Exhibit Hall | Forum 6
Purpose: Active bone marrow (ABM) can serve as both a target for disease and an organ-at-risk in EBRT and radiopharmaceutical therapy. Identification of ABM with CT and MRI could improve treatment planning accuracy and increase accessibility of bone marrow (BM) imaging/segmentation to a larger patient population. This work aims to benchmark the ability of proton density fat fraction (PDFF) MRI and non-calcium multi-energy CT (MECT) to delineate ABM for individual subjects relative to the gold standard of [¹⁸F]FLT-PET imaging.
Methods: PDFF MRI, non-calcium MECT, and [¹⁸F]FLT-PET images were obtained for three Wisconsin Miniature Swine™ subjects. For each subject, a BM contour was created on the CT image and propagated to the MR and PET images. ABM was defined in the PET image as pixels within the BM contour with SUV(bw) above the mean. Prospective ABM contours were created on the MECT image through a lower HU value threshold, above which the marrow is considered active. Prospective ABM contours were defined on the MRI through an upper fat fraction threshold below which the marrow is considered active. These volumes were compared to the PET-determined ABM volume by calculating the DICE coefficient between the MECT and MRI ABM contours and PET ABM contour.
Results: For the three swine subjects in the PDFF MRIs, maximum DICE coefficients of 0.744, 0.777, and 0.732 occurred at maximum fat fraction values of 50, 55, and 65, respectively. For these same subjects in the MECT images, maximum DICE coefficients of 0.654, 0.63, and 0.654 occurred at minimum HU values of 20, 10, and -10, respectively.
Conclusion: Image value thresholds were determined for PDFF MRI and non-calcium MECT to maximize similarity between ABM contours as compared to FLT-PET. Both PDFF MRI and MECT show promise in identifying ABM distribution for individual subjects without the need for PET scans.
Funding Support, Disclosures, and Conflict of Interest: Funding was received from Siemens and General Electric (GE) Healthcare
Image Analysis, Segmentation, Bone Structure