User Manual
Contents
1. Figure 3 1 Heart Analysis Tool methods All OpenCV related operations are labelled with OpenCV in brackets Download the OpenCV library including the user manual at http sourceforge net projects opencvlibrary 3 3 Loading And Saving Sequences In the following section you can read all about how to save and open files and which formats are supported It is suggested to use multi page TIFF files On the one hand you only have one big file on the other hand the application is optimised to handle this format Another possibility is the use of JPG BMP PNG or TIFF files In that case the sequence is loaded from a set of images where you have to specify a source directory The source directory should only contain the images for the sequence with the file extension as noted above to avoid errors It is recommended to open at least a sequence with three images because most algorithms require three images Further the application was not tested with 3 3 Loading And Saving Sequences such a small sequence even if it is possible to open a single image Moreover the application tries to restore all primitives if a valid CSV file is available Same as loading you have the possibility to save the sequence into one big multi page TIFF or as a set of JPG BMP PNG or TIFF files but again only the multi page TIFF file is totally supported Furthermore if you save the sequence as a multi page TIFF an additional CSV file is created This CSV file2. IS OEE ae ae RE ets OB alle a aS 12 3 5 The Bottom MentiBal gis Sse a diy ae Adie a Bay SE SE 12 Bel AN iirc tds i vw Ee eee coat Sant I Sina ent fact es 12 30a Pause as ES eee eee tg ht aa eg 12 3 5 3 Display The Previous Frame 12 3 5 4 Display The Next frame a bor HA 12 3 5 5 Original Sige l Kaserne Ban ld 13 3 5 6 Stretch Fit Default x dig ie Bia teh sae cg ans 13 39 2 Single Vi ss Ol cae Bia Joni E ell oy eae Georg 13 3 5 8 Combined View Default 13 35 9 Statistical View as di is A 13 A A Meer ody A a 13 3 5 11 Refresh Rate Control 2 2 242 202 Pa 13 3 5 12 Time Area Label luis a Eo Sho Ee Ee 14 3 5 13 Position Slider 2224 05 an a re 14 3 6 The Side Bar Ze AA A Bi lee BE 14 3 6 1 Quick nfo Labels ut u ote a igh ae 14 3 0 2 Progress BAT os ee 15 3 6 3 Progress Label ct a A gen Bede eg 15 3 7 Various Instructions ars 8 ade eh an ar he 15 0 LONE A ker See Bra de ee 15 ASA WOVE sors ce er ee de E EA 15 IT Moving homie amp dws amp Adee e A eng Beh aS 15 Bord Threshold shaw tee lata Da ti te ee yes 15 3 7 9 Operations Mens una ba A Fat 15 OG lable Barc e o a oe AE we 16 37 7 Export Eine Prehle e whee EA EA RO A 16 ge Toggle Mask us nen SE AGA ar 16 Pay ac TACNA x 3a dl nes 16 4 Operations 19 AT PT a a A MS ke ner Ee Rei Seh tor 19 4 1 1 Detect Area st Ara a A a ate hg 19 4 1 2 Find Centre shit 2 Ble ey Ee oe Yb Spe oe 19 4 1 3 Detect Area Ell
3. OpenCV Read the OpenCV manual 3 2 for additional information 29 4 9 Edge Detection 4 8 3 Median OpenCV The median filter is very effective against single error pixel 4 8 4 Bilateral OpenCV This filter reduces the colour space Read the OpenCV manual 3 2 for additional information 4 9 Edge Detection In this section two edge detectors are described shortly 4 9 1 Canny This own implementation of the Canny edge detector has the advantage that it doesn t execute the thresholding with hysteresis for a better visibility of the strength of the corners 4 9 2 Canny OpenCV The Canny edge detector return good results for almost all tasks It uses non maximum suppression to eliminate multiple corners and double thresholding with hysteresis to trace weak corners 4 9 3 Sobel OpenCV Use this detector only for primitive tasks because it only returns good results on absolute corners 4 10 Morphologic With the consecutive execution of dilation and erosion you can simulate the morphological closing and opening 4 10 1 Dilate Morphological dilation for grey scale images 4 10 2 Erode Morphological erosion for grey scale images 30 4 11 Image Addition 4 11 Image Addition This approach was implemented to find a connection between the single frame and the sum of some or all images 4 11 1 Total Sum This method calculates the average of all frames 4 11 2 Total Sum Prepared This me
4. to enter a lower and an upper threshold All values between and equal are set to black all others to white 4 3 Masking Masking is very important especially to obviate error pixel or regions which are not previously of interest All necessary methods to perform this are described in the following section A mask is displayed in a blue transparent colour In most 23 4 3 Masking cases you select the Drawing Mode 3 4 9 for defining the mask Look out that the lines or curves of your region do not overlap 4 3 1 Create Mask First define a region with the Drawing Mode 3 4 9 and select Create Mask afterwards The selected area mask is displayed in a blue transparent colour 4 3 2 Cut Mask First Frame If you want to cut a special region also called ROI Region Of Interest you only have to surround your preferred area at the first frame and execute this command Consequently you only see your selected region all other pixel are set to white 4 3 3 Cut Mask All Frame Similar to the previous method you can define a region of interest but in this case you have to draw a region at all frames 4 3 4 Detect Background This method uses different algorithms to determine the background or static ele ments You can use this method followed by Cut Mask First Frame to automatic detect the region of interest 4 3 5 Blend Out It is sometimes useful to blend out the border With this function you can define the width of the bord
5. contains all data except for the sequence itself A detailed description of the structure of the CSV file follows in this section The following commands are listed in the file menu 3 3 1 Open File To open a file simply click on the first button of the Top Menu Bar 3 4 or select it via the top menu File Open File A file dialogue appears where you can choose the selected multi page TIFF file Remember that the file should contain at least three pages 3 3 2 Open Directory Use this command if you want to open a set of files from a directory A dialogue appears where you can select your source directory Please read the top section to avoid problems 3 3 3 Open File Resized Choose this operation either on large sequences or to speed up the processing time A dialogue appears where you can enter the new height of the image The new minimal height of the image have to be 64 pixel otherwise it is set to 64 pixel automatically If using images with a greater height than width it is possible that the resizing fails Depending on the aspect ratio of your image you has to select a large size than 64 to receive a minimal width of 64 pixel In case you receive a distorted image you should enable the Safe Mode 3 4 17 It is not suggested to set the new height greater than the original one as it does not cause better results This command only works on multi page TIFF files 3 3 4 Open Subsequence Use this command on very large sequen
6. edge image 4 7 7 Difference Image Difference image simply subtracts the previous frame of the current one The background is set to white for better visibility We take the absolute values of the subtraction to get non negative values This method is very sensitive to noise 4 7 8 Remove Background This method subtracts the average image of all images Normally the result is much better than the Difference Image variant Same as in the previous method the background is set to white and absolute values are taken 4 7 9 Remove Background signed The only difference between to Remove Background and Remove Background signed is that no absolute values are taken The background is set to grey 127 negative values are displayed darker positive values brighter 4 7 10 Remove Background EX Unlike the previous versions the current state of the heart decreasing or increas ing influences the result For this you have to calculate the correlation coefficient before and set the maximum extraction When the heart increases or decreases the previous and next frame is also used to determine movement pixel 4 8 Smoothing Smoothing is very important for processing images in general In most cases the Gauss filter is sufficient 4 8 1 Gauss OpenCV Using the sigma value you can specify the intensity of smoothing If you use a of 1 2 you receive a soft smoothing if you use a o of 4 0 you gain a very intensive smoothing 4 8 2 Blur
7. gt 3 institut fUr informatik Heart Analysis Tool User Manual Christian Premm 21 September 2009 Contents 3 4 Getting Started System Requirements 2 1 Minimal Requirements ats kd gp ee ae dt 2 2 Recommended Requirements 1 224 22a Han Beck Hin 2 3 Tested Hard and Software 04 2 ar ea Be Program Overview Sal Generaly Le Past derer 32 BEEN rar edea a wartete 3 3 Loading And Saving Sequences 20 4 3 3 1 3 3 2 3 3 3 3 3 4 3 3 5 3 3 6 3 3 7 Th Top Menu Bar ita a a is Bee as AS tg 3 4 1 3 4 2 3 4 3 3 4 4 3 4 5 3 4 6 3 4 7 3 4 8 3 4 9 3 4 10 3 4 11 3 4 12 3 4 13 3 4 14 3 4 15 3 4 16 en NN NNN Open Ele sr ee tas te Ra atte a to E See She sd E a ewok Hee Open Directory 2 eck ek aa Ma ee Open File Resized un 2 dod Gre ck re see ee ge Open Subsequence 2 2 2 22 2m nn Save EE Dive Hill dans de a ee ed eke aa Export Mack do autre Data e it a ue tele as OPEN ya ed a Do Gh eas de o tds TO she DAVES Ae ama as oe as e a a o oe 3G Draw Drawing Mode ia ti rd iia A Ae Meets nee Ap Ne a Shea aa hee hes oo Del Draw e ake S34 feu ti da PARES ar ee Be phy er ee EY Markers Cal a Re ee a Os a de EA EKUE ow eRe a Rs a arse Batt mee te op Be ee Stas naher Be DIZES Put ey MBAS zur une zn is O A de Duration 2 eg Er Bee ee ee ee EH oO OO OO OO AW AAWAAAONNNANNANNANNADAOOKR q A EH il SALT Settings e Sate ake Mace See ale as See 10
8. in um if using the marker Furthermore the size of any calculated area is displayed in um not in pixel Size specific values are displayed on the Statistical Window on the Log Window or on the Bottom Menu Bar 3 5 3 4 16 Duration With this button you are able to set the real duration of the sequence in ms After clicking the Duration button enter the real duration of the sequence in ms After assigning a duration to the sequence the elapsed time is displayed in the Bottom Menu Bar 3 5 3 4 17 Settings This button opens the Settings dialogue box which contains three tabs Each tab will be described in the following three sections General Tab In this tab you can set general options of the program e The first two labels displays the current default path for single and multi page TIFF files This path will be set on a successful loading of a sequence The path can not be changed e The next radio buttons give you the opportunity to start the application with the Single View 3 5 7 or Combined View 3 5 8 On slow systems or systems with an onboard video card it is suggested to start with the Single View For more details read the Trouble Shooting 6 To take effect you have to restart the application e If you set the check box maximise on start the application starts maximised To take effect you have to restart the application e Default the program shows the required time of every operation in brackets in the L
9. the image Keep in mind that some methods require a black background others a white background especially when using binary images 4 5 2 Normalise Image Normalise Image scales expands the whole sequence between 0 and 255 Nor mally you gain a better contrast after using this method 4 5 3 Remove Flickering On some sequences you notice a jitter because the brightness changes as result of the illumination Use this method to adjust the brightness automatically 4 5 4 Normalise Image EX Normalise Image EX scales each frame between 0 and 255 Since the scaling has been based on every single frame this method adjusts the brightness automati cally 26 4 6 Frequency 4 5 5 Remove Flickering Extended In this case the adjustment of the brightness is calculated from one of the four corners On some sequences you have to use this method because Remove Flick ering 4 5 3 does not lead to correct results The corner for the regularisation is selected via the correlation coefficient 4 6 Frequency In the next section frequency related operations will be discussed 4 6 1 Calc Correlation Coefficient This method calculates the correlation between the first and all other frames The result is displayed in the Statistical Window Furthermore the frequency and pulse of the heart is calculated and displayed in the Log Window If you want to compare the correlation coefficient with another except for the first frame select a different
10. the same order 7 o 282 299 i a r po se se elo fo y Mai Area Figure 3 3 Bottom Menu Bar 3 5 1 Play If you click on this button the playback of the sequence starts You can set the playback speed with the Refresh Rate Control 3 5 11 Normally the playback is stopped if you start an operation to decrease the progress time 3 5 2 Pause Click on the Pause button to stop the playback 3 5 3 Display The Previous Frame This button shows the previous frame 3 5 4 Display The Next frame This button shows the next Frame 12 3 5 The Bottom Menu Bar 3 5 5 Original Size 1 1 If you want to display the sequence with the original aspect ratio and the original size use this button 3 5 6 Stretch Fit Default This button fits the sequence into the window which looks best You can also use this button after resizing or moving the sequence to restore the view In some cases it is useful to have the correct aspect ratio or the original size For this case use the button described before 3 5 5 3 5 7 Single View The Single View consists of one window where the current sequence is shown 3 5 8 Combined View Default The Combined View is a combination of all views It consists of one window with the current sequence another window with the original sequence the Statistical Window and the Log It is suggested to use this view because it is the most comfortable one and you see all necessary information simultan
11. to the specific method is placed in brackets behind the keyword 2 System Requirements Due to time pressure it was not possible to test the program on all systems but it should work well with a video card with at least 64 MB of memory Especially when using a notebook or any other systems with a video card which uses shared memory you should read the Trouble Shooting 6 A fast CPU is advantageous 2 1 Minimal Requirements OS Windows XP SP1 CPU 1 5 GHz RAM 512 MB Hard Drive 5 0 MB or more free space Video Video Card with 64 MB of memory Input Keyboard and mouse 2 2 Recommended Requirements OS Windows XP SP3 CPU 3 0 GHz or more RAM 2048 MB or more Hard Drive 20 0 GB or more free space Video Video Card with 256 MB of memory Input Keyboard and mouse 2 3 Tested Hard and Software OS Windows XP SP1 SP2 SP3 CPU e AMD XP 2400 2 0 GHz e AMD 64 3200 2 0 GHz e AMD 64 3800 2 4 GHz e Intel Centrino 1 5 GHz 2 3 Tested Hard and Software e Intel DualCore T3400 2 16 GHz e Intel DualCore E6400 2 2 GHz e Intel DualCore E8400 3 0 GHz RAM 512 MB 2048 MB Hard Drive 10 0 GB or more free space Video e ATI Radeon X800 256 MB e ATI AMD Radeon HD4890 OC 1024 MB e Nvidia 4200 64 MB Nvidia FX 5900 ultra 256 MB e Nvidia 8800gt 512 MB e Intel GMA 4500 shared memory e Intel 855GME shared memory Input Keyboard and mouse The shared memory systems work fine with
12. using the Horn and Schunck algorithm Read the OpenCV manual 3 2 for additional information 4 7 2 Optical Flow LK OpenCV This method performs a calculation of the optical flow using the Lucas Kanade algorithm Read the OpenCV manual 3 2 for additional information 4 7 3 Optical Flow BM OpenCV This method performs a calculation of the optical flow using the block matching algorithm Read the OpenCV manual 3 2 for additional information 4 7 4 Segment Motion OpenCV This method uses the motion tracking algorithm of the OpenCV library It requires about 20 steps for initialisation Read the OpenCV manual 3 2 for additional information 4 7 5 Snake Image OpenCV Snake Image works similar as Find Correlation First you draw a polygon around the heart only at the first frame The edge points should be set on significant positions like corners or other eye catching spots Notice that the lines itself are ignored only the points are of interest The algorithm tries to fit the polygon at all other frames 4 7 6 Find Correlation First you have to draw a polygon around the heart only at the first frame The edge points should be set on significant positions like corners or other eye catching spots Notice that the lines itself are ignored only the points are of interest The 28 4 8 Smoothing algorithm tries to trace the significant points using the correlation coefficient You sometimes gain better results on an
13. D Exact Chan Vese 3D Chan Vese 3D Exact Chan Vese 2D Ellipse F1 Help F3 Stretch F4 1 1 F5 Single view F6 Combined view F7 Stat view F8 Log Space Toggle play pause Num Strech Num 1 1 Alt F4 Exit 32 5 Hotkeys Hotkey Command Alt S Settings Strg A Abort Strg C Copy Strg D Draw Strg S Draw Zero Strg M Marker Strg O Open Strg R Restore Strg S Save Strg X Exit Strg Y Redo Strg Z Undo Left Cursor Previous image Right Cursor Next image Entf Del Draw Note On some keyboards Strg is labelled as Ctrl and Entf as Del 33 6 Trouble Shooting In some cases systems with a video card with shared memory have problems to show the Combined View correctly In this case open the program settings and select Single View on the first page General tab To take effect you have to restart the program After restarting it the program displays the Single View instead of the Combined View default If you close the application it might happen that an error message appears which can be ignored It is recommended not to use the Combined View in this situation Another problem can appear on systems with an old video card which doesn t support non power of two textures In this case an error may be shown in the Log Window and the sequence will not be shown correctly or the window will may be kept blank Simply open t
14. Tracking Calculate Area Image Enhancement Invert Other Find Center Other Calc Ellipses Masking Use As Mask 22 4 2 Thresholding 4 2 Thresholding Thresholding is very important for isolating certain pixel In this section you get a brief introduction to various thresholding methods 4 2 1 Threshold To White All pixel with a value above the selected number are set to white all others remain constant 4 2 2 Threshold To Black All pixel with a value below the selected number are set to black all others remain constant 4 2 3 Threshold Binary All pixel with a value above the selected number are set to white all others to black You can also execute the binary threshold by double clicking in the Statistical View into the histogram 4 2 4 Threshold Auto This method allows you to determine the value of the threshold for a given per centage and performs a Binary Threshold 4 2 3 with the caculated value For example you estimate that 50 of your data is in range you enter 50 and recieve a threshold where approximate 50 of all pixel of the whole sequence are in range Because of the usage of discrete values it may happen that you recieve a threshold of 58 or 46 instead of 50 In the worst case you have to select another value since it is not possible to find a threshold with your chosen value 4 2 5 Double Threshold Use this threshold if you want to isolate a certain grey scale value You are able
15. all result of the operations are mentioned Used parameters are quoted in brackets 4 1 1 Detect Area This sequence detects the area of the heart without any pre knowledge or segmen tation It is very sensitive to noise which means that single error pixel influences the calculation of the area negatively Note Use this method only for testing purposes e Image Enhancement Remove Flickering e Motion Tracking Remove Background Thresholding Threshold Auto 2 Smoothing Median OpenCV 3 e Masking Convex Hull OpenCV 4 1 2 Find Centre This method determines the centre by calculating the average of all defined mo tion pixel Furthermore a calculation of the regression line is done to receive the orientation of the heart e Image Enhancement Remove Flickering e Motion Tracking Remove Background e Thresholding Threshold Binary 247 19 4 1 Auto Smoothing Median OpenCV 3 Other Find Center 4 1 3 Detect Area Ellipse With this sequence you fit an ellipse around the heart The size of the ellipses is determined by all counted motion pixel Use this method only for testing purposes Image Enhancement Remove Flickering Motion Tracking Remove Background Smoothing Gauss OpenCV 1 20 Thresholding Threshold Binary 251 Other Find Center Image Enhancement Invert Other Calculate Area Other Calc Ellipses 4 1 4 Detect Area Motion Simila
16. ar informs you about the overall progress whereas the right progress bar informs you about the temporary progress On simple operations both bars work identically 3 6 3 Progress Label The progress label displays the overall process in percent same as the left progress bar 3 7 Various Instructions 3 7 1 Zooming You can zoom in and zoom out by simply scrolling the mouse wheel You can reset the view by selecting Original Size 1 1 3 5 5 or Stretch Fit 3 5 6 3 7 2 Moving To move or drag the image or sequence hold the middle mouse button and move the mouse The Statistical View can t be moved 3 7 3 Moving Points After Drawing Mode 3 4 9 you can move or change the polygon or spline by holding down the left mouse button on the yellow points and move it You can stop moving the point by releasing the left mouse button You are able to move all yellow points connected to red lines except the centre point 3 7 4 Threshold To perform a Binary Threshold 4 2 3 double click on the desired position into the histogram of the current frame Statistical View 3 7 5 Operations Menu The most important menu can be called by clicking the right mouse button anywhere in the frame 15 3 7 Various Instructions 3 7 6 Status Bar The left part of the status bar informs you about the memory usage or a more detailed description of the selected operation The right part shows the current path of the sequence 3 7 7 Export Li
17. are Chan Vese vu ei de ee ee RE Beh ww 26 Image Enhancement tee heee a eee nn Hae 26 A nyert ig an oy teas Bak ee ta es Ee i ee AT oe g 26 102 Normalise Image ds La N ER ew 26 4 5 3 Remove Flickering a Bar ee aes BO AE 26 4 5 4 Normalise Image EX us Sana a ea 26 4 5 3 Remove Flickering Extended 44246 2 20202 27 Erequienit e Isa aa a alas a E Sok ATS ek eke 27 4 6 1 Calc Correlation Coefficient a cusco 26 102 Cale Luminarc ii es oea ea a o ea 27 4 6 3 Cross Correlation o oo oe ie asa 27 46A Sonograma A Ta ea Roe kB Dass DT Motion Tracking zu sr ir Ban ed Re he ho Ba 28 4 7 1 Optical Flow HS OpenCV 28 4 7 2 Optical Flow LK OpenCV Ls diia a a 28 4 7 3 Optical Flow BM OpenCV 28 4 7 4 Segment Motion OpenCV Vamos aoe ee 28 4 7 5 Snake Image OpenCV 28 4 7 6 Find Correlation e wet dt a 28 4 7 7 Difference Image oy oe engen dns 29 4 7 8 Remove Background AS 29 4 7 9 Remove Background signed 29 4 7 10 Remove Background EX ur Ge 2 au za 2 24 24 aS 29 ALS TOOL MING Ka do AAA Dale ent er ze pre 29 4 8 1 Gauss OpenCV atte AAA A 29 4 8 2 Blur OpenCV ek ae eS oe ee a Be 29 4 8 3 Median OpenCV za IAEA 30 4 8 4 Bilateral Op EV sr tiere a 30 49 Edge Detection e s sa ia eune rar Bar ara 30 AOL CAB a e re li eh re 30 4 9 2 Canny OpenCV Erbsen 30 4 9 3 Sobel OpenCV a A Les ar Ar a a si 30 4 10 Morphologie te 2 43 Ss
18. are disabled In the following sections the commands are listed as in the Top Menu Bar with the same name and the same order ik Copy S E S s Ops Auto Draw Draw0 Del Draw 2 4 a Undo Redo Edges Marker Extr Size Duration Settings Abort LJ Open Save Restore Figure 3 2 Top Menu Bar 3 4 1 Open Identical to 3 3 1 3 4 2 Save Identical to 3 3 5 3 4 The Top Menu Bar 3 4 3 Restore On Restore the original sequence is reloaded and all graphic primitives like lines and arrows are deleted As the original image is kept in memory this command performs a lot of faster than reloading the sequence via Open File 3 3 1 3 4 4 Copy As in many other applications you can copy the current view frame to the clipboard In case you have selected Combined Single and Statistical View an image of the current sequence or graph is copied to the clipboard After choosing the Log View the containing text is copied The image frame is copied to the clipboard as it is which means if you max imise the application you receive a higher resolution of the copied image On specific resolutions it can happen that a distorted image is copied to the clip board In this case resizing the application in small steps helps 3 4 5 Undo On Undo the last operation is undone The number of undoes depends on the number of buffered sequences You can change this number in the Settings 3 4 17 Default you
19. can restore one operation step Some data like the mask or midpoint can not be restored 3 4 6 Redo On Redo the last operation is redone The number of redoes depends on the num ber of buffered sequences You can change this number in the Settings 3 4 17 Default you can restore one operation step Some data like the mask or midpoint can not be restored 3 4 7 Ops This button opens a menu with all possible operations which can be applied to the sequence As it is the most important and most extensive part it is described separately in chapter 4 3 4 8 Auto Identical to 4 1 3 4 The Top Menu Bar 3 4 9 Draw Drawing Mode Select Draw if you want to mark or surround a region in the sequence You enter the drawing mode by clicking the Draw button Then simply click into the current sequence to set the points of the polygon or spline 3 4 12 You abort drawing with a double click Drawing is required for most Masking Operations 4 3 It is suggested that your lines of your polygon do not overlap otherwise some operations may fail Drawing can only be performed on the current sequence neither in the Statistical View and nor in the Log View On most operations only drawing at the first frame takes effect 3 4 10 Draw 0 This mode is very similar to the previous one 3 4 9 but you can playback 3 5 1 the sequence while drawing moreover the polygon or spline is only drawn on the first frame 3 4 11 Del Draw This comman
20. ces 2000 pages and more with a resolution of 480x256 to select only a subset of all images A dialogue appears where you see the total number of images and where you can select the range of the subsequence This command works only on multi page TIFF files 3 4 The Top Menu Bar 3 3 5 Save File Select this command to save the current sequence An additional CSV file sec tion ref with the same name will also be saved where you can find nearly all information like the values of the graphs or calculated velocity vectors You can also execute this command via the second button of the Top Menu Bar 3 4 3 3 6 Save Files Choose this command if you want to save a set of images files instead of one multi page TIFF file If the dialogue appears you can select the file format and the directory where the files are saved An additional CSV file will not be created 3 3 7 Export Mask If the required information is held in the mask you have the opportunity to save the mask as a multi page TIFF file 3 4 The Top Menu Bar In the Top Menu Bar you find almost all important commands If you highlight a button keep the mouse a short time over the button without moving a more detailed description appears The menu is very intuitive so you may ignore this section After loading a sequence The Top Menu Bar should look the same as in Figure 3 2 In case no sequence is loaded or the sequence is currently processed most buttons of the menu bar
21. d removes all graphic primitives of the current sequence 3 4 12 Edges This button allows you to toggle between polygons and splines smooth curves It is easier sometimes to start in polygon mode and switch to the spline mode to do the fine tuning afterwards 3 4 13 Marker For some operations Export Line Profile Size Sonogram and Calc Luminance you require a line as input To draw a single line enter the marking mode via the Marker button and click first on the start position and afterwards on the end position Accordingly you can select the preferred operation Furthermore you can use the marker to calculate the size of a region To receive the size in um not in pixel you have to set the size before 3 4 15 3 4 14 Extr With this button you can manually set a frame number with the maximal extrac tion of the heart or segment To do this select the frame with the slider of the Bottom Menu Bar 3 5 where the heart or segment seems to be largest and click the Extr button A red line appears at your selected position in the Statistical View 3 5 9 3 4 The Top Menu Bar If no frame with the maximal extraction is selected the first one is taken 3 4 15 Size To calculate the real area or a distance a reference value is required Use the Marker 3 4 13 to select a region from which you know the exact size If you click on the Size button a dialogue appears where you can enter the size in um After doing this you receive the size
22. eously On slow playback disabling the OpenGL text may help 3 5 9 Statistical View The Statistical View consists of one window containing the histogram with some graphs like the correlation coefficient 3 5 10 Log This view logs all operations with their parameters for either reconstructing your current progress easily or for reproducing the automatic calculations 3 5 11 Refresh Rate Control With this control you can adjust the refresh rate in ms The range is between 20 ms and 2000 ms To use refresh rates below 50 ms you require a fast system On slow playback disable the OpenGL text 3 4 17 13 3 6 The Side Bar 3 5 12 Time Area Label This label displays the time in ms and the area in um Before time and area are displayed you have to set the duration 3 4 16 and the area 3 4 15 3 5 13 Position Slider Use this control for sliding through the sequence or for going to a specific position 3 6 The Side Bar varse INR RR DGGE 5 3 Progress Figure 3 4 Side Bar 3 6 1 Quickinfo Label This label gives you information about the current mouse position and the grey scale value at this position In case of the Statistical View you get the value of the graph The shown RGB value does not provide you the real value in case of masks or for example red lines It only considers the colour information of the grey scale sequence 14 3 7 Various Instructions 3 6 2 Progress Bar The left progress b
23. er A a e 30 AA SL A EEE ea aa a ed Ge 30 ALOJA Erode Nah al tte ee a ee ib Er We 30 4 11 Image Addition aa 22a aa lt A Se ke a ee 31 AT Total Sun a der A E R E 31 4 11 2 Total Sum Prepared 22 8 as ae 31 4 11 3 Average Image d Sy ok poked eS OR E dis 31 A o 2 un hod oc Ae te Bae the Ye RAS eg Ag Make tle 31 4 12 1 Copy Image Brom e ade rs 31 AND A fhe a SS ES Er Soe See a ees es es 31 4 12 3 Calculate Ellipses aorta e ds aan dur ag 31 4 12 4 Calculate Area ada RISES GAD 31 5 Hotkeys 32 6 Trouble Shooting 34 1 Getting Started This user manual gives you a detailed overview of the Heart Analysis Tool It provides you with all necessary information how to use the program and it gives you a short introduction to the algorithms which have been used To start the program simply double click on the executable Heart Analysis Tool exe No further files or configurations are required If the program does not start or if it throws any error messages please have a look at the System Requirements 2 or Trouble Shooting 6 As the application needs no additional files and only 3 MB of disk space it can be sent via mail or it can be put on an USB device It can be transferred to other systems it does not need any installation and it can be even started from a network drive without problems In the following chapters keywords like methods operations and buttons are printed in capital letters In the majority of cases the reference
24. er in pixel for blending out If you disable soft corners you only receive a expanded border without blending You can also perform an inner and outer blending by disabling only outer mask 4 3 6 Convex Hull OpenCV As the name implies the function of the OpenCV library calculates the convex hull with the Slansky algorithm All values smaller than 255 white are used for creating the hull 4 3 7 Fit Line To Edges Use this method if you have sharp hard edges In that case draw a polygon near this corners and perform this operation afterwards to fit the polygon to the edges 24 4 4 Segmentation 4 3 8 Trace Borders Some operations return a mask as a result which is as storage intensive as the sequence itself Trace Borders gives you the ability to draw a line stipple contour around the mask automatically 4 3 9 Use As Mask Use this command to set the current image as mask and restore the original sequence All white pixel represent the mask 4 4 Segmentation In the following section you get a short introduction to some segmentation algo rithm In our tests the Chan Vese 3D algorithm have returned the best results 4 4 1 Chan Vese 3D This method performs a segmentation with the modified Chan Vese algorithm As input use a binary image The background has to be black 0 the foreground white 0 contrary to most other operations To save time set a value below 100 for the maximum number of iterations This method is
25. he program settings and select on the second page Program tab start in safe mode One major problem is the huge amount of memory usage of the program If the opened TIFF sequence is near on the limit of free memory it can lead to program termination In most cases you will receive an error message in the Log Window if you try to open the sequence which informs you that the sequence exceeds your free memory In that case you have two possibilities 1 Open the sequence resized 2 Open only a subsequence You can find both commands in the file menu In particular have a look at the Chan Vese 3D algorithm If the program crashes at the beginning of the operation try to reopen the sequence with the two suggestions mentioned before The application writes some settings to the Windows registry If any tool is blocking the access to the registry please add the Heart Analysis Tool to the exceptions If you want to delete the registry entry go to Start Run and type regedit into the prompt and delete the entry HKEY_CURRENT_USER Software Heart Analysis Tool It might happen that you get two versions of the program 1 Heart Analysis Tool exe 2 Heart Analysis Tool colour exe The second version supports the colour output This feature visualises contents with colour and it is currently in experimental state The colour visualisation was 34 6 Trouble Shooting removed from the first version because it takes more memory which
26. ipse 222mm nenn 20 4 1 4 Detect Area Motion nee eh u WUE 20 4 1 5 Chan Vese 2 ed toe te a le ld ER od oe 21 4 1 6 Chan Vese 2D Exact Be wa ket bod a 21 Gedo Ohan Vese gDr n aigan ye aa Ga a a ee Bee aE 21 4 1 8 Chan Vese 3D Exact a Er de 22 4 1 9 Chan Vese 2D Ellipse x cia 22 os ae Peg a E 22 4 2 4 3 4 4 4 5 4 6 4 7 HTeSHOIdine so ihe Skee e Mace See ees as oe 23 4 2 1 Threshold To White 2 24 2 2 ea ow a eile ER a 23 4 2 2 Threshold To Blacks em narra 26 Dare 26 aS 23 4 2 3 Threshold Binary 41 as ae de bi Glas dete amp Mase Ae ae 23 4124 Threshold Altes EA a A ee 23 4 2 5 Double Threshold 2 2 22 a ad 23 Maskil at he a Gs Bi e 23 4 3 1 Create Mask 2 0 a ahoa let ee ae ei 24 4 3 2 Cut Mask First Frame aus a Ad a a 24 4 3 3 Cut Mask All Frame 25 3 Sf 2 af ger 24 4 3 4 Detect Background 2 234 anf 2 228 he a oe ea 24 A Blend Out mess ete ger a he a ee a lie 24 4 3 6 Convex Hull OpenCV u 45 aero bya ee ae 24 4 330 EM Lite To Edgese ea ana 2 De ine 24 4 3 8 Trace Borders e Cedo oa eh Eke hg WG 25 4 3 9 Use As Mask tasa ada Sn Er Sr FAR ER 25 Segmentation s N denen p ee y Sk A ee 25 4 4 1 Chan Vese Bil 6 an er A a nz 25 4 4 2 Chan Vese 2D Visualisation 25 4 4 3 Chan Vese 2D a 2 a a ES ne a et 25 4 44 Meanshift Segmentation OpenCV 25 4 4 5 Pyramid Segmentation OpenCV 26 4 4 6 Detect Largest Segment o 26 4 4 7 Prep
27. is strongly required to open large sequences It is recommended to use the first version Whenever you have a question feel free to contact me premm christian chx2 com 35
28. ment of the foreground can be read at the specific operations at the Operations chapter If you need another background for a specific method simply use Invert 4 5 1 or press the i button on the keyboard The application works only with one byte unsigned char grey scale images As a result the range of integer values is from 0 to 255 3 2 OpenCV This application makes extensive use of the OpenCV library Open Source Com puter Vision library It is a very fast and comprehensive image processing tool box As most methods of the OpenCV library are very complicated we suggest to read the OpenCV manual This user manual doesn t describe most of those 3 3 Loading And Saving Sequences S Heart Analysis Tool 01 xj File Operations View Help N MM amp 3 E v o 9 4 o 0 amp o Open Save Restore Copy Undo Redo Ops Auto Draw DrawO DelDraw Edges Marker Extr Size Duration Settings Abort QuickInfo i x7 y 235 ii I R 22 G 22 ku 300 pages loaded Size 480 x 256 Path a D PROGRAMMING Infmath Projekt Daten Zebrafisch Heart Overlayed tif Memory usage 122MB peak 122MB Frequency gt Calc Correlation Coefficents position 6 5 T2 or sequence 12 4 beat duration 72 00frames 5 344sec y w a a fe 5 al Time su Y ib mo e a Sail s n D PROGRAMMING Infmath Projekt Daten Zebrafisch Heart Overlayed tif
29. mment or a more detailed description is arranged between the name and the value like Corr coeff SO to S4 represent the values of the Statistical View If the graph of the Statistical View is empty this section does not exist 10 to 13 show the value of the graphic primitives like lines or arrows For example if you want to evaluate the optical flow you require the arrow points of section 11 Pairs of values are separated by a comma frames usually by a semicolon Heart Analysis Tool 1 0 0 0 Number of images 49 Dimension width height 480 256 Path D Daten Zebrafisch New 29 1 2008 50 Frames csv Duration ms 0 Size um 0 Size marker pixel 0 Marker pos x1 y1 x2 y2 0 0 0 0 Max extraction frame number 0 Center x y 15 1 Information of the statistics view s0 Corr coeff 1 000000 0 973490 0 973490 0 979962 Internal data 10 Points frame number x0 y0 x1 y1 17 3 7 Various Instructions 11 Arrow points frame number x0 y0 x1 y1 18 4 Operations 4 1 Auto Automatic operations are a sequence of algorithms or operations with the ad vantage that you need no further knowledge of segmentation motion tracking etcetera To get an idea what the application causes simply open the sample sequence and perform the operations of this section As it is only a set of al gorithms you can read the details at the specific operations only the name the order and the over
30. ne Profile To export an intensity line profile select your desired region with the marker Afterwards select Export Line Profile from the top menu File Export Export Line Profile and select the path to save the CSV file The file contains all intensity values along a line of each frame A shortened example of the file follows below Frame 0 120 120 124 122 118 113 113 113 117 Frame 1 116 114 118 113 114 109 112 113 108 Frame 2 121 118 120 116 113 112 118 116 120 Frame 3 124 122 121 117 116 114 116 108 112 Frame 4 125 122 124 113 113 110 112 109 113 Frame 5 124 121 122 114 120 116 116 112 112 Frame 6 120 118 120 112 114 112 113 108 109 3 7 8 Toggle Mask In some cases you want to see or use the mask but it is actually hidden You can force to show or hide the mask by selecting View Toggle Mask from the top menu 3 7 9 The CSV File On saving the file as a multi page TIFF a CSV file with all necessary information is created An shortened example of the file follows below The first few lines represent the header of the file and contain various information to validate the file The proper data starts at the 12th line with Duration ms Interpreting the file should be straightforward First you have to locate the name of interest like Duration ms After a line break the real value is arranged In some cases either 16 3 7 Various Instructions a co
31. og Window To disable the output uncheck the show process time box 10 3 4 The Top Menu Bar e The last checkbox gives you the opportunity to load the sequence automat ically on start up This options only work on multi page TIFF files Program Tab Use this tab to configure program specific options To take effect you have to restart the application e In the first input field you can enter the buffer size It defines the number of possible undoes and redoes You receive the number of undoes and redoes by subtracting the input value with one In case of two default you can undo and redo one operation once except for sequences Use a value greater than two only on small sequences because the demand on memory grows enormously e Use start in save mode with an old video card which does not support power of two textures or OpenGL 2 0 It is also suggested to enable the safe mode if you work on small resolution sequences smaller than 128x128 pixel e Depending on your system you can increase the performance by disabling the OpenGL text If this checkbox is enabled you will not see any OpenGL texts as the minimum and maximum values of the Statistical Window This option increases the refresh rate in the Combined and Statistical View e The application uses smooth textures to render the sequence This feature has the effect that the sequence can be stretched or extended and it is still looking good If you disable it each pixel i
32. one with the max Extr button 3 4 14 4 6 2 Calc Luminance Before you can calculate the luminance you have to use the Marker Select a region overlapping the heart Calc Luminance determines the average brightness over all frames and displays the result in the Statistical Window 4 6 3 Cross Correlation Select the preferred area using the Drawing Mode Similar to Calc Correlation Coefficient the Cross Correlation calculates the correlation between all frames but in this case only the ROI is used An alternative is using Cut Mask All Frames to select your ROI and use Calc Correlation Coefficient afterwards It should lead to more detailed results 4 6 4 Sonogram First you have to select a region overlapping the heart with the marker If you select the Sonogram afterwards a small resizeable window with the sonogram is displayed 27 4 7 Motion Tracking 4 7 Motion Tracking In this section you find some algorithms like optical flow or segment motion to isolate the movement The optical flow is displayed with arrows Because of the different working of the three implementations you have to set the output size in pixel of the arrows Consequently you have to possibilities one is to take the average of the surrounded pixel the other is to display the optical flow of one pixel via the arrow Further this means you only have the possibility to or 4 7 1 Optical Flow HS OpenCV This method performs a calculation of the optical flow
33. out using the Combined View 3 5 8 3 Program Overview In the following few sections you get a detailed description of the whole applica tion its usage and appliance Primary the application was designed to calculate automatically the volume area of the heart of the zebrafish To do this you require a two dimensional video sequence with the pumping heart which formats are supported will be explained later Via the GUI you are able to resize and playback the whole se quence and perform various operations Since it has not been possible to calculate the area of the heart automatically so far the program provides you with many image manipulation and segmentation algorithms to calculate the area manu ally step by step Consequently the application can also be used to test various segmentation and motion tracking algorithms especially from the OpenCV li brary 3 2 If the application is executed the first time you receive a blank screen and no sequence is loaded This can be changed in the options menu to load the last recently used sequence automatically After loading a sequence and calculating the correlation coefficient the application should look like in Figure 3 1 3 1 General Some methods operate with a white background others with a black background Mathematically the background is black 0 The reason of either using a white or a black background is the better visibility of dark elements on a white background The require
34. r to Detect Area 4 1 1 this method uses no pre knowledge or enhanced segmentation but in this case the motion pixel are determined by an OpenCV algorithm Use this method only for testing purposes Image Enhancement Remove Flickering Motion Tracking Remove Background Image Enhancement Normalise Image Segmentation Segment Motion OpenCV Image Enhancement Normalise Image Morphologie Erode Thresholding Threshold Binary 249 Image Enhancement Invert Masking Convex Hull OpenCV 20 4 1 Auto 4 1 5 Chan Vese 2D This sequence detects the area of the heart by using the Chan Vese segmentation stops after 20 iterations It is very insensitive to noise and it produces smooth corners To get better results restrict the area with a mask at the first frame e Frequency Calc Correlation Coefficients e Image Enhancement Remove Flickering e Motion Tracking Remove Background e Thresholding Threshold Auto 10 e Image Enhancement Invert e Segmentation Chan Vese mu 2 00 11 0 80 12 1 20 vis 0 max it 20 e Segement Detect Largest Segment e Motion Tracking Calculate Area e Masking Use As Mask 4 1 6 Chan Vese 2D Exact It is the same sequence as Chan Vese 2D 4 1 5 but it stops either on saturation or after 1000 iterations 4 1 7 Chan Vese 3D This sequence detects the area of the heart by using the modified Chan Vese segmentation for three dimen
35. s displayed You can see the pixel as a square on zooming in e In the last checkbox you can decide whether you want to use a smooth blue or a normal silver progress bar It is strongly recommended not to use the smooth one because it decreases the performance of the application enormously Preprocessing Tab In this program you can choose between four operations which are executed automatically when loading a sequence e See Auto Normalise EX 4 5 4 e See Auto Remove Flickering 4 5 3 e See Auto Normalise 4 5 2 11 3 5 The Bottom Menu Bar e See Auto Correlation Coefficient 4 6 1 It is useful to enable the Auto Correlation Coefficient checkbox 3 4 18 Abort This button is enabled if any operation is in process If you abort the opera tion by clicking the button the progress stops immediately In most cases it is not possible to continue your work because the current sequence is distorted or incomplete In this cases restore the sequence 3 4 3 or use the Undo button 3 4 5 3 5 The Bottom Menu Bar In the Bottom Menu Bar you can find most view specific controls If you highlight a button keep the mouse a short time over the button without moving a more detailed description appears The menu is very intuitive so you can ignore this section After loading a sequence the bottom menu bar should look like as Figure 3 3 In the following sections the commands are listed like in the Top Menu Bar with the same name and
36. sional data stops after 20 iterations It is very insensitive to noise and it produces smooth corners Chan Vese 3D yields to the best results To get better results restrict the area with a mask at the first frame e Frequency Calc Correlation Coefficients e Image Enhancement Remove Flickering Motion Tracking Remove Background Thresholding Threshold Auto 10 Image Enhancement Invert 21 4 1 Auto Segmentation Chan Vese 3D mu 2 00 11 0 80 12 1 20 max it 20 Segement Detect Largest Segment Motion Tracking Calculate Area Masking Use As Mask 4 1 8 Chan Vese 3D Exact It is the same sequence as Chan Vese 3D ref but it stops either on saturation or after 1000 iterations 4 1 9 Chan Vese 2D Ellipse This sequence detects the area of the heart by using the Chan Vese segmentation stops after 20 iterations It is very insensitive to noise and it produces smooth corners This method uses the data calculated by the Chan Vese algorithm to determine the size and position to fit an ellipse around the heart To get better results restrict the area with a mask at the first frame Frequency Calc Correlation Coefficients Image Enhancement Remove Flickering Motion Tracking Remove Background Thresholding Threshold Auto 10 Image Enhancement Invert Segmentation Chan Vese mu 2 00 11 0 80 12 1 20 vis 0 max it 20 Segement Detect Largest Segment Motion
37. thod calculates the sum of all movement pixel and scales it 4 11 3 Average Image With this command you can sum a set of images and take the average of them It has a similar effect as Gauss smoothing Additional the sequence seems to run slower and the movement is better visible 4 12 Other In this section you find some methods without a special classification 4 12 1 Copy Image From Depending on the settings responsible for the buffer you have the possibility to use a previous sequence instead of the current one 4 12 2 Centre This method calculates the centre of all none white pixel It also calculates the regression line through a set of points 4 12 3 Calculate Ellipses If the area and the midpoint has already been calculated this method uses the information to fit an ellipse around the heart 4 12 4 Calculate Area If a mask is set this method calculates the number of all mask pixel otherwise the number of all non white pixel 31 5 Hotkeys an O a lt Command Abort Remove Background Calc Correlation Coefficient N pzz man Draw Remove Flickering Invert Create Mask Restore Normalise Image Sobel Set extraction Zero Draw Numpad 1 2 or Numpad 2 3 or Numpad 3 4 or Numpad 4 5 or Numpad 5 6 or Numpad 6 7 or Numpad 7 8 or Numpad 8 9 or Numpad 9 E o rm Detect Area Find Center Detect Area Ellipse Detect Area Motion Chan Vese 2D Chan Vese 2
38. very memory intensive so be careful you have enough free memory 4 4 2 Chan Vese 2D Visualisation It is similar to Chan Vese 2D 4 4 3 but in this case the actual progress is displayed in real time This operation performs slower than without visualisation 4 4 3 Chan Vese 2D This method performs a segmentation with the Chan Vese algorithm As an input use a binary image The background has to be black 0 the foreground white 0 contrary to most other operations To save time set a value below 100 for the maximum number of iterations 4 4 4 Meanshift Segmentation OpenCV For this segmentation use the original data so no preprocessing is necessary To get further information read the OpenCV manual 3 2 25 4 5 Image Enhancement 4 4 5 Pyramid Segmentation OpenCV For this segmentation use the original data so no preprocessing is necessary To get further information read the OpenCV manual 3 2 4 4 6 Detect Largest Segment This method detects the largest connected segment and closes its holes It can also be used for closing holes only The background has to be black 0 the foreground can be any colour except black 4 4 7 Prepare Chan Vese With this sequence you can prepare a binary image which is required for the Chan Vese segmentation 4 5 Image Enhancement In this section you find basic methods to improve the quality of the sequence except for invert 4 5 1 Invert With this operation you can invert
Download Pdf Manuals
Related Search