Diffusion-weighted MRI Radiomics Classification of Spinal Bone Tumors

Presentation Format: Oral presentation.

Purpose or Learning Objective: To classify spine bone tumors using diffusion- and T2-weighted magnetic resonance imaging radiomics-based machine learning.

Methods or Background: This retrospective study approved by the local ethics committee included 101 patients with histology-proven bone tumor of the spine (22 benign, 38 primary malignant, and 41 metastatic). All tumor volumes were manually segmented by a radiology resident experienced in musculoskeletal and oncologic imaging on morphological T2-weighted sequences. The same region of interest (ROI) was used to perform radiomic analysis on an apparent diffusion coefficient (ADC) map. A total of 1,702 radiomic features was considered. Feature stability was assessed through small geometric transformations of the ROIs mimicking multiple manual delineations. Intraclass correlation coefficient (ICC) quantified feature stability. Feature selection consisted of stability-based (ICC > 0.75) and significance-based selections (ranking features by decreasing Mann-Whitney p value). Class balancing was performed to oversample the minority (i.e., benign) class. Selected features were then used to train and test a support vector machine (SVM) to discriminate benign from malignant spine tumors using 10-fold cross validation.

Results or Findings: A total of 76.4% of the radiomic features were stable. The quality metrics for the SVM were evaluated as a function of the number of selected features. The radiomic model with the best performance and the lowest number of features for classifying tumor types included eight features. The metrics were 78% sensitivity, 68% specificity, 76% accuracy, and area under the curve 0.78.

Conclusion: SVM classifiers based on radiomic features extracted from T2- and diffusion-weighted imaging with an ADC map are promising for classification of spine bone tumors. Radiomic features of spine bone tumors show good reproducibility rates.

No conflict of interest has been declared by the author(s).

Publication History

Article published online:
03 June 2021