems In Vivo animal CT Clinical CT Focal spot size 15 200 um 370 700 um XGq pone 10 300 W 60 100 kW Spatial resolution 50 200 um 400 1250 um Minimum Scan times 18s 0 33 0 5 s Flat panel CCDs Solid state detectors silicon CCDs UFC 30 100 mm 200 700 mm Detector Field of measurement Dose As Low As Reasonably Achievable Table 1 Micro CT vs clinical CT comparison 2 6 9 2 2 Fundaments of CT image Signal Formation The basic idea of Computed Tomography as it was previ
58. issing two processes running independently one for managing the system and the other one for processing the acquired data With all libraries and applications for the control of micro CT system successfully running and cooperating we planned a step and shoot acquisition of a soft tissue rat sized phantom with cylindrical shape and 4 equiangularly spaced metallic threads placed longitudinally over its surface To assess the correctness of the step and shoot implementation we performed a preliminary test simulating the whole acquisition process with no x ray radiation To this end we included an option in the dag_sys command synopsis which disables the x ray source no_xray This implies that tube related operations were not perform until the step and shoot scheme worked properly from start to end and images were correctly saved to disk The final test consisted in the acquisition of 360 projection images for a scan angular span of 360 with a 1 step retrieving a 600x600 projection binning 4 per angular position To obtain an appropriate signal level at the detector we selected the working point in terms of X ray source voltage and beam filtration that maximized the pixel signal for rat sized phantoms Figure 27 shows the plot showing the optimal voltage as a function of filtration for rat sized samples extracted from 22 Rat size 50 mm Sul Sideal 40 50 60 70 80 90 100 110 Voltage kV Figure 27
59. los ltimos a os la aparici n del TAC de rayos X de haz c nico CBCT en sus siglas inglesas ha hecho posible el desarrollo de sistemas de imagen dedicados para entornos de aplicaci n espec ficos Hoy en d a este tipo de sistemas son ampliamente utilizados en radioterapia y cirug a guiada por imagen imagen de regiones espec ficas e imagen precl nica de peque o animal A pesar del gran avance experimentado por la CBCT en la ltima d cada las diferencias con los sistemas de TAC convencionales hacen de esta t cnica un rea con un gran potencial para la investigaci n especialmente debido a las complicaciones introducidas por el uso de un ngulo de cono mayor y por el uso de detectores bidimensionales En el caso de sistemas micro TAC para imagen precl nica a las complicaciones anteriormente presentadas se a ade la necesidad de obtener una mayor resoluci n que permita la observaci n de las estructuras anat micas en peque o animal con el mismo nivel de detalle proporcionado por los sistemas cl nicos El objeto de este proyecto consiste en el desarrollo de un conjunto de librer as software de bajo nivel para el control de los elementos hardware de sistemas de adquisici n de imagen micro TAC de rayos X de alta resoluci n basados en fuentes de rayos X microfoco y en detectores planos Las librer as implementadas deben ofrecer una interfaz independiente del hardware subyacente y la posibilidad de a adir nuevos dispositivos sin neces
60. ls The resulting value corresponds to the projection measurement p p fat Le 2 2 The inherent difficulty lies in the fact that it is impossible to retrieve an infinite number of line integrals which will unequivocally recover the image A compromise must be sought between image quality for diagnosis and a sufficient number of lines 10 Geometrical configurations for CT acquisition The geometrical configuration of a CT system describes the arrangement of its hardware components Provided we will devote section 2 3 to deeply explain the components of a CT scanner as well as its design criteria in short it needs a x ray source and a sensor to record the attenuated x ray intensity Each and every component is mounted on the mechanical part of the system allowing translation and rotation movements for highly accurate sampling processes 11 The evolution of the CT systems has been partly related to the geometrical configurations from its irruption in the XX century to these days The first generation of systems turns to punctual x ray sources that emitted a beam with cylindrical narrow shape and incident on a single detection element To acquire enough projections of the area of interest it combined displacement and rotation movements The acquisition time for parallel beam scanners was quite long increasing the radiation dose and the risk of image quality degradation due to patient motion Figure 5 A shows the geometrical arran
61. mage to reconstruct can be thought of as a cross section through the specimen in which intensity values represent the attenuation coefficient Each attenuation measurement i e its projection value along its ray 1 0 gives a value of the integrated attenuation coefficient along the object which corresponds to the Radon transform through the object The inverse Radon transform provides an approximation of the image but significantly blurred The source of the blurring presented by the backprojected image is the overestimation of low frequencies and underestimation of high frequencies when adding data from all projections Adding a frequency dependent weight to the projection data can compensate this difference in sampling density 10 Note in figure 7 the overlapping of the different projections in the low frequency area and how weighting can compensate this effect A Ideal Sampling B C Real Sampling Real Sampling Weight weighting function Figure 7 Sampling patterns of the frequency space ideal situation A real situation for backprojection approach B and the compensated situation C 10 13 Chapter 2 Principles of CT imaging Analytical algorithms such as the Filtered Back Projection FBP follow this principle In general terms this algorithm is widely used because of its low sensitivity to space domain interpolation error and its reconstruction speed It is based on the concept that each slice is obtained by conv
62. me gets more complex interacting with the encoders for retrieving the exact position of mechanical components will grow in importance Also it will be worth to discuss the benefits of including correction algorithms at this hardware level or delegating the correcting tasks to higher level applications The library for controlling the x ray source from Oxford Instruments used in the Argus CT system will be implemented in the near future so that the interface can be integrated in the Argus system used in the department Regarding its commercial use this software will be included in at least two more micro CT systems that are being manufactured by SEDECAL 7 The first one already mentioned in chapter 3 is the test bench for the biomedical laboratory in Carlos III University This design will be used by students in their laboratory practices for learning how micro CT system works and for research purposes It is a system that allows plenty of geometrical configurations thanks to the multiple linear and rotational axes as it can be seen in the prototype in figure 33 Figure 33 UC3M Test bench prototype The second system is one of the targets defined in the work project Improving imaging platforms for the IMI funded PreDiCT TB project a research on model based preclinical development of anti tuberculosis drug combinations One of the working packages includes the design and implementation of an in vivo small animal CT imaging with e
63. nergy equal to the difference between the energy of the photon and the binding energy of the electron Outer shell electrons are the most likely to be knocked The vacancy left by the electron is almost instantaneously filled up by another electron from a higher energy level and a secondary photon results from the energy difference The energy carried by secondary photons in biological tissues is usually low enough to ignore them The photoelectric effect is the dominant effect in high Z materials and low x ray energies and it is considered the ideal process for x ray projection data formation Compton scattering In this type of interaction the impinging photon interacts with an electron from the outer shell of an atom in the medium The collision frees the electron from the atom whilst the incident x ray photon is deflected or scattered with partial loss of its initial energy and a different direction as illustrates figure 4 B that depends on the energy lost during the interaction Not much information regarding the location of the interaction or Chapter 2 Principles of CT imaging the photon path can be retrieved with such a wide deflection angle and therefore Compton scatter is usually a source of degradation in image quality Since Compton scattering is an interaction between photons and outer shell electrons its probability shows a dependency on the electron density of the material characteristic se radiation x recoil electron
64. ng every acquisition error in the reconstructed image This systematic discrepancy between real attenuation and the computed CT value is known as CT artifact and can be recognized in the reconstructed image in which undesirable lines shadows rings or other effects appear The sources of artifacts can be classified in four categories artifact due to physics beam hardening lack of photons subsampling aliasing due to subject patient motion metallic implants truncation due to electronics and mechanics poor calibration and limitations of the system and due to multi slice scanning data inconsistencies may appear in CBCT because Tuy Smith condition is not hold 15 There are two strategies to combat artifacts through correction or through avoidance 5 In figure 8 we can visually assessed artifacts belonging to the different categories defined above Figure 8 Artifacts due to failure of electronics a patient motion b beam hardening c partial volume effects d metallic implants e or patient exceeding the field of measurement f 2 15 Chapter 2 Principles of CT imaging Image pre processing gain and dark maps Raw image data provided by the FP detector is contaminated by structured noise that shows as deterministic patterns arising from the structure of the imaging sensor This correlated noise has to be removed to enhance the quality of the projection data and avoid artifacts in the reconstructed volume In particular
65. nhanced soft tissue contrast that will make use of the presented libraries 69 70 Discussion and future work Annex A System initialization log when simulating step and shoot acquisition The log below shows the messages printed out by the software while the initialization of the system is taking place Remember that before launching any acquisition protocol the system must be initialized as it was explained in section 5 4 so this will be printed out when the acquisition system is started for the first time or after abort an acquisition in progress The values are the ones read from the configuration files passed through command line in the real case scenario O k K K K K K K K K K K K K K K K K K K K K K K K K KK K K K K K K K K K K K K K K K K K K K ok VERSION ACQ 5 0 Subversions Daq CT Library version 2 0 Motion Control Version 1 0 X ray Tube Management Version 0 1 with Serial I O Version 0 1 X ray Flat panel Management Version 1 0 71 72 Annex A System initialization log when simulating step and shoot acquisition LELLLLLLLLLLLELLLLLLLLELLLLLLLLLLLLLLLELLLLLLLLLLLELLLLLLLLLLLELLLLLLLI init CT Opening CT configuration file OK init CT File CT params txt was read E OY Y configuration summary 7 7 ele ele eel kok kk k Physical Args HOME_X_POS 5 HOME_Y_POS 0 RING_START_POSITION 0 000000 BED_STOP_MAX 0 BED_STOP_MIN 471220 JOY_VEL 300000 BEDOFFSET 335 PIXELS_X 2400 PIXELS_Y
66. ntly used open languages with a wide range of compilers Its object oriented design compasses low level libraries in a single interface preventing the high level applications from dealing with direct hardware interactions and isolating device drivers code that really needs access to the hardware from the complex acquisition schemes making use of them The aforementioned factors made this acquisition middleware a highly portable interface Performance has been validated on the small animal fDOT CT system designed by J Aguirre and A Sisniega 35 The low level libraries have also been integrated in a user friendly acquisition interface in the platform Lab View see figure 32 that it is indeed internally used in the department 67 68 Discussion and future work Replace dialog box with Process 1 code Return to Idle state when complete S new picture new picture Crees 71 Index Array m Figure 32 LabView snapshot virtual instrument The snapshot vi is a call to the method in detector class Nevertheless there are still some features that we will like to improve and to incorporate in the interface Although the continuous mode has been implemented it has not been evaluated yet Neither has the instruction that performs the flat field calibration acquisition since calibration data is currently acquired by an independent process and not as part of the micro CT data acquisition process As the acquisition sche
67. ode is not activated If the user has specified the detector area to use the number of images to acquire and the format file beforehand then the constructor of the detector object will take these values into account and construct the corresponding object with user defined values With the aim of retrieving or modifying the parameters after the initialization get set functions were implemented for every parameter of the detector managing class so that they can be easily accessed within the high level application The library also provides methods for querying for parameters for parameters that are dynamically updated during the image acquisition such as the captured field counter or the buffer counter This information is of paramount importance for synchronization tasks during data acquisition Table 5 provides a reference of the functions we have just mentioned Function Description date Query about image resolution dimensions colors and bits per pixel dom Query about image frame buffers size number of buffers number of boards set_UD_ExpTime Sets exposure time get_tam_tot_x Query about the number of pixels in x coordinate set_tam_tot_x Sets the number of pixels in x coordinate get_tam_tot_y Query about the number of pixels in y coordinate set_tam_tot_y Sets the number of pixels in y coordinate get_ave_fr Query about the size of the averaging block set_ave_fr Sets the size of the averaging block get_trigger
68. of a higher image resolution to observe anatomical structures with the same level of detail obtained for clinical CT The purpose of this project is the development of a set of low level software libraries to control hardware elements of high resolution micro CT systems based on microfocus x ray sources and flat panel detectors The implemented libraries must offer an interface independent of the underlying hardware and the possibility of adding new hardware elements with no modifications in the high level applications The designed libraries consisted of a set of individual modules controlling x ray detectors x ray sources and control systems for motor drives providing a transparent interface to the high level application The low level libraries were integrated into a high level application for the acquisition and basic processing of micro CT image The individual components were tested on two existent micro CT systems assessing the functionality of the libraries The complete software architecture was included in an existent micro CT system and a set of CT data was acquired and reconstructed proving the validity of the developed software for the successful generation of CT volumes Keywords x ray CT micro CT frames images acquisition C vil viii Resumen La Tomografia Axial Computarizada TAC es una de las t cnicas mas ampliamente utilizadas para la obtenci n de informaci n anat mica en vivo y de forma no invasiva Durante
69. olid arrows the direction of the first displacement and dashed arrows the direction of the second one Grey arrows illustrate the triple movement improve Method ACCULACY inci aan ib 40 Figure 20 Flow chart illustrating the stages for performing a movement using API Controllers a A a a dd 41 Figure 21 Flow chart illustrating the stages for performing motion commands for Technosoft controllers syner cct 42 Figure 22 Angular projection depiction naar ea 48 Figure 23 Diagrams of step and shoot acquisition schema eee 51 Figure 24 Flow diagram for snapshot and sequence acquisition esses 54 Figure 25 Flow diagram of the process managing the detector and frame grabber 55 Figure 26 Flow diagram of the process dumping the acquired datasets 56 Figure 27 Relative signal compared to an ideal flat energy response detector as a function of x ray tube working voltage and xray beam alumninum filtration 59 Figure 28 Projection acquired at 225 66 KV 90 uA 60 Figure 29 Dark map left and flood map right for the same exposure settings 66 kV 90 uA Acquired prior step and shoot projections ocooooccccoccccnoncccnoncnononcconananonnnanos 61 Figure 30 Raw data top row and processed data bottom row with two different contrasUbrightness adjustment aueh lee 62 Figure 31 Reconstructed image with a Sheep Logan left and a Hann right filter The color bar
70. olving the image with a ramp filter presenting some limitations due to its finite projection bandwidth and finite projection order 10 Alternatively the iterative approaches assign specific attenuation values for each element without a priori knowledge of the subject itself and initially assuming homogeneity Attenuation parameters are updated depending on whether the calculated projection is overestimated or underestimated requiring several iterations before converging One example of this kind of methods is the Algebraic Reconstruction Technique ART 13 However despite the computational load being higher than in analytical approaches because of the need of performing several forward and backward projection operations the use of GPUs and their massive parallel computing capabilities has notably speed up iterative algorithms 14 Limitations of the CT technique The accuracy of projection measurements is threatened by the difference between ideal conditions and those of real scenarios Some of the assumptions or approximations may no longer hold in real imaging conditions Such non idealities impact diagnosis performance because of the distortions they introduce in the image which in some cases mime pathologies Some sources of errors are scattered radiation that mainly adds a low frequency bias that increases with radiated volume to the attenuation measurements non linearities of detector due to dark current biasing hysteresis or g
71. on the right of both images represents the attenuation values 63 Figure 32 LabView snapshot virtual instrument The snapshot vi is a call to the method in deteet r Class NON 68 Figure 33 UC 3ND Vest bench Prototypes span 69 xiv Table Index Table 1 Micro CT vs clinical CT comparison 2 6 9 seen 8 Table 2 Comparison of current micro CT system models sss 21 Table 3 X ray flat panel detector characteristics 25 Table 4 Dexela x ray flat panel detector characteristics eee 27 Table 5 Functions to access parameters ann asia 33 Table 6 Detector control nc iii a 33 Table 7 Acquisition mode functions a ae ero hoc ERE RE eB Ree eene Un ee Pew eee Se EN QR 35 Table 8 Attributes techno axis and api axis classes ooooccconoccccnononcnononononononononnncnnnn ocn 36 Table 9 Motion POHCHOEHS Cents essentiellen 43 Table 10 X ray source information structure within tubemanager esses 44 Table 11 Tubemanager methods wen al 45 Table 12 Description of dag_sys methods eene 46 Table 13 Acquisition rutine default values u ee 47 Table 14 Motor shell commands 2 42 2m aaa 57 Table 15 Tube information sheet essen e o enatis 58 TABLE INDEX xvi Chapter 1 Motivation and Objectives During the last decades the numbers of animal models of human disease have largely increased and their use
72. one of the main subsystems of a micro CT system detector gantry and x ray source 4 1 Software architecture System Interface detector station tubemanager Detector module Motion module Xray Source module Dex_det api_func Pixci_det techno_functions api_axis techno_axis Figure 15 Interface architecture The diagram in figure 15 shows a schematic depiction of the software architecture proposed in the present project The upper box on the diagram daq sys in figure 15 shows the high level application for the acquisition of CT data making use of the individual hardware libraries The developed libraries integrated into the system acquisition interface daq_sys are the detector and frame grabber module detector the motion module station and the x ray source interface tubemanager In order to provide a transparent interface to the high level applications regardless of the underlying hardware additional classes were required as can be seen in the lowest box see figure 15 Dex_det api_func api_axis and techno_axis were developed entirely as part of this project Pixci_det and techno_functions were part of an older version of the software and needed to be converted into C classes and modified to meet the hardware requirements And the C libraries api_basic_func serial_io and serial_io_n needed no modifications In figure 15 this classification is represented as coloured names black for the new developments li
73. ously outlined in the introduction lies on the following physical principles The x ray beam goes through the object body being attenuated during its travel The total attenuation suffered by the beam is dependent on the different attenuation coefficients of the traversed tissues and the traversed distance across each tissue The information about the accumulated attenuation is carried by the emergent radiation which is recorded forming a projection From the last fact it can be deduced that if the beam is emitted from a set of multiple angles surrounding the body and forced to cross through the same point for every angle we will retrieve enough information to determine the punctual attenuation coefficient Repeating this procedure over all the points of interest we acquire enough information to estimate the spatial distribution of attenuation coefficients within the subject In other words the 3D distribution of the x ray attenuation is reconstructed from either 1D stacked or 2D projection data X ray radiation in medical CT usually is on the range that comprises from 20 keV to 140 keV where three different radiation matter interaction mechanisms whose net effect is the aforesaid attenuation prevail Photoelectric effect It refers to the interaction between the photon and an electron in the medium in which the photons releases all of its energy and ejects the electron from the atom figure 4 A The electron is ejected with a kinetic e
74. pot size is 35 um Motion system The detector and x ray source are allocated on a rotating gantry This gantry consists of an 80 cm diameter aluminium plate that holds both components and a rotating motion stage model RV350PP Newport Corp Irvine CA The sample is placed inside the FOV of the system by means of a linear motion stage equipped with a stepper motor There is also a rotation stage that is either built upon a mechanical ring or the system holding the object to be scanned The motorized stages included in the system are managed by servo drives with integrated motion controller boards model ISCM 4805 Technosoft S A Neuchatel Switzerland or DM 224i API Controls Inc NY 30 31 The individual drives communicate via a CAN bus which end side is connected to the acquisition computer by a RS 232 serial link 27 28 Chapter 3 Hardware 3 5 Working station All the components included in the systems were connected to a regular personal computer in charge of the control of the whole system In the case of Argus and fDOT CT models the acquisition computer ran a Linux operating system Fedora Core 9 Whereas a Windows operating system Windows 7 was required for the Super Argus and UC3M Bench scanners The abovementioned personal computer was configured to host both operating systems so that the developed libraries could be tested and their compatibility validated Chapter 4 Software interface design This
75. s provide an overview of the states of the system Shared Buffer 1 Acquisition process Shared Buffer 2 Data to disk process Steady state Once the system is initialized the shared memory objects are ready to be used buffers have memory allocated and mutex semaphores are unlocked Each process uses a pointer to the current buffer 49 Chapter 4 Software interface design Shared Buffer 1 Acquisition process Shared Buffer 2 Shared Buffer Acquisition process Shared Buffer 1 Acquisition process Shared Buffer 2 50 Data to disk process Shoot state The acquisition process locks the mutex of the corresponding buffer in this case buffer 1 and acquires the averaged image from the flat panel at the current position Then it writes the data to the shared buffer for which he has exclusive access the data to disk process cannot read data yet Finally it releases the shared memory area and updates its pointer to next buffer buffer 2 Step state The acquisition process updates the gantry position Meanwhile the data to disk process after locking the semaphore accesses the shared buffer where the dataset acquired at the previous gantry position has been stored buffer 1 Once the data to disk process has finished reading the data it releases the shared memory area and updates its pointer to next buffer buffer 2 The data to disk process stores the d
76. sition scheme with continuous x ray beam flux Provided a fast detector is used this approach achieves better acquisition times Nonetheless it may increase the effect of image lag and the number of mechanical imperfections and synchronization flaws A popular alternative is the acquisition of a stack of frames for each angular projection figure 22 with the gantry steady during the acquisition step and shoot While the gantry rotates no image is acquired and the radiation beam is blocked to avoid the delivery of useless dose to the subject under study and to minimize the effect of image lag Flat panel detector Axis of rotation Object a X ray source M Trajectory Figure 22 Angular projection depiction As it was mentioned at the beginning of this section the acquisition interfaces requires of two executing processes The processes were synchronized using semaphores and shared memory objects of BOOST Interprocess library Version 1 53 0 34 Deepen on the step and shoot approach the acquisition process was implemented as an event driven finite state machine with states that take advantage of the maximum data rate provided by the detector 18 thus reducing acquisition time and delivered dose Since we were interested in acquiring an averaged image for each angular position we chose to acquire in sequence mode and to store the datasets in two shared memory areas in ping pong style The following graph
77. t division by the dark corrected gain map according to the following expression 20 for a pixel with coordinates u v R u v Or u v C u v G u v OG u v 2 4 Where R is the acquired data Oz is the offset data acquired for the temperature conditions of the acquired data G is the gain map and Og is the offset image for the temperature conditions of the gain map While the flat field correction approach provides sufficient image quality for most applications it inherently assumes a perfect linear gain curve for each pixel In real detectors however it remains some degree of non linearity in the gain curve of the pixels 21 causing the equalization of the pixel response to leave slight differences between the response of adjacent pixels that in turn translate into artifacts in the reconstructed image This effect is more evident for signal levels far from the calibration point 2 3 Micro CT components Scanner configuration An overall view of a fDOT micro CT system with shielding elements in place but without outer casing is shown in figure 9 It is used as reference model for any micro CT system therefore DOT specific components are labeled but not described Figure 10 shows a schematic depiction of the connections and communication channels between the different parts of the scanner Shielding etector holder Control Unit X Ray source Frame Grabber Figure 10 Block diagram of a mi
78. t to the required position The specified distance value has to be converted into controller internal units before transmitting the order The difference in execution between the API and Technosoft controllers relays on the fact that motion commands are immediately executed in API systems whereas in Technosoft systems a first command programs the required movement and a second command starts the actual movement If anything goes wrong controllers send a command to immediately abort movement Basic steps for moving an axis are summarized in the following flow diagrams in figure 20 for API controllers and figure 21 for Technosoft controllers Move API Setup Configure inputs Read config file Initialize communication Read user request Build message Transmit order to execute motion Wait Usleep More requests Home Check drive faults Check drives status Turn off standing power API close Figure 20 Flow chart illustrating the stages for performing a movement using API controllers 41 Chapter 4 Software interface design Move Technosoft Setup Configure axes Read config file Set detector in position Set bed in position Open communication port Read user request Program movement Wait Usleep Update motion immediately ore requests NO Close port Figure 21 Flow chart illustrating the stages for performing motion commands for Technosoft controllers 42
79. tTrgLemo i I 1 C7940DK 02 Figure 16 Wiring and communication diagram of the flat panel sensor and frame grabber 18 Regardless of the flat panel model the C class either pixci_det or dex_det which imports the XCLIB library must contain the following parameters Detector coordinates which include the upper and lower limits of the active area User selected averaging block size Path to the format file Flag for external triggering Variable storing the image exposure time 31 Chapter 4 Software interface design The classes interfacing implement functions of three different categories detector initialization and access to parameters control tasks and acquisition routines The set of functions is explained in the sections below Detector initialization and access to parameters The detector object can be initialized using a set of default parameters or using the values specified in the system configuration file In the event of using the empty constructor the default parameters for the configuration of the detector are extracted from predefined configuration files Regarding the detector area the upper horizontal and vertical coordinates are initialized to zero and the lower horizontal and vertical coordinates to 1 meaning to use the complete detection area Frame grabber status flags are assigned to zero indicating that the detector is not open it is idle and that the continuous acquisition m
80. the manuscript However a more detailed description of the specific models and their components is included below FDOT CT is built upon a Hamamatsu flat panel detector Technosoft controllers for the mechanical system and a Hamamatsu x ray source Argus PET CT system also uses a Hamamatsu sensor but the motion controllers are manufactured by API Controls and the x ray source by Oxford Instruments Test bench for Carlos III University has a Dexela flat panel detector Technosoft motion system and a Hamamatsu x ray source SuperArgus system relays on a Dexela flat panel detector Technosoft motion system and a Hamamatsu x ray source These model characteristics are graphically summarized in figure 12 Micro CT system Flat panel detector Mechanical system X ray source Hamamatsu aricontrois fDOT CT Super Argus fDOT CT Argus fDOT CT Argus UC3M Bench Super Argus UC3M Bench UC3M Bench Super Argus Figure 12 Graph relating the components of CT systems with the target systems of this project and the hardware manufacturers 3 2 Flat panel detectors Two different flat panel detectors were contemplated for the development of the controlling libraries A description of their main features and a brief explanation of its main parts are provided in the following subsections Hamamatsu detector One of the x ray flat panel detectors used is the model C7940DK 02 figure 13 A Hamamatsu Photonics K K Hamamatsu Jap
81. tigaci n A los compa eros del Lim con los que no s lo es un placer trabajar sino que es con los que apetece desconectar al terminar la jornada Hay trabajos que poco menos que invitan a dejar la sonrisa en la puerta y recogerla a la salida y luego est este en el que c mo entres sin ella te dibujan una antes de que te des cuenta A mi familia gracias por vuestro apoyo incondicional por vuestros consejos por vuestra ayuda constante por vuestras palabras de nimo y por vuestra gran paciencia Sin vosotros llegar hasta aqu hubiera sido imposible A mis amigos y compa eros de universidad A los amigos de siempre A todos con los que he compartido experiencias Gracias vi Abstract Computed tomography CT is one of the techniques most commonly used to retrieve anatomical information of living subjects non invasively During the last years the advent of x ray Cone Beam CT CBCT has enabled the development of dedicated imaging systems for specific environments Nowadays these systems are widely used in radiotherapy and image guided surgery imaging of specific regions and preclinical imaging in small animals Despite the advance experienced by CBCT during the last decade the differences with conventional CT make it a subject of vigorous research mainly due to the issues introduced by the use of a larger cone aperture and two dimensional detectors Micro CT systems for small animal imaging add to the commented issues the need
82. x ray photon o x ray photon Zur _ en ee Oo gt O 82 photoelectron gt scattered x ray photon a b Figure 4 Photoelectric interaction a and Compton interaction b 10 Rayleigh scattering This interaction occurs when electrons of an atom are set in vibration by an incoming x ray photon emitting radiation with a wavelength equal to the one of the electromagnetic wave that made them vibrate i e the original photon and a slightly different direction The interactions that produce the highest amount of absorbed or scattered photons are the photoelectric mechanism for low energy photons or the Compton scattering for photons above 50 60 keV depending on the medium whereas Rayleigh scattering is noticeable at low photon energies The attenuation effect is quantified by the Beer Lambert law which describes the exponential relation between the attenuated intensity after crossing the sample and the incident intensity considering a uniform material in terms of density and atomic number and a monochromatic incident x ray beam I I e 0 2 1 Where Io refers to the non attenuated x ray intensity 4 is the unit length attenuation coefficient of the traversed material and L is the length of material Since the slice of a human or animal body is not uniform and presents variable materials with different density values the overall attenuation characteristics are estimated from the measured line integra
83. ype Motion type Negative and positive limits Active unit flag Configured flag Maximum displacement Identification integer Displacement units counts radians degrees or distance units Current position Velocity Nominal velocity Acceleration Nominal acceleration Homing type Motion type Clockwise and counterclockwise limits Api status Full step enable Degrees per step Power in motion and standing Current Home ring offset VO mapping Table 8 Attributes techno_axis and api_axis classes In addition to the classes managing the properties of each axis a second set of low level classes api_functions and techno_functions provides the functionality for performing motion input output and configuration tasks The classes forming the different levels are described in greater detail in the following paragraphs The object of the station class on its creation manages the dynamic allocation of the memory space for the api_axis or techno_axis objects one per axis in the system The default number of axes is four user modifiable and they are arranged on an array so that the station object can easily address the appropriate axis for each action Each controller has a character identifier which is initialized to NULL until the configuration file is read The same happens with the active unit flag initialized to zero or the communication port whose default value is defined as a macro in the header of the station

Desarrollo Software de librerias para el control y la adquisicion de

Contents

Download Pdf Manuals

Related Search

Related Contents