Opportunistic Osteoporosis Diagnosis via Texture-Preserving Self-Supervision, Mixture of Experts and Multi-Task Integration

2025-06-25

AI Summary

Osteoporosis opportunistic screening is hindered by DXA scarcity, underutilization of unlabeled vertebral CT data, cross-device/system domain shifts, and failure to integrate clinical priors (e.g., spatial BMD distribution). To address these challenges, we propose a CT-based diagnostic framework integrating self-supervised learning, gated mixture-of-experts (MoE), and multi-task learning. Specifically, radiomics-guided texture-preserving self-supervised pretraining enhances utilization of unlabeled vertebral CT volumes; device-invariant feature disentanglement coupled with a gated MoE architecture improves generalizability across centers and imaging devices; and joint vertebral localization and BMD regression explicitly models anatomy–density correlations. Evaluated across four clinical centers, our method significantly outperforms state-of-the-art approaches—achieving a 5.2% AUC improvement—and attains clinically deployable generalizability and screening accuracy.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Multi-modal Vision
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Osteoporosis, characterized by reduced bone mineral density (BMD) and compromised bone microstructure, increases fracture risk in aging populations. While dual-energy X-ray absorptiometry (DXA) is the clinical standard for BMD assessment, its limited accessibility hinders diagnosis in resource-limited regions. Opportunistic computed tomography (CT) analysis has emerged as a promising alternative for osteoporosis diagnosis using existing imaging data. Current approaches, however, face three limitations: (1) underutilization of unlabeled vertebral data, (2) systematic bias from device-specific DXA discrepancies, and (3) insufficient integration of clinical knowledge such as spatial BMD distribution patterns. To address these, we propose a unified deep learning framework with three innovations. First, a self-supervised learning method using radiomic representations to leverage unlabeled CT data and preserve bone texture. Second, a Mixture of Experts (MoE) architecture with learned gating mechanisms to enhance cross-device adaptability. Third, a multi-task learning framework integrating osteoporosis diagnosis, BMD regression, and vertebra location prediction. Validated across three clinical sites and an external hospital, our approach demonstrates superior generalizability and accuracy over existing methods for opportunistic osteoporosis screening and diagnosis.

Problem

• Underutilization of unlabeled vertebral CT data for osteoporosis diagnosis
• Systematic bias from device-specific DXA measurement discrepancies
• Insufficient integration of clinical knowledge like spatial BMD patterns

Innovation

• Self-supervised learning preserves bone texture
• Mixture of Experts enhances device adaptability
• Multi-task learning integrates diagnosis and prediction