Diffusion-weighted MRI radiomics of spine bone tumors: feature stability and machine learning-based classification...

Area of Science:

• Radiology
• Oncology
• Machine Learning

Background:

• Spine bone tumors require accurate classification for effective treatment.
• Radiomics offers a quantitative approach to medical imaging analysis.
• Differentiating benign from malignant spine tumors is clinically significant.

Purpose of the Study:

• To assess the stability of radiomic features from spine bone tumors.
• To evaluate the performance of machine learning models for tumor classification using radiomics.
• To utilize diffusion- and T2-weighted MRI for radiomic analysis.

Main Methods:

• Retrospective analysis of 101 spine bone tumor cases (benign, primary malignant, metastatic).
• Manual segmentation of tumor volumes on T2-weighted MRI and radiomic analysis on ADC maps.
• Feature stability assessment using intraclass correlation coefficient (ICC) and machine learning (SVM) for classification.

Main Results:

• 76.4% of radiomic features exhibited good stability.
• An 8-feature SVM model achieved 78% sensitivity, 68% specificity, 76% accuracy, and 0.78 AUC for tumor classification.
• The model demonstrated effective discrimination between benign and malignant spine tumors.

Conclusions:

• Radiomic features from T2- and diffusion-weighted MRI are valuable for classifying spine bone tumors.
• The developed SVM classifiers show potential for clinical application.
• Radiomic features extracted from spine bone tumors exhibit high reproducibility.