Clinical utility of foundation models in musculoskeletal MRI for biomarker fidelity and predictive outcomes

TL;DR

The paper presents a modular, open-source framework that converts routine musculoskeletal MRI into standardized quantitative biomarkers using promptable foundation segmentation models (SAM, MedSAM, SAM2) fine-tuned across 12 heterogeneous datasets. It demonstrates that biomarker measurements derived from these segmentations achieve high concordance with expert references and enable clinically useful applications, including a three-stage knee MRI triage system and longitudinal risk modeling for knee replacement and incident OA using 48-month landmark analyses. The approach emphasizes biomarker fidelity and a pipeline- and model-agnostic architecture to support scalable deployment and independent validation, showing how automated measurements can drive both current workflow efficiency and future precision medicine. The results indicate robust segmentation performance across tissues and protocols, with strong biomarker agreement (ICC up to 0.996), meaningful reductions in radiologist workload, and clinically relevant predictive power, underscoring the potential of foundation models to operationalize precision MSK imaging.

Abstract

Precision medicine in musculoskeletal imaging requires scalable measurement infrastructure. We developed a modular system that converts routine MRI into standardized quantitative biomarkers suitable for clinical decision support. Promptable foundation segmenters (SAM, SAM2, MedSAM) were fine-tuned across heterogeneous musculoskeletal datasets and coupled to automated detection for fully automatic prompting. Fine-tuned segmentations yielded clinically reliable measurements with high concordance to expert annotations across cartilage, bone, and soft tissue biomarkers. Using the same measurements, we demonstrate two applications: (i) a three-stage knee triage cascade that reduces verification workload while maintaining sensitivity, and (ii) 48-month landmark models that forecast knee replacement and incident osteoarthritis with favorable calibration and net benefit across clinically relevant thresholds. Our model-agnostic, open-source architecture enables independent validation and development. This work validates a pathway from automated measurement to clinical decision: reliable biomarkers drive both workload optimization today and patient risk stratification tomorrow, and the developed framework shows how foundation models can be operationalized within precision medicine systems.

Figures (14)

• Figure 1: Musculoskeletal MRI segmentation study design.
• Figure 2: Summary of dataset composition, subject demographics, and imaging protocol characteristics.
• Figure 3: Zero-shot and fine-tuned performance of SAM-family models across musculoskeletal MRI.
• Figure 4: Analysis of MRI acquisition parameters and segmentation agreement for fine-tuned SAM models in musculoskeletal MRI.
• Figure 5: Comparative evaluation of segmentation models using ground truth and automated bounding box prompts across musculoskeletal MRI datasets.