Tumor Segmentation of Whole-Body Magnetic Resonance Imaging in Neurofibromatosis Type 1 Patients: Tumor Burden Correlates

Abstract

PURPOSE: To use software segmentation of whole body MRI (WBMRI) to quantitate total body tumor volume (tumor burden) in patients with Neurofibromatosis type 1 (NF1) and examine associations between demographic factors and tumor burdens. METHODS: Patients with NF1 underwent WBMRI, and the images were reviewed by a board-certified radiologist for tumors. Each tumor was classified as superficial or internal, and discrete or plexiform. Tumors were then segmented by a trained assistant using a semi-automated software-based tool (SliceOMatic, TomoVision, Québec, Canada). Segmentation times were recorded. Segmentation yielded the total quantity and tumor burden of superficial, internal, and plexiform tumors. Correlations between the segmentation data and patient demographic factors were examined. RESULTS: 15 patients were included (42.3 +/- 13.6 years, 10 female, 5 male). Segmentation time ranged from 20 to 60 minutes, and yielded 2328 tumors (1582 superficial, 746 internal, and 23 plexiform). One tumor (0.04%) was malignant, all others were benign. The number of tumors per patient ranged from 14 to 397. Tumor burden ranged from 6.95 cm3 to 571 cm3. Individual tumor volume ranged from 0.0120 cm3 to 298 cm3. Significant correlation was found between total volume of superficial tumors and height (ρ = 0.5966, p < 0.02). Male patients had larger overall tumor burdens (p < 0.05) and larger superficial tumor burdens (p < 0.03). Patients with a negative family history of NF1 tended to have more tumors (p < 0.05). CONCLUSION: Segmentation of WBMRI in patients with NF1 is possible and can elucidate meaningful relationships between disease phenotype and demographic factors.