Role of Apparent Diffusion Coefficient Map-Based First- and High-Order Radiomic Features for the Discrimination of Sacral Chordomas and Chondrosarcomas with Overlapping Conventional Imaging Features

PURPOSEChondrosarcomas arise from the lateral pelvis; however, midline chondrosarcomas (10%) display similar imaging features to chordoma, causing a diagnostic challenge. This study aims to determine the diagnostic accuracy of apparent diffusion coefficient (ADC)-based radiomic features and two novel diffusion indices for differentiating sacral chordomas and chondrosarcomas.METHODSA retrospective, multireader review was performed of 82 pelvic MRIs (42 chordomas and 40 chondrosarcomas) between December 2014 and September 2021, split into training (n = 69) and validation (n = 13) data sets. Lesions were segmented on a single slice from ADC maps. Eight first-order features (minimum, mean, median, and maximum ADC, standard deviation, skewness, kurtosis, and entropy) and two novel indices: restriction index (RI, proportion of lesions with restricted diffusion) and facilitation index (FI, proportion of lesions with facilitated diffusion) were estimated. One hundred seven radiomic features comparing patients with chondrosarcoma versus chordoma were sorted based on mean group differences.RESULTSThere was good to excellent interobserver reliability for eight of the 10 ADC metrics on the training data set. Significant differences were observed (P <.005) for RI, FI, median, mean, and skewness using the training data set. Optimal cutpoints for diagnosis of chordoma were RI > 0.015; FI < 0.25; mean ADC < 1.7 × 10-3 mm2/s; and skewness >0.177. The optimal decision tree relied on FI. In a secondary analysis, significant differences (P <.00047) in chondrosarcoma versus chordoma were found in 18 of 107 radiomic features, including six first-order and 12 high-order features.CONCLUSIONThe novel ADC index, FI, in addition to ADC mean, skewness, and 12 high-order radiomic features, could help differentiate sacral chordomas from chondrosarcomas.