Table 1.
Segmentation of thigh CT images into regions of fat, muscle, and bone based on user-defined Hounsfield unit ranges for tissue density.
Fig 1.
Step-by-step histogram method for segmentation of thigh tissue into fat, muscle, and bone and then calculating the cross-sectional area of each region.
Prior to beginning, the file location and pathway of the CT scan images should be known. CT grey level images will often be in DICOM file format and may show the extension *.dcm though SliceOmatic can read image files in other formats too. SliceOmatic segments tissue into regions using a script file run by the program. The user must create this file by pasting the following script text into Notepad, or similar program, and saving as an ANSI text file whose name must end with the extension “.scp”. The script sets the color for each “tag” and the density ranges to be applied when the Region Growing (“region”) or Histogram Segmentation (“histo”) modes are used for segmentation. Script Text to be copied into Notepad File titled “Thigh Segmentation Tutorial.scp” is shown.
Fig 2.
SliceOmatic is opened by double clicking the desktop icon or by following the Windows Start menu > All Programs > Tomovision > SliceOmatic.
The open SliceOmatic window displays the Main Menu bar at the top and the Display Area on the left contains the word “WARNING”. The areas on the upper right contain mostly test feedback that can be ignored for this tutorial and the lower right contains the Control and Tools Areas which will later contain needed control buttons, sliders, and results. All images in Figs 2–12 have been reprinted from SliceOmatic v5.0 rev7 under a CC BY license, with permission from Tomovision, original copyright 2017.
Fig 3.
The script file (Thigh Segmentation Tutorial.scp) presets SliceOmatic analysis settings and is applied by dragging the file over into the Display Area.
Alternatively, the script can be applied at startup every time SliceOmatic opens by following the Main Menu > File > Config > Startup Script File. Add the full path and script filename to the box, toggle OFF to ON, and apply and save.
Fig 4.
The CT grey level images or DICOM files should be available in a convenient known location such as a desktop folder.
A CT scan is opened by dragging the image file into the Display Area. SliceOmatic can open multiple images at the same time but it is recommended that the user focuses on single images until familiar with the program and the analysis method.
Fig 5.
The thigh of interest is isolated by temporarily tagging all tissues with a density ≥ -190 HU (i.e. fat, muscle, and bone) using the region growing mode.
Coloring will usually not escape the border of the thigh since it is mostly surrounded by areas with density less than the -190 HU limit (e.g. air at -1000 HU). This limit was set by the script text (region: 1 min on -190.00) shown in Fig 1. The Region Growing options are made available in the lower right of the Control Area by clicking in the Main Menu on Mode > Region Growing. The region growing options to be clicked are: smallest brush size (a bug in the program prevents a larger brush from being used), the temporary color (NDF i.e. red), and Grow 2D. The histogram shows that the peaks for fat (left peak) and muscle (right peak) are included in the red areas with a density lower limit of -190 HU (left image). The peak on the far left is air. The bone peak is to the far right in the histogram but too small to be seen without narrowing the range of the y-axis by moving the gray slider bar down (right image).
Fig 6.
The thigh is then colored by clicking anywhere in its area.
Sometimes areas touching the thigh (e.g. the opposite thigh or exam table) are also colored. Editing these inappropriately colored areas down to only the area of interest is presented later in this tutorial.
Fig 7.
While the thigh is temporarily tagged red (NDF), further analysis is limited to this area by using Tag Lock to exclude untagged areas from segmentation.
The size of tagged areas can also be shown. Click in the Main Menu on Tools > Tag Lock and Tag Surface/Volume to show these tools in the Control Area (note: the size of the SliceOmatic window likely needs to be maximized to see these tools and they may appear next to one another as shown here). Click Tag Lock “none” to exclude areas without a tag. The area of the red tag, i.e. total thigh area (161.8 cm2) is now shown and should be recorded or can be calculated later as the sum of all regions.
Fig 8.
Segmentation into fat, muscle, and bone can now be performed based on the densities defined in Table 1 and set by the script in Fig 1.
Click in the main menu on Modes > Histogram Segmentation to show the segmented thigh. If the opposite thigh also colors, then the tag lock was not set to “none”.
Fig 9.
Next click compute in the Control Area to apply the segmented changes to the thigh and show the size of each tagged region (calculated as area under the curve in the histogram).
The size for areas of NDF and VLDM may be the only ones seen, but the slider bar allows scrolling to see all areas including Bone at the far right. Also click Preview so that it is turned off or a bug in the program will interfere with the next step.
Fig 10.
The skin and bone marrow tag with a mixture of colors assigned to the various densities of fat and muscle.
Since this could interfere with the results, particularly muscle, we have chosen to edit or brush the skin and structures within the subcutaneous fat (e.g. great saphenous vein) to be normal density fat and brush bone marrow to be bone. From the main menu click on Modes > Edit. A brush size and the desired colored tag must be chosen. The brush is used by holding the left mouse button while dragging the cursor over the area to be edited. This process is easy for marrow since it is surrounded by bone and easy for skin since Tag Lock “none” protects outside the thigh. Brushing can be made easier by increasing the image size in the Display Area by clicking the keyboard plus (+) sign. Similarly, minus (-) decreases the image size. When choosing Edit mode, if the image reverts to the red colored thigh, then either Compute was not selected or Preview was not turned off as instructed in Fig 9. When brushing is complete, if uncolored pixels are present, they can be brushed the desired color after Tag Lock “none” is turned off.
Fig 11.
As mentioned in Fig 6, sometimes an attempt to isolate the thigh of interest tags additional areas.
Should this happen, it appears easiest to proceed through the instructions in Fig 9. The Editing skills learned in Fig 10 can also be used to limit the tagged areas to the thigh of interest. To edit the touching leg, tag lock “none” will need to be turned off and the inappropriately colored area brushed to the color “none”. The final tagged image can be saved by clicking Main Menu > File > Save TAG Files. The program uses the same filename as the original DICOM file (*.dcm) but adds to the extension (*.dcm.tag) and also saves the TAG file in the same location as the DICOM file. The TAG file will open automatically in the future when the DICOM file is opened. If the original DICOM file needs to be analyzed again, then a copy of the *.dcm file can be saved and analyzed under a new filename.
Fig 12.
The TAG Surface Areas or results are exported as a CSV file (default name “results.csv”) and the parameters to be exported can be selected.
The current analysis was limited to area (cm2) as opposed to volume or HU average, minimum, or maximum by choosing Main Menu > File > Config > File > Surface Units (in cm2). Unselect any unwanted results and Apply & Save. Export the results by clicking Write Result File in the Control Area.
Fig 13.
The CSV file (results.csv) can be opened and manipulated using excel and is saved in the same location as the original DICOM file.
When additional scans are analyzed and ready to Write Result File, the same result.csv filename will be offered. If chosen, an option will be given to append result.csv by adding the new results to the previous results. The results layout in the CSV file is shown. In addition to the CT scan filename, the columns also contain the scan position and thickness of the scan slice (neither needed for this tutorial), and the surface area (cm2) for the regions of interest as abbreviated in Table 1.
Table 2.
Effect of CT image editing of skin and bone marrow on contributions of muscle, fat, and bone to total thigh area.
Table 3.
Effect of image editing on cross-sectional areas of thigh tissues and their change after 12-wks of resistance training (N = 24).