AngioQuant
Contents
1. 237 23 77 46 267 0 90 50 294 0 69 04 176 24 76 64 246 0 87 80 214 1 68 60 196 25 70 87 237 077 63 201 055 60 205 cooooumoou 2c Figure 2 A CSV file produced by AngioQuant opened in Microsoft Excel old the high threshold removal of edge tubules yes no and the prune size The last lines of the CSV file give the raw data that is the lengths and sizes of each tubule complex in each image as well as the number of junc tions in each tubule complex For each image each of these parameters lengths sizes junctions take up one column The data are sorted in a de scending order according to the lengths of the tubule complexes The data in each row correspond to the same tubule complex For example in the CSV file of Figure 2 the longest tubule complex in img1 bmp has the length 856 93 pixels its size is 3611 pixels and it has 16 junctions From the raw data it is possible to derive statistics describing the data Such statistics include the number of tubule complexes in the image the total and mean lengths and sizes of all tubule complexes in each image as well as the total number of junctions and the mean number of junc tions per tubule complex These seven statistics are given for each image in the CSV file above the raw data The same statistics are displayed on the AngioQuant window in the single image mode 4 Option Reference Below you can find information on all the elements of the AngioQuant graphical user in2. sizes and numbers of junctions Threshold slider Adjust the threshold in the low thresholding or high thresh olding image The slider is visible when the low thresholding or high thresholding image is displayed Threshold textbox Select the threshold in the low thresholding or high thresh olding image The textbox is visible when the low thresholding or high thresholding image is displayed Reset pushbutton Reset the low or high threshold to the automatically se lected value The pushbutton is visible when the low thresholding or high thresholding image is displayed 5 Function Reference Below you can find a brief description of all the functions m files of the AngioQuant Toolbox If you are using the AngioQuant Toolbox you can modify these functions to change the functionality of AngioQuant Note that the AngioQuant Toolbox is distributed under the GNU General Public License If you do not know what this means visit http www gnu org copyleft gpl html angioquant m The main function of AngioQuant Starts the graphical user interface bwthin m Perform thinning on a binary image using the method by Guo 1992 This is used as the skeletonization method in AngioQuant eqillum m Correct uneven illumination from a grayscale image by a poly nomial least squares surface fit fends m Find end points in a binary skeleton image fjunc m Find junction points in a binary skeleton image givehthresh m Give a threshold for a graysc
3. AngioQuant Antti Niemist antti niemisto tut fi v1 33 August 11 2005 Contents 1 Introduction 2 Installation 2 1 Standalone Version de 8 ahh ore a 2 1 1 Installing the MCR libraries e 43 4 99 ES 2 1 2 Installing AngioQuant a amp a amp as oh a oe Bee oR ues 21 9 Starting UD ar voce eon e kW A ee ADR 22 MATLAB toolbox version cres 2 2 1 Installing the AngioQuant Toolbox 222 Stating UD oen a y oe oboe 9d dca E Sd QULA E S d 3 Usage 3 1 Image quality requirements llle 3 2 Aarti CAON ex qox om uer deed Je Av deed Aer df 3 2 1 Opening animage ccce xoa e ou eoe Da Oe do 32 2 V SMON ru cuu miu eue iuc awe C BW OA Segmentation ar rrari kE E RN a RE ACA 324 JSkeletonization 4 cx zc ER RA X RU o DOuLD PRUNING 12379 Ve X cae eh aoe b eoa iub odi 3 2 6 Single step quantification i935 ere bx D Saving the results T g ren nanya a ea ua YS 3 3 Batch pro ssig ssis aout y os SpA ed wou ALY wad Wop AC dd 3 3 1 Preparing and starting a batch 3 3 2 Parameter selection 1 254 a c4 3 3 3 Saving the results aoaaa 390 99093 3 4 Viewing the results aaau aaa 3 4 1 Visualization in AngioQuant 3 4 2 Reading CSV hless v sts 42 ee eea a ine oaa a 4 Option Reference 4 1 Menuitems 0 00 00 arre es u arnau ero 242 Ter tems oten boe eho he ah X0 otk e KE ook o NO 5 Function Reference 11 11 12 13 1 Introduction An
4. age see Section 3 2 2 The higher the value the stronger the smooth ing is The default value is 1 which means that no smoothing is done AngioQuant also asks whether the tubule complexes that touch the edges of the image should be removed Removing these complexes would be mo tivated by the fact that measuring lengths of tubule complexes that are not completely seen in the image introduces a bias towards small complexes However we recommend not to remove these complexes because it too may cause a bias towards small complexes This is because large complexes are more likely to touch the edges of the image than small complexes In the extreme case the image might only contain large complexes that all touch 8 the edges of the image Removing all these large complexes would clearly result in a false quantification As the last parameter AngioQuant asks the user to give a prune size see Section 3 2 5 The choice of this value should in general be based on experience in analyzing similar images but typically a value around the default 10 pixels is suitable Finally AngioQuant asks whether you would like to give descriptive codes for the different images If you choose yes AngioQuant prompts for the codes one by one and shows each respective image on the AngioQuant window The images will be sorted in the CSV file alphabetically according to the codes that you give If you choose not to give descriptive codes the images will be sorted a
5. ale image such that a given percentage of the darker unique graylevels will be marked as objects This function is used to find the low threshold in AngioQuant hthtesh m Threshold a grayscale image by marking a given percentage of the darker unique graylevels as objects This is used as the low thresh olding method in AngioQuant 13 hystthresh m Apply hysteresis thresholding for segmentation of an image This function is not used in AngioQuant It may be used to replace segmcombine m othresh m Threshold a grayscale image using the method by Otsu 1979 This is used as the high thresholding method in AngioQuant rmbranch m Remove short branches and small objects from a skeleton im age This is used as the pruning method in AngioQuant rmedge m Remove objects that touch the boundaries of a binary image segmcombine m Apply a variation of hysteresis thresholding for segmen tation of an image This is used as the segmentation method in AngioQuant skellen m Calculate the length of a skeleton taking into account that the distance between two diagonally connected pixels is not 1 but square root of 2 superimpimg m Create superimposed image from a grayscale image and red green and blue masks This function is used for visualization in AngioQuant thresh m Threshold a grayscale image using a given threshold 14
6. art the m file angioquant 3 Usage All the information in this section applies both for the standalone as well as the toolbox version of AngioQuant 3 1 Image quality requirements To get the best quality results from AngioQuant you need to make sure that your images are of high quality First the imaging should be done soon after the angiogenesis experiment is finished in the laboratory This makes sure that the staining is strong enough to provide a sufficient contrast between the tubules and the rest of the image Second it is important to make sure that the whole image area is well in focus to avoid blurring of the tubules Finally it is good to have the whole image evenly illuminated Even though in most cases AngioQuant is able to correct uneven illumination an image that is evenly illuminated is likely to provide more accurate results than an unevenly illuminated image It is important to realize that if your images are of low quality you are unlikely to get satisfactory quantification results The imaging stage is there fore crucial in making sure that the quantification is accurate You also need to make sure that you save your images in a file format that AngioQuant is able to open The supported file formats are e Windows Bitmap BMP Tagged Image File Format TIF TIFF JPEG Image JPG JPEG Compuserve Graphics Interchange Format GIF PC Paintbrush PCX Portable Graymap File Format PGM Portable Network Graphic
7. ays lost when the image is saved in the JPEG format When the skeleton or the pruned skeleton image is displayed on the AngioQuant window the respective lengths and sizes of the tubules as well as the numbers of junctions branching points in the tubules can be saved in a comma separated values CSV file by choosing Save Data from the File menu The CSV file can then be opened in a spreadsheet program such as Microsoft Excel For information on interpreting the CSV file opened in a spreadsheet program see Section 3 4 2 The default CSV separator is the comma However in some systems the separator must be the semicolon in order for the spreadsheet program to be able to read the CSV file correctly If the CSV file does not look okay when you open it in your spreadsheet program change the CSV separator 7 by choosing Separator Semicolon from the CSV menu and save the CSV file again If you are using AngioQuant under the Windows operating system you cannot save the image or CSV file over a file that is currently opened in another application such as Excel 3 3 Batch processing 3 3 1 Preparing and starting a batch The batch mode allows you to analyze a large number of images without the need to open each image separately in AngioQuant Before starting the batch you need to put all the images that you wish to include in the batch into the same directory on your hard drive Note that a batch cannot include images that are located o
8. ccording to the filenames 3 3 8 Saving the results Once all the parameters have been set AngioQuant analyzes the images without any user intervention Please note that running a batch containing a large number of images may take several hours If you have a lot of images you may want to consider running the batch overnight Once the batch is ready to be saved a popup window appears with the text Batch processed successfully Press the OK pushbutton to move on to the save dialog Use the save dialog to select the name and location of the comma separated values CSV file The CSV file will contain the quantifi cation data on all the images in the batch It can be opened in a spreadsheet program such as Microsoft Excel For more information on interpreting the CSV file opened in a spreadsheet program see Section 3 4 2 AngioQuant also saves the original images with the skeleton overlaid in red in a subdirectory skel in your batch directory You can use these images to assess the quality of the obtained results If the skeleton is not consistent with the underlying image the quantification results are not re liable and the analysis should be done again using modified parameters Note that with high resolution images the displayed skeleton may seem discontinuous even when it really is not This is due to the limited resolu tion of computer screens that is the screen cannot display all the data that the image contains You should zoom
9. ectory e g on the desktop 2 Copy the file MCRInstaller exe into the directory that you created 3 Run the file MCRInstaller exe Four files are extracted into the temporary directory 4 One of the extracted files is Setup exe Run this file and follow the instructions 2 1 2 Installing AngioQuant The AngioQuant software can be downloaded from the AngioQuant home page at http www cs tut fi sgn csb angioquant The name of the file to be downloaded is AngioQuant zip Installation steps 1 Create an empty directory for AngioQuant 2 Extract the contents of the zip file AngioQuant zipinto the directory that you created 2 1 3 Starting up AngioQuant is started by double clicking on the file AngioQuant exe Be fore trying to start AngioQuant make sure that you have installed the MCR libraries see Section 2 1 1 of this manual 2 2 MATLAB toolbox version The MATLAB toolbox version of AngioQuant AngioQuant Toolbox re quires MATLAB 6 5 R13 or higher with the Image Processing Toolbox The toolbox version can be run on any platform for which MATLAB is available The toolbox version is mainly intended for advanced users 2 2 1 Installing the AngioQuant Toolbox The AngioQuant Toolbox is installed by extracting the contents of the zip file AngioQuant Toolbox zip into an empty directory 2 2 2 Starting up 1 Start MATLAB 2 Go to the directory in which the AngioQuant Toolbox was installed 3 St
10. el size parameter a high value results in heavier smooth ing If the size of the kernel is one default smoothing does not have any effect on the original image 3 2 3 Segmentation To detect and quantify the tubules you must create a binary image based on the original smoothed image This is done by pressing the Segment pushbutton It is now important to check that the binary segmentation re sult overlaid in red on top of the original image accurately represents the tubules The quality of the eventual results is dependent on the accuracy of the segmentation result and segmentation is therefore the most crucial step of the overall quantification procedure If the segmentation result is unsatisfactory a better result may be ob tained by repeating the segmentation step with modified parameters First you need to use the popup menu in the top right corner of the AngioQuant window to display the Low thresholding image A slider now appears below the image Use this slider to make sure that at least a part of each connected tubule complex is overlaid with the red color while at the same time none of the fibroblast or background areas are overlaid with red Sec ond you need to use the popup menu to display the High thresholding image Now use the slider below the image to make sure that all the tubules are completely overlaid with the red color It is likely that simultaneously some undesired fibroblast or background areas get overlaid wi
11. esult in a false quantifi cation 3 2 4 Skeletonization In order to be able to measure the lengths of tubules they must be re duced to arcs that are one pixel in width This is done by pressing the Skeletonize pushbutton The resulting skeleton is displayed overlaid in red on the original image Tubule junctions branching points are dis played as green dots If you do not want to display the junctions use the Show junctions checkbox Checking or unchecking this checkbox does not have an effect on the quantification results rather it is just a visualiza tion option With high resolution images the displayed skeleton may seem discon tinuous even when it really is not This is due to the limited resolution of computer screens that is the screen cannot display all the data that the im age contains You can zoom in to check for continuity by using the zoom tool First press the Zoom tool pushbutton and then left click on the im age at the region that you want to zoom in Clicking again with the left mouse button results in further zooming in To zoom out click with the right mouse button 3 2 5 Pruning The skeleton that is obtained in the skeletonization step usually contains spurious short arcs that split from the main lines and are not consistent with the tubules of the original image These spurious arcs are removed using the Prune pushbutton It is important to select an appropriate Prune size This parameter defines the maxim
12. gioQuant is an automated image analysis tool for quantification of an giogenesis It is designed for in vitro angiogenesis assays that are based on co culturing endothelial cells with fibroblasts AngioQuant uses networks of connected tubules as a basic unit in the quantification These networks are called tubule complexes AngioQuant provides quantitative and repeat able measurements of the lengths and sizes of tubule complexes as well as the numbers of junctions branching points in the tubule complexes The resulting quantification data are saved in the comma separated values CSV format The CSV files can be easily opened in a spreadsheet program such as Microsoft Excel A snapshot of the AngioQuant software can be seen in Figure 1 In in vitro angiogenesis assays that are based on co culturing endothelial cells with fibroblasts endothelial cells are stained so that the tubules can be detected However there is usually also some background staining of fibroblasts AngioQuant has been specifically designed to detect only the tubules and not the more lightly stained fibroblasts This manual contains detailed information on the installation and use of AngioQuant For technical details of the used image processing methods please consult the manuscript Antti Niemist Valerie Dunmire Olli Yli Harja Wei Zhang and Ilya Shmulevich Robust quantification of in vitro angiogene sis through image analysis IEEE Transactions on Medical Image P
13. in to check for continuity If you are using AngioQuant under the Windows operating system you cannot save the images or CSV file over a file that is currently opened in another application such as Excel 3 4 Viewing the results 3 4 1 Visualization in AngioQuant In the single image mode when the quantification results are ready after the skeletonization or pruning stage AngioQuant displays the most important summary statistics on the main window These statistics are number of tubule complexes e total length of tubule complexes e mean length of tubule complexes 9 total size of tubule complexes e mean size of tubule complexes total number of junctions and e mean number of junctions per tubule complex The lengths and the sizes of tubule complex are measured in pixels The lengths are measured from the skeleton of the tubules whereas the size is measured from the segmentation result The size measurement thus takes into account the thickness of the tubules and the sizes are therefore larger than the lengths It is also possible to view the quantification results as histograms of the tubule lengths sizes and numbers of junctions The histograms are shown by pressing the Show histograms pushbutton It is also possible to zoom in the image displayed on the AngioQuant window First press the Zoom tool pushbutton and then left click on the image at the region that you want to zoom in Clicking again with the left mouse b
14. n a CD or other read only media The batch is started by selecting Process Batch from the File menu 3 3 2 Parameter selection The user makes a number of parameter selections before AngioQuant starts analyzing the images Sometimes it may be difficult to know which choices you should make In fact it is often the case that the single image mode is used with some representative images in order to find the suitable param eters The batch mode is then used to process a large number of images at once First AngioQuant asks if the optimal values for the low and high thresh olds see Section 3 2 3 should be found separately for each image The al ternative option is to use the same thresholds for all images However it is recommended to let AngioQuant find the optimal thresholds for each image separately because not only may the illumination differ between the images in the batch but under different treatments the intensity of the tubules may vary as well If you chose to use the same low and high thresholds for all the images AngioQuant may next ask if the parameters used with last opened image or the last saved batch should be used If you answer no to both of these ques tions AngioQuant next asks whether the low and high thresholds should be selected automatically or manually If you opt for manual selection of the thresholds AngioQuant next prompts for them AngioQuant then prompts for the kernel size for smoothing of the origi nal im
15. rocessing in press 2 Installation AngioQuant is available as a standalone version and as a MATLAB tool box Most users should use the standalone version However users who have an installation of MATLAB version 6 5 or higher may want to use the toolbox version MATLAB is a registered trademark of The MathWorks Inc http www mathworks com 2 4 Standalone version The standalone version of AngioQuant is available for Microsoft Windows based systems Most users should choose to install the standalone version of AngioQuant 2 1 1 Installing the MCR libraries The MCR libraries must be installed before installing AngioQuant The MCR libraries are available for download on the AngioQuant homepage athttp www cs tut fi sgn csb angioquant The name of the file to be downloaded is MCRInstaller exe The copyright of the MCR libraries is held by The MathWorks Inc 2 img1 bmp AngioQuant High thresholding bd Kernel size Smooth 1 Segment C Remove edge tubules Skeletonize Show junctions Prune size 10 Prune Quantify Results Number of tubule complexes Total length of tubule complexes Mean length of tubule complexes Total size of tubule complexes Mean size of tubule complexes Total number of junctions Mean number of junctions how histograms Figure 1 A snapshot of the AngioQuant software Installation steps 1 Create a temporary dir
16. s PNG Portable Pixmap File Format PPM The use of the JPEG images is not recommended because JPEG uses lossy compression In other words some fine details of the images are al ways lost when the image is saved in the JPEG format This may cause the results obtained with AngioQuant to be sub optimal If your images are in JPEG format converting to another format before using AngioQuant does not help 3 2 Quantification Please note that many of the quantification steps described in this section are computationally very intensive Some of the steps may take from couple of seconds to several minutes to complete The processing time depends strongly on the size of the image being processed as well as the computing power of the computer used We recommend to use a computer with an Intel Pentium 2 0 Ghz processor or equivalent 3 2 1 Opening an image To open an image in the single image mode choose Open from the File menu Choose the image that you want to process using the dialog that opens The image is then displayed on the AngioQuant window If you open a color image it is converted into a grayscale image 3 2 2 Smoothing Smoothing is an optional step that in most cases does not have to be per formed However in some cases smoothing an original image helps to cre ate a good binary image in the segmentation step Section 3 2 3 Smoothing is done using the Snooth pushbutton Smoothing is controlled by the user defined Kern
17. terface This section serves as a quick reference for users that are already familiar with AngioQuant 4 1 Menu items File Open the File menu Open Open an image to be processed in the single image mode see Section 3 2 1 Save Image Save the currently displayed image on the disk see Sec tion 3 2 7 Save Data Save the quantification data related to the currently dis played image see Section 3 2 7 11 Process Batch Quantify a batch of images see Section 3 3 Exit Exit AngioQuant CSV Open the CSV menu Separator Comma Select the comma as the separator in the CSV files saved by AngioQuant Separator Semicolon Select the semicolon as the separator in the CSV files saved by AngioQuant Help Open the Help menu About Display general information on AngioQuant 4 2 Other items Select image popup menu Select the displayed image As you proceed with the quantification steps more images corresponding to the different steps become available in the popup menu You can use the popup menu to go back to an earlier step of the quantification process Smooth pushbutton Smooth the original image see Section 3 2 2 The pushbutton is enabled when the original image is displayed Kernel size textbox Select the size of the kernel to be used in smoothing the image The higher the value the heavier the smoothing Segment pushbutton Segment the image to recognize the tubule areas see Section 3 2 3 The pushbutton is enabled
18. th red but it does not matter as long as low thresholding was done correctly Finally press the Segment pushbutton to perform segmentation with the modified parameters Again remember to check that the segmentation result accu rately represents the tubules If the result is still unsatisfactory go back to fine tune low and high thresholding The high thresholding process can be thought of as selecting all poten tial tubule complexes The selection of the actual tubule complexes is done in the low thresholding process The overlaid read areas in the low thresh olding result can be thought of as markers That is the idea is to put a red mark on all tubule complexes and all those potential tubule complexes that got marked are selected as actual tubule complexes At this point you can also choose to remove tubules that touch the edges of the image This is done using the Remove edge tubules checkbox Removing these complexes would be motivated by the fact that measur ing lengths of tubule complexes that are not completely seen in the image introduces a bias towards small complexes However we recommend not to remove these complexes because it too may cause a bias towards small complexes This is because large complexes are more likely to touch the edges of the image than small complexes In the extreme case the image might only contain large complexes that all touch the edges of the image Removing all these large complexes would clearly r
19. the CSV file contain the descriptive codes of the im ages if applicable and the names of the image files If the descriptive codes were given the results for different images of the batch are sorted alphabeti cally according to the codes Otherwise the results are sorted alphabetically according to the filenames The next five lines of the CSV file contain the parameters that were used to analyze each image These parameters are the kernel size the low thresh 10 Fal bate D E F G H 1 J 1 code neg control pos control treatment 2 filename imgl bmp img2 bmp img3 bmp 3 4 kernel size not used not used not used 5 low threshold 119 113 116 B high threshold 138 131 141 7 remove edge tubules No No No 8 prune size 10 10 10 9 10 number of complexes 26 39 34 11 lengths sizes junctions lengths sizes junctions lengths sizes junctions 12 total 2486 70 9385 24 3210 06 11221 29 2110 62 6669 13 mean 95 64 360 96 0 92 82 31 287 72 0 74 62 08 196 15 0 38 14 data see below see below see below see below see below see below see below see below see below 15 856 93 3611 16 503 73 1902 11 194 78 271 32 785 2 474 69 1527 5 189 95 628 17 151 31 656 3 399 62 1621 7 173 71 668 18 138 12 424 2193 21 902 2 156 99 517 SH 19 110 20 405 0 153 60 455 0 127 44 423 20 98 64 266 1144 05 504 0 127 23 417 21 87 88 379 0 114 04 36 1 120 95 571 22 77 74 386 0 102 33 580 190 50
20. um length of a spurious arc The default value is 10 which means that all arcs that have 10 pixels or less are removed The prune size should be chosen in such a way that the pruning step results in a clean skeleton that accurately follows the topology of the tubules 3 2 6 Single step quantification The smoothing segmentation skeletonization and pruning steps of Sec tions 3 2 2 through 3 2 5 can be done as a single step using the Quantify pushbutton The Kernel size and Prune size parameters as well as the status of the Remove edge tubules checkbox have the appropriate effect on the quantification Single step quantification is useful if you are able to rely on the automated selection of the thresholds in the segmenta tion stage Note that single step quantification always performs the quantification starting from the original image and using automated selection of thresh olds If you have modified the thresholds and then press Quantify the modified thresholds are ignored 3 2 7 Saving the results At any time the image displayed on the AngioQuant window can be saved by choosing Save Image from the File menu The supported file formats are e Windows Bitmap BMP Tagged Image File Format TIF TIFF JPEG Image JPG JPEG Portable Network Graphics PNG Portable Pixmap File Format PPM Saving images in the JPEG format is not recommended because JPEG uses lossy compression In other words some fine details of the images are alw
21. utton results in further zooming in To zoom out click with the right mouse button The displayed image can also be viewed in an external window To do this press the External image pushbutton You can have as many such external windows as you like Viewing the images in external windows can be useful because then you can display at the same time two or more im ages that you want to compare You can also resize these external windows Zooming is possible by selecting the appropriate tools from the toolbar of the external window 3 4 2 Reading CSV files As described above the quantification results are saved in a comma sepa rated values CSV file in the single image as well as the batch mode CSV files are a standard format for saving data in a text file format They do not appear evenly spaced in a text file but they can be interpreted easily by third party software If you have a spreadsheet program such as Microsoft Excel installed on your system you can load the CSV files into the spread sheet You will be able to save the spreadsheet from Excel in its own internal format if required A part of a CSV file that was saved in the batch mode of AngioQuant is shown in Figure 2 Note that the quantification results for each image take up four columns CSV files that are saved in the single image mode of AngioQuant are similar The only difference is that naturally then the CSV file contains data for only one image The first two lines of
22. when the original smoothed low thresholding or high thresholding image is displayed Remove edge tubules checkbox Select whether the tubules that touch the edges of the image are included in the quantification or not For rec ommendations see Section 3 2 3 The checkbox is disabled when the skeleton or the pruned skeleton is displayed Skeletonize pushbutton Skeletonize the segmented image see Section 3 2 4 The pushbutton is enabled when the segmentation image is displayed Show junctions checkbox Select whether the junctions of the tubules are shown as green squares on the skeleton The checkbox is enabled when the skeleton or the pruned skeleton is displayed Prune pushbutton Prune the skeleton see Section 3 2 5 The pushbutton is enabled when the unpruned skeleton is displayed Prune size textbox Select the maximum size in pixels of tubules to be re moved in the pruning step Quantify pushbutton Perform all the quantification steps as a single step see Section 3 2 6 12 Zoom tool pushbutton Zoom in on the displayed image After pressing the pushbutton left click on the image at the region that you want to zoom in Clicking again with the left mouse button results in further zooming in To zoom out click with the right mouse button External image pushbutton Show the currently displayed image in an ex ternal window Show histograms pushbutton View the quantification results as histograms of the tubule lengths
Download Pdf Manuals
Related Search
AngioQuant angioquant angioquant software