Opportunistic Bone-Loss Screening from Routine Knee Radiographs Using a Multi-Task Deep Learning Framework with Sensitivity-Constrained Threshold Optimization

2026-04-22

AI Summary

Osteoporosis is frequently underdiagnosed due to insufficient screening, and the gold-standard dual-energy X-ray absorptiometry (DXA) has limited accessibility. This work proposes STR-Net, a multi-task deep learning model that leverages routine knee radiographs to simultaneously screen for bone loss, classify its severity, and estimate T-scores—all without requiring additional imaging. The approach introduces a task-aware representation routing module and a sensitivity-constrained threshold optimization strategy, integrating clinical variables into a shared backbone network for weakly coupled T-score regression. Experimental results demonstrate strong performance: an AUROC of 0.933 for screening (sensitivity: 0.904), an AUROC of 0.898 for severity grading, and a correlation coefficient of 0.801 between estimated and DXA-derived T-scores, achieving high sensitivity alongside effective multi-task synergy.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Machine Learning: Multi-instance/Multi-view Learning
• Search and Optimization: Evaluation and Analysis

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Background: Osteoporosis and osteopenia are often undiagnosed until fragility fractures occur. Dual-energy X-ray absorptiometry (DXA) is the reference standard for bone mineral density (BMD) assessment, but access remains limited. Knee radiographs are obtained at high volume for osteoarthritis evaluation and may offer an opportunity for opportunistic bone-loss screening. Objective: To develop and evaluate a multi-task deep learning system for opportunistic bone-loss screening from routine knee radiographs without additional imaging or patient visits. Methods: We developed STR-Net, a multi-task framework for single-channel grayscale knee radiographs. The model includes a shared backbone, global average pooling feature aggregation, a shared neck, and a task-aware representation routing module connected to three task-specific heads: binary screening (Normal vs. Bone Loss), severity sub-classification (Osteopenia vs. Osteoporosis), and weakly coupled T-score regression with optional clinical variables. A sensitivity-constrained threshold optimization strategy (minimum sensitivity >= 0.86) was applied. The dataset included 1,570 knee radiographs, split at the patient level into training (n=1,120), validation (n=226), and test (n=224) sets. Results: On the held-out test set, STR-Net achieved an AUROC of 0.933, sensitivity of 0.904, specificity of 0.773, and AUPRC of 0.956 for binary screening. Severity sub-classification achieved an AUROC of 0.898. The T-score regression branch showed a Pearson correlation of 0.801 with DXA-measured T-scores in a pilot subset (n=31), with MAE of 0.279 and RMSE of 0.347. Conclusions: STR-Net enables single-pass bone-loss screening, severity stratification, and quantitative T-score estimation from routine knee radiographs. Prospective clinical validation is needed before deployment.

Problem

• osteoporosis
• bone-loss screening
• knee radiographs
• opportunistic screening
• bone mineral density

Innovation

• multi-task deep learning
• opportunistic screening
• sensitivity-constrained optimization
• bone mineral density estimation
• knee radiographs