Photographing & Segmenting Muscle Fibers in Stained EOM Crossections! JM Miller, PhD
Eidactics, San Francisco! Quantitative histology of extraocular muscles (EOMs) conventionally uses sampling to avoid having to count, or worse, measure the thousands of fibers in each crossectional slice. Unfortunately this tissue is highly inhomogeneous, and unbiased sampling is difficult. We address this problem by segmenting essentially all the muscle fibers in stained EOM slices and calculating their individual areas, excluding voids, connective tissues, blood vessels, and nerves. A completely automated process probably cannot do this, and it is usually impractical to delineate tens or hundreds of thousands of fibers by hand." Muscle fibers appear in crossection as small roundish objects, ideally stained uniform red. Paraffin embedded samples fibers shrink, creating surrounding voids, making segmentations easy. Elsewhere, contrasting connective tissue separates muscle fibers. But some segmentations are difficult, where fibers are tightly packed, or poorly stained. In many of the former cases we will need to break such groups into individual fibers manually. Conversely, a single fiber may be split into fragments by histological processing, causing automatic methods to find multiple fibers where there is only one. We will need to merge these fragments manually to get a correct, unbiased result. Some fibers will simply be missed by automatic processing."
We therefore propose a combination of interactive and automated processes, including:" • Manual selection of Regions of Interest (ROIs) or masking to avoid or remove non-muscle structures that would be wrongly identified as muscle fibers in subsequent automatic processing." • Manual and automatic filtering to clean up images (eg, mathematical morphology operations)," • Separation of tightly-packed fibers using a Watershed algorithm." • Background removal by intensity and shape-based filtering, where thresholds can be set interactively while assessing their overall effect on the image." • Tools to find and manually fix residual segmentation errors." • Calculation of individual fiber areas to a “live” spreadsheet, where suspect entries can be selected to view corresponding image regions for possible editing. A "roundness" index (eg, the ratio of min and max dimensions of segmented objects), and variance measures would help find segmentation errors.
2014 July 4 Friday
Page 2! of 12 !
Photographing & Segmenting Muscle Fibers
System Requirements! • Media Cybernetics Image-Pro Premier 9.0.4 or better (IPP-9) with the Grains app." • Windows 7, 64-bit (which we run under Apple Bootcamp)"
Histological Processing! Mason Trichrome stain was chosen to distinguish muscle from other tissue components. We pin each muscle to a popsicle stick approximating normal length and width, and leave it in 150 ml of Z-Fix (from Anatech Ltd) for 4 days. We then transfer pinned samples to small vials filled with fresh Z-Fix, and send them for histological processing." The most troublesome result of poor histological processing is fiber cracking, referred to as "parched earth" artifact, a sign that the tissue has been over-processed (excessive dehydration or clearing). During segmentation, this results in fragments being identified as fibers. Often, these errors must be repaired manually. It is essential to minimize cracking, although it would be useful to develop a version of the Grains app (see below) that avoided following cracks by preferring contours without high or sudden increases in curvature."
Photographing Masson's Trichrome Stained Slides! Cleaning Slides! (1) Blow slide clean with compressed air." (2) Apply a drop of lens cleaning solution to each side." (3) Tear off a fresh piece of lens cleaning paper and gently wipe."
Leitz Microscope Set-up & Calibration! (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Place the Calibration Slide on the stage, centered on the “0.01 mm” pattern." Gently tap the slide from the edge and top to be sure it’s squarely positioned and flat on the stage." Select the x16 objective for x20 total magnification to the camera." Swing the condenser in (for a smaller, more intense illumination spot)." Set illumination power to 12 volts." Set iris diaphragm below condenser to "6"." Set the focus of both eyepieces to the faint black ring." Adjust the interocular distance for binocular viewing." Adjust fine and coarse focus knobs until the scale on the slide is clear and sharp." Viewing through one eye at a time, adjust each eyepiece for best focus, thereby compensating for differences in the optical power of your two eyes."
Viewing Slides! (1) Place a slide on the stage, gently tapping from the edge and top to be sure it’s squarely positioned and flat on the stage." (2) Select an objective to view at x3.125 or x20 magnification" (3) Swing condenser out for x3.125 mag, or in for x20 mag" (4) Set illumination power at 9 volts for x3.125 mag, or 12 volts for x20 mag" (5) Set iris diaphragm below condenser at "6"" (6) Adjust fine and coarse focus knobs until the slide is clear and sharp."
2014 July 4 Friday
Page 3! of 12 !
Photographing & Segmenting Muscle Fibers
Canon EOS 10D Camera Settings - Masson’s Trichrome! (1) 1/4000 Shutter Speed" • Turn the Main Dial, next to the Shutter Button, to adjust shutter speed" • See EOS 10D Instruction Manual, pg 86" (2) 1600 ISO Film Speed" • Press “Drive-ISO” button on top of camera, near small LCD Panel" • Adjust ISO Film Speed with Quick Control Dial, near large LCD Monitor" • See EOS 10D Instruction Manual, pg 149" (3) 4500 [K]elvin for Color Balance" • Press Menu Button next to large LCD Monitor" • Use Quick Control Dial to scroll to Color Temp" • Press Set Button in center of Quick Control Dial" • Use Quick Control Dial to adjust Color Temp" • See EOS 10D Instruction Manual, pg 53" (4) Iris diaphragm at “6”" (5) RAW Files (CRAW)" (6) MT Destination Folder (Pick the correct place in the folder hierarchy)"
Canon EOS 10D Camera Settings - Hematoxylin and Eosin!
!
(1) (2) (3) (4) (5) (6)
1/4000 Shutter Speed" 1600 ISO Film Speed" 4500 [K]elvin for Color Balance" Diaphragm iris at “3”" RAW Files (CRAW)" HE Destination Folder (Pick the correct place in the folder hierarchy)"
Aperture 3.5 Software! (1) (2) (3) (4) (5)
Connect the Canon EOS 10D camera via USB to the computer (Ros)." Turn the Main Switch, under the camera’s large LCD Monitor, to the ON setting." Aperture software should launch in Import mode. Exit Import mode." Select “File” » “Tether” » “Start Session”" Edit Tether Settings:" • Store Files: Folder to save new images in" • Subfolders: None." • Version Name: Select Edit, and format with (Custom Name) and (Counter). Edit the custom name with desired file name. Start Counter at 1 with 3 digits." • Name Text: Do not edit. It will be the same as Version Name. " • Check box “Apply to Master filenames.” This applies the Version Name to the filename in Finder." • Metadata: select Eidactics, which is a copy of “Basic Info” with updated fields." • Backup to: None." • Effect: None." • Check box “Show HUD” to display the Capture button on screen."
2014 July 4 Friday
Page 4! of 12 !
Photographing & Segmenting Muscle Fibers
• Click Start Session." (6) Click the Capture button to take a photo." (7) To verify camera settings used, select the photo and click the “Info” tab. Within the “Info” tab, select “EXIF Info” from the pull down menu." Shutter Speed and ISO Film Speed were chosen to make the histogram of photo lightnesses centered and as wide as possible." Focus can be optimized by viewing slides at 20 mag. Then, rotating the fine focus -200 microns (2 full rotations in the negative direction) should produce optimal focus x3.125."
Scanning the Slide! View the Calibration Slide through the eyepieces and through the camera to understand the relationship between what you see in the former and what will be recorded by the later. Take successive photos in a systematic way across the sample, spaced so that overlap of adjacent frames is not excessive, but all parts of the sample are captured. To do this, you need to understand how the camera images will be cropped. First, circular cropping removes peripheral parts of the image that are affected by lens aberrations. Then, square cropping removes regions that would result in excessive overlap when the 2D montage is created. A square crop mask with rounded corners therefore seems a good choice." Inadequate overlap of these cropped images will, of course, leave gaps. These can subsequently be filled in. Excessive overlap will cause Photoshop’s Photomerge function to fail – ⅓ linear overlap is said to be optimal."
Processing CRAW Slide Images into Montages! We use only the uncompressed CRAW image. Set the camera control software to ignore (or minimize) the JPG image that usually accompanies it."
(1) Convert CRAW files to PSD! Reason: Batch cropping in the succeeding step does not work with CRAW files." (a) (b) (c) (d) (e)
Highlight all CRAW files in Finder and drag to open in Photoshop CS5.1" "Camera Raw 6.3" plugin will start." Press "Select All"" Press "Same Images…" and save with .psd extension in a new folder." After saving, exit by pressing "Cancel""
(2) Record Cropping Actions! Reason: Record actions with a sample image in order to then apply that recording to other images as a batch process." (a) (b) (c) (d) (e) (f) (g) (h) (i)
Go to View > Rulers (Enable)" Go to Photoshop > Preferences > Units & Rulers" Change "Units > Rulers: percent" then press OK" Open a PSD microscope slide image in Photoshop" Go to Windows > Actions (Enable)" Click the "Actions" tab" Click the folder icon to "Create New Set" and name it "Auto Batch Crop"" Click the paper icon to "Create New Action" and name it "Crop circ2500+squ2048px"" Click "Record""
2014 July 4 Friday
Page 5! of 12 !
Photographing & Segmenting Muscle Fibers
(j) Click "Elliptical Marquee Tool" on the side toolbar OR press the (M) key" (k) Check the tool's settings:" • Feather: 0 px" • Anti-Alias: disabled (not checked)" • Style: Fixed Size" • Width&Height: 2500px" (l) Using the x,y rulers align your mouse cursor to 50% on both axes of the open image" (m)Mouse click while holding down the Option key and a circle should appear centered on the image" (n) Click “Rectangular Marquee Tool”" (o) Set Width&Height: 2048px" (p) Place cursor at 50% on both axes, and Option-Click" (q) Click "Refine Edge…" and set the following:" • View: On Layers" • Everything else unchecked or at zero values."
(r) (s) (t) (u) (v)
• Output to: New Layer" • Remember Settings: Enable (Checked)" In the Layers window, select "Background" layer and delete (press Trashcan icon)" Go to Photoshop > Preferences > Units & Rulers" Change "Units > Rulers: mm" then press OK." Click "Stop Recording" on the Actions window" Close the image without saving"
(3) Automate Batch Crop! (a) Go to File > Automate > Batch… and select:" • Set: Auto Batch Crop" • Action: Crop circ2500+squ2048px" • Source: your folder with PSD files you wish to crop" • Check only the box "Suppress File Open Options Dialogs"" • Destination: make a new folder for your cropped images" • File naming: "Document name" + "circ2500+squ2048px" + "extension"" (b) Click OK"
(4) Photomerge! (a) (b) (c) (d) (e) (f)
Go to “File > Automate > Photomerge…”." Select “Reposition”. This translates and rotates, but does not otherwise distort the component images." Use: Files" “Browse” to select the cropped images to be merged." Check only "Blend Images Together"" Click OK. Photomerge process may pause at the end of “Align Selected Layers Based On Content” (progress ~20%) briefly, for a long time, or (so far as we know) forever. Reducing number of images and image overlap may solve this problem." (g) Save the multilayer montage (eg: “R57970A S2 MT x20 btm multilayer.psd”)." (h) Flatten Image, and save as a different file (eg: “R57970A S2 MT x20 btm flat.psd”)."
2014 July 4 Friday
Page 6! of 12 !
Photographing & Segmenting Muscle Fibers
(5) Optimize Color Contrast! (a) In the flattened montage, use “> Image > Adjustments > Levels” to optimize the gamut of each of the Red, Green & Blue channels (adjust the left and right Input Level sliders), while keeping the background as close to neutral grey as possible (adjust the center Input Level slider)" (b) Save under a new file name (eg: “R57970A S2 MT x20 btm.psd”)." (c) Copy only the finished montage (eg: “R57970A S2 MT x20 btm.psd”) to the "Rabbit EOM Histology Montages" folder. Only this folder will be SugarSync’ed to make it available to all lab members. The “Rabbit EOM Histology Picts” folder will exist on Yori (to which the component images were acquired), on J2D (which JMM will ChronoSync to the Yori folder), and on the Time Capsule backups of Yori and J2D."
IPP-9 Image and Data Files! Image-Pro Premier 9 opens, processes, and saves image files. Some IPP-9 operations modify the image itself, whereas others create data in “image overlays”, which are included1 when the file is saved as TIFF (use ZIF compression)." Set » File » Options » Application » Display: Individual Object Display Limit = 20,000."
Multiple people may work on these files over a long period so it is essential that naming conventions be strictly followed. Files will be stored as in the following example:" Rabbit EOM Histology Analysis/ R57976a S2 MT x20 btm (BPX)/ R57976a S2 MT x20 btm 20131118jmm (psd).tif"
(read-only)
R57976a S2 MT x20 btm 20131202rr (grains ew7s8).tif "
(read-only)
R57976a S2 MT 20131202rr.iqo R57976b S2 MT x20 btm (saline)/ R57976b S2 MT x20 btm 20131118jmm (psd).tif "
(read-only)
The initial part of folder & file names, eg, “R57976a S2 MT x20 btm” never varies in any way - it identifies the the sample across the many stages of processing, and must be searchable. “Date-time-person” appended to each modified file version saved. Usually 1 day resolution is sufficient (format: yyyymmdd), but depending on frequency of saving, 1 hour (yyyymmddhh, 24 hour clock), or even 1 min resolution (yyyymmddhhmm) may be Overlays are stored in TIF tags using an IPP-9 format. It is possible to save overlays to separate images (eg. make the image white and burn the overlay), but it will not be a vector object. 1
2014 July 4 Friday
Page 7! of 12 !
Photographing & Segmenting Muscle Fibers
useful. Initials of the person creating the file are appended. Date-time-person is followed by a “hint” in parentheses. The hint helps communicate something about the sample or the stage of analysis completed, eg:" – this sample was treated with bupivacaine." – this is a control sample, treated with saline." psd – Photoshop was used to remove non-muscular structures." ew7s8 – the Grains app, with Edge Width = 7 and Sensitivity = 8 was run on this file." fixing – manual refinement of Grains app segmentation underway." fixed – manual refinement done." Important files should be made read-only so they are not inadvertently overwritten (see examples above)." BPX
saline
The Custom Ribbon Contains Most Necessary Tools! Our workflow uses the tool groups in the Custom ribbon (a toolbar near the top of the workspace), roughly left to right:" • Filters » 2D Filters
may be useful to prepare images."
• Regions of Interest,
eg, limits the region to which the Grains app is applied, which saves time when testing its
parameters." sets filters applied by the Grains app after the image is separated into regions, and determines measurements computed on measurement objects."
• Measurements » Types
toggles the filters. When a fiber fails to be identified, rerun Grains with filters off so see if it was in fact segmented, but was subsequently filtered out."
• Measurements » Ranges
• Measurements » Options • Select
•
toggles the Measurements Options details panel."
tools select, delete, show, and hide measurement objects."
tools create, delete and move notes on an image overlay. Do not Burn-In annotations, which would make them part of the image, damaging the raw data." Annotate
• Direct » Polygon
draws measurement objects that need to be manually added."
draws lines to split groups of fibers wrongly identified as one. Draw multiple split lines and then apply them all with a single click of the Apply button. Do not use the Split All button, which will automatically process the entire image according to it’s own rules."
• Split
combines multiple measurement objects into one (even if the display does not show it)."
• Relative » Merge
puts measurement objects into multiple named classes for subsequent identification or selective processing. We define various classes to record image processing."
• Class
• View
tools show results."
Preparing an Image for Segmentation! Two sorts of image preparation may be useful: Gaussian, Morphological, and other overall processing, and masking or removal of troublesome non-muscular regions. Gaussian Blur can help fill the larger cracks in fibers, Morphological Opening (eg, with a 7x7 structuring element) can fill smaller ones, and Morphological Closing can cleanup the background. these operations can be done in IPP-9."
2014 July 4 Friday
Page 8! of 12 !
Photographing & Segmenting Muscle Fibers
IPP-9 has tools for manual editing of objects2, but Adobe Photoshop may be easier to use. In Photoshop choose an appropriately-sized, soft-edged brush, sample local background, and paint over any vasculature, nerves, or connective tissues that might be wrongly segmented as muscle, and would fail to be discarded by Background Removal (see below). Resample local background as needed to avoid painting spurious edges." It is also possible, before running Grains, to use Regions of Interest » Options to define ROIs that avoid clumps of connective tissue, nerves, and blood vessels, which might otherwise be identified as muscle fibers (figure =>), but the current version of Grains applies a mask in its first stage, so this may be unnecessary3." The Grains app begins by optionally masking the image. A mask can therefore be constructed to exclude non-muscular tissues that would not be eliminated by the background-removal filters of Stage 4. It must not remove any muscle. The mask is a binary image, black for regions to be eliminated. It can be created using any or all of the following methods:" • Use Segment » Smart » Smart Segmentation to select some Objects and Background areas (figure =>). Smart Segmentation will use the selected Recipe Options to segment the image." • Mask and Overlay » Create Mask • Convert » Mono 8bpp » Apply
to remove small black areas inside white regions (eg, use the 7 x 7 Octagon structuring element)."
• Filters » 2D Filters » Morphological » Close
to expand white regions, assuring no muscle is masked out."
• Filters » 2D Filters » Morphological » Dilate
• Draw ROIs around regions to be excluded using the Regions of Interest » Polygon Tool, and setting them to 0 (Math » Calc » Set 0 » Apply)." The resulting mask can then be saved and loaded into Grains."
Segmenting an Image with the Grains App! The Grains app, Grains.ipx, is an add-in module for IPP-9, designed to segment “particles” or “grains”. It can separate particles even if their borders are not completely closed. If the Grains app panel does not appear in the left sidebar, open it from Apps, or if necessary, from Search » Open Project (both in Apps ribbon). It has a few parameters that need to be set:" • Edge Type = Step
- for dark muscle fibers on light background."
- small values detect thinner edge boundaries, but are more sensitive to noise (values 6-10 have been useful in the past)."
• Edge Width
2
By default, objects are locked following a Count-Size operation; use the Unlock Objects menu item to unlock all objects.
3
Thanks to Yuri Gaidoukevitch, Senior Software Engineer at Media Cybernetics for developing these Grains modifications. 2014 July 4 Friday
Page 9! of 12 !
Photographing & Segmenting Muscle Fibers
- high sensitivity values detect low luminance “ridges”, tending to separate fibers into fragments (see “Watershed Separation” below). Values that are too low will fail to separate distinguishable fibers (values 4-8 have been useful)."
• Sensitivity
• Mask
- optionally used to mask-out areas in the image that would be wrongly identified as muscle fibers."
- when checked, the Grains app is automatically run whenever a parameter is changed. This is useful for trying parameters, inside a small ROI."
• Interactive
• Measure Grains
- runs the Grains app."
Test Grains Parameters on Part of an Image! • Use Select » Square Marquis to draw an ROI around a typical part of the image, and then Select it." • Use Measurements » Ranges to turn off filtering so you can see the unfiltered separation." • Set Measurements » Options » Segmentation Options » Clean Borders = None. Otherwise, this is set to All, which discards regions on borders of the image or ROI (if fibers, they would be partial), but for setting Grains parameters, we want nothing discarded." • Check Interactive. Now, vary Edge Width and Sensitivity until muscle fibers are well separated. Segmentation errors will subsequently be repaired manually, and your judgment of what constitutes a good Grains separation should be informed by the difficulty in doing the various types of repair. For example, merging fragments must be done one group at a time, and each merge takes several seconds to process. Also merged fragments continue to show as fragments, albeit in a distinct color. Large numbers of such merges may make the display confusing."
Run the Grains App! • Set the active Class to “Grains”. This will attach the label “Grains” and a yellow color to each object found by the Grains app." • Optionally, select a previously-created Grains » Mask." • Click Grains » Measure Grains to run app." Show or hide the segmentation with Select » Lighting." The Grains app performs a series of operations. After applying the Mask, if any, edges are extracted and smoothed with 2 filters, both of which use the Grains » Edge width parameter (figure above) to define filter size”" • A Variance Filter extracts edges. This filter associates with each pixel the statistical variance of pixel intensities in its neighborhood. Pixels near grain edges have high variance (due to intensity differences across the edge, plus noise), whereas pixels inside grains and in the background have lower variance (noise only). Smaller filter kernels detect thinner edge boundaries, but are more sensitive to noise." • A Gaussian Filter then smooths extracted edge contours. Larger kernels increase smoothing." The image is then separated into regions with a Watershed algorithm, using seed points in uniform areas, between edges found by the Variance Filter. Watershed treats the image as a (greyscale) luminance landscape, gradually filing with water to a uniform height. There will be multiple pools at first, but as the water rises pools merge. Every point at which pools merge is on a “ridge”, and the set of these points are “watershed lines”, which are used to separate objects. Low ridges, which may reflect only noise, are ignored by choosing a “pre-flood” level, below which merging of pools is ignored:" • Watershed Pre-flood ignores Watershed ridge points up to a luminance level determined by Grains » Sensitivity." • The Watershed algorithm then divides the image into regions."
2014 July 4 Friday
Page 10 ! of 12 !
Photographing & Segmenting Muscle Fibers
Regions (treated as wholes) are then retained or discarded according to Measurements » Types » Selected Measurements/Filters. This list actually serves two functions: [1] All items in the list appear as columns in View » Data Histogram, and [2] Items with a “√” are used to filter the results of Watershed separation. Measurements » Types. Measurements » Ranges turns this filtering on and off." But only some filters are applied by Grains (which ones???). Others can subsequently be applied manually. Useful Measurements Filters Values Histotest-lmr Region: Area (very large & small regions are likely non-fibers) Region: Roundness (elongated regions are likely groups or non-crossections)
150-20000 1-2.5
Region:Intensity, Mean
10-160
Region:Intensity, Red
185-255
Region:Intensity, Green
0-140
Region:Intensity, Blue
0-100
Region: Std Dev (high variability regions may not be muscle fibers)
0-65
Object: Class (distinguish Grains, Split, Added & Merged fibers)
(no filter)
Finally, Segmentation Options » Clean Borders = All discards regions bounded by image edges (segmented fibers clipped by image borders are incomplete and must be ignored)." The appearance of segmented regions can be altered in Measurements » Options » Measurements Options » Appearance Options » Segmentation Options » Appearance."
Grains App Files! • Grains.IPX – is a ZIP container from which project files are extracted when it is loaded. Grains is a generic open module, which can be viewed and modified using the Scripting Workbench."
defines the UI (User Interface)."
•
GrainsApp.VB
•
ModuleGrains.VB
is the Visual Basic code that does the separation.
2014 July 4 Friday
Page !11 of !12
Photographing & Segmenting Muscle Fibers
• MuscleFibers.IQO – contains all settings in the Measurement and Count/Size ribbons, including RGB intensity filtering. IQO files can be loaded and saved in Measurements » Options » Segmentation Options. • MuscleFibers.GRO – contains Grains app parameters for segmenting muscle fibers. GRO files are loaded and saved in Grains (figure above)." • MuscleFibers.ISG – contains Smart Segmentation » Recipe Options (only needed to create the optional image mask)."
Manually Refining a Segmented Image! Accurate fiber size statistics may require manual repair of incorrect segmentations after Grains has run. We keep track of these refinements by assigning them to “classes”. Classes are distinguished by color in the main window, and by color and Feature Name prefix in the Measurement Table." It seems useful to perform the refinements on one fascicle at a time, and generally in the order given below, although several cycles of refinement will usually be necessary." Use Select » Hide or Show Overlay to see the identified fibers (“measurement objects”) or the raw image." 1. Result of Running Grains App (class: “Grains”): No manual modifications." 2. Splitting Groups of Fibers (class: ”Split”): Multiple fibers, and fibers with attached background are split. Oddly, IPP-9 re-colors only one of the resulting objects. Use Split to draw lines on the image. Lines must completely transect the group to be split, and if they extend into neighboring fibers, chunks will be removed from those fibers. Splitting will sometimes produce tiny, eg, single pixel, measurement objects. These can be seen and selected in the Histogram, and deleted4" 3. Removing Non-Fibers: Use Select » Select to select the wrongly identified fiber, and delete it with Select » Delete. Fibers with extreme Areas, Roundnesses, or other values are immediately suspect." 4. Adding Fibers (class: ”Added”): Fibers or parts of fibers that fail to be automatically identified must be added manually. Use Direct » Polygon to outline the omitted fiber or part. If a part,then merge the fragments, as below." 5. Merging Fiber Fragments (class: “Merged”): Click to select the first fragment to be merged, Control-Click to select subsequent fragments, and then click Relative » Merge. Because visible contours of merged fibers are not merged in the measurements overlay, we tag them with a distinct color so we know the fragmentation has been resolved. Unfortunately, where merged fibers are adjacent the display can become ambiguous. Merging is best delayed until most splitting and removing of non-fibers is done." Do not use “undo” (CTL-Z). Due to a program bug, if there are more than 200 measurement objects, all measurement data will be lost! Instead, make frequent use of Capture » Save for Analysis, which saves the image and all overlays under a unique filename (similar to Save As in TIFF format)."
Measuring a Segmented Image! • Click View » Data Histogram and select the measure to be shown in the histogram from the drop-down menu." • Select the measures computed for each segmented object using Measurements » Types (figure =>)." • Measurements can be saved as IQM files in the Measurements data table."
4
The Histogram canals be used to filter measurement objects, which is not what we want to do here. 2014 July 4 Friday
Page 12 ! of 12 !
Photographing & Segmenting Muscle Fibers
Modifying the Grains App! Add-in modules are managed as “projects” in the development kit: Click Macros » Scripting Workbench." is the Grain app UI."
➡
GransApp.vb
➡
ModuleGrains.vb
is the code."
Suggested Program Improvements! • A “lasso tool” would be useful to select irregular areas from which a large number of measurement objects need to be deleted." • Muscle fibers are frequently wrongly segmented into fragments, following cracks caused by histological processing. It would seem a natural addition to the Grains app to make it prefer "roundish" contours, or to enforce some degree of smoothness or maximum curvature of segmentation lines." • Due to a program bug, if there are more than 200 measurement objects, all measurement data are lost! Raising the limit as panned in IPP-9.1 is not a solution." • Measurement objects combined using Relative » Merge do not change their appearance in the measurements overlay. As a workaround, we tag them with a distinct color so we know they’ve been merged, but if multiple merged objects abut, this becomes ambiguous. Merge should be fixed to update the overlay display, as does Split.
2014 July 4 Friday
