Estimating the whole bone-marrow asset in humans by a computational approach to integrated PET/CT imaging

PURPOSE: Despite their relevance in clinical medicine, the extension and activity of the bone marrow (BM) cannot be directly evaluated in vivo. We propose a new method to estimate these variables by combining structural and functional maps provided by CT and PET. METHODS: BM extension and glucose uptake were estimated in 102 patients undergoing whole-body PET/CT because of a history of nonmetastatic melanoma. Image analysis assumed that the BM is surrounded by compact bone. An iterative optimization scheme was applied to each CT slice to identify the external border of the bone. To identify compact bone, the algorithm measured the average Hounsfield coefficient within a two-pixel ring located just inside the bone contour. All intraosseous pixels with an attenuation coefficient lower than this cut-off were flagged as 1, while the remaining pixels were set at 0. Binary masks created from all CT slices were thus applied to the PET data to determine the metabolic activity of the intraosseous volume (IBV). RESULTS: Estimated whole-body IBV was 1,632 ± 587 cm(3) and was higher in men than in women (2,004 ± 498 cm(3) vs. 1,203 ± 354 cm(3), P < 0.001). Overall, it was strictly correlated with ideal body weight (r = 0.81, P = 0.001) but only loosely with measured body weight (r = 0.43, P = 0.01). The average FDG standardized uptake value (SUV) in the thoracic and lumbar vertebrae was 2.01 ± 0.36, Accordingly, intraosseous voxels with SUV >= 1.11 (mean spine SUV - 2.5 × SD) were considered as active "red" BM and those with SUV <1.11 as "yellow" BM. Estimated red BM volume was 541 ± 195 ml, with a higher prevalence in the axial than in the appendicular skeleton (87 ± 8 % vs. 10 ± 8 %, P < 0.001). Again, red BM volume was higher in men than in women (7.8 ± 2.2 vs. 6.7 ± 2.1 ml/kg body weight, P < 0.05), but in women it occupied a greater fraction of the IBV (32 ± 7 % vs. 36 ± 10 %, P < 0.05). Patient age modestly predicted red BM SUV, while it was robustly and inversely correlated with red BM volume. CONCLUSION: Our computational analysis of PET/CT images provides a first estimation of the extension and metabolism of the BM in a population of adult patients without haematooncological disorders. This information might represent a new window to explore pathophysiology the BM and the response of BM diseases to chemotherapy