An Anatomically Guided and Optimization-Refined Radiomics Framework for Opportunistic Osteoporosis Assessment from...

Area of Science:

• Radiology
• Medical Imaging Analysis
• Biomedical Engineering

Background:

• Osteoporosis significantly contributes to vertebral compression fractures (VCFs).
• Dual-energy X-ray absorptiometry (DEXA) for bone mineral density (BMD) assessment is not always accessible in spine surgery.
• There is a need for alternative osteoporosis screening methods in routine imaging workflows.

Purpose of the Study:

• To develop and validate an automated radiomics framework for opportunistic osteoporosis assessment using lumbar spine MRI.
• To evaluate the framework's ability to identify osteoporosis-related bone quality changes.
• To explore the feasibility of estimating BMD from MRI-derived radiomic features.

Main Methods:

• Anatomically guided, optimization-refined radiomics pipeline for lumbar spine MRI (L1-L4).
• Hierarchical template matching for vertebral localization and ROI refinement using multiple constraints.
• Extraction of handcrafted radiomic features (statistical, textural, gradient, frequency, shape).
• Classification using Support Vector Classification (SVC) and NeurodynamicSVMRBFTanh.

Main Results:

• Robust and anatomically consistent vertebral localization achieved across diverse MRI data.
• NeurodynamicSVMRBFTanh framework demonstrated high performance: 85.2% accuracy, 100.0% sensitivity for osteoporosis screening.
• Exploratory analysis showed feasibility of estimating DEXA-derived BMD from MRI radiomics (15-20% MAPE).

Conclusions:

• Lumbar spine MRI-derived radiomics capture clinically relevant information for osteoporosis assessment.
• The proposed framework offers a practical, automated tool for opportunistic osteoporosis screening where DEXA is unavailable.
• This approach may enhance early detection and management of osteoporosis-related skeletal complications.