Erosion Identification in Metacarpophalangeal Joints in Rheumatoid Arthritis using High-Resolution Peripheral Quantitative Computed Tomography

One of our goals as part of the study group for extreme computed tomography in RA, or spectra, is to develop better image analysis tools to understand bone damage in rheumatoid arthritis. Applying high resolution peripheral quantitative computed tomography, or HRPQCT for short, to rheumatoid arthritis, researchers have shown that this modality is far more sensitive than standard X-rays to detect bone erosions. And have detected bone changes in response to treatment in as little as three months.

One of the major challenges is accurately identifying bone erosions in HRPQCT images. Most researchers rely on expert rheumatologists with extensive training. The analysis, while powerful, is also very time consuming.

Our group has developed a new open source software module that can be used to analyze bone erosions in HRPQCT images of rheumatoid arthritis patients. We have shown that the results are comparable to previously established methods. Implementation of this module will allow us to more efficiently investigate the effects of altered treatments and subclinical inflammation on bone damage progression in rheumatoid arthritis.

To begin, launch the 3D Slicer software. Then launch the training module on the 3D Slicer's toolbar. Click on the dropdown menu to locate the bone analysis modules and hover the mouse above it, then click on training.

Click proceed to load all necessary files. Next, hover over the input volume and select a volume dropdown menu and select a scan type. Choose the mask from the input mask, select a volume dropdown menu.

Verify the input volume against the measurement ID and the metacarpophalangeal joint in both sections. Click on output erosions, select a segmentation, to create a new output segmentation. Then click the dropdown menu and select create new segmentation.

Now, click on the dropdown menu labeled seed points, none. And then choose create new point list to create a new seed point. Scroll through the slices, pinpoint erosion sites, and press the red dot/blue drop button to place a seed point in the corresponding area of interest.

To change the seed point size, modify the percentage size in the text box labeled seed point size. To reorient the image to the location of the seed point in the other two planes, hold down the Shift button as the cursor is being used to move the seed point. After placing the seed points, click on the get erosions button to initiate the erosion measurement algorithm.

When the measurements are complete, the module will provide feedback on the seed point placement. The module showed that erosion identification against expertly annotated images resulted in accurate feedback and segmented gray scale images. However, incorrect seed point placement results in an error prompt indicative of no erosion.

The module also accurately identified cysts when a seed point was placed on it. Wrong seed placement on the cyst resulted in a corresponding error prompt. An insufficient number of seed point placements also resulted in corresponding error prompt.

To begin launch, the 3D Slicer software. For image input of the identified bone erosion sites, click on the button labeled data on the 3D Slicer window. Then press choose files to add, and locate individual images to add the files.

To import an entire directory of images at once, press the choose directory to add option, then search and insert the preferred directory. Next, click on the dropdown menu of the 3D Slicer's toolbar. Hover the cursor over the bone analysis modules and click on automatic mask.

Under the tab, step two, automatic mask, choose an input volume from the input volume dropdown menu. Select create new label map volume in the output segmentation dropdown menu. If another output label is preferred, click on create new label map volume as, and provide the desired label.

Input the number of bones to be masked in the given text box. Then select or mirror or dual threshold from the algorithm dropdown menu. Click on get mask to run the algorithm and output a result in the corresponding image directory.

If the image contains multiple bones, a separate mask will be saved for each one. To import a binary mask onto 3D Slicer, copy the mask files onto the load masks directory located in the BAM folder. Go back to the 3D Slicer window and click on the tab step three, manual correction.

To load the mask files and automatically convert them to slicer segmentations, click on load. To make changes to the mask, use the tools provided in the manual correction tab. To start statistical analysis, click on the erosion volume module from the dropdown menu on the toolbar.

Load an image and its corresponding mask. Select the input image and mask, create an output segmentation, and place the seed points on the erosion. Click on the get erosions button to initiate the erosion measurement algorithm.

If the size and shape of automatically detected erosion volumes are unsatisfactory, manually edit them under the tab step five, manual correction. Now, click on the tab labeled step six, statistics, to export the computer data into a spreadsheet file. Next, input the erosion volume in the dropdown menu labeled input erosions.

Now, provide the gray scale image and its voxel width in millimeters in the text box under the master volume menu. Click on get statistics to generate a spreadsheet file in the erosions output data directory located in the BAM folder.