Towards trustworthy artificial intelligence in musculoskeletal medicine: A narrative review on uncertainty quantification
Knee Surgery, Sports Traumatology, Arthroscopy · 2025
DOI: 10.1002/ksa.12737 ↗Kurz gefasst
Übersichtsarbeit in der Zeitschrift Knee Surgery, Sports Traumatology, Arthroscopy, 2025. Verfasst von Amir M Vahdani und 3 weiteren Autor:innen. Beteiligte Einrichtung: Tehran University of Medical Sciences, School of Medicine, Tehran, Iran. Hauptthemen laut MEDLINE: Deep Learning, Artificial Intelligence, Musculoskeletal Diseases, Musculoskeletal System. Die Arbeit wurde bisher 7-mal zitiert (Crossref) und nennt 123 Literaturangaben.
Abstract (Originalsprache)
Introduction
Deep learning (DL) models have achieved remarkable performance in musculoskeletal (MSK) medical imaging research, yet their clinical integration remains hindered by their black‐box nature and the absence of reliable confidence measures. Uncertainty quantification (UQ) seeks to bridge this gap by providing each DL prediction with a calibrated estimate of uncertainty, thereby fostering clinician trust and safer deployment.
Methods
We conducted a targeted narrative review, performing expert‐driven searches in PubMed, Scopus, and arXiv and mining references from relevant publications in MSK imaging utilizing UQ, and a thematic synthesis was used to derive a cohesive taxonomy of UQ methodologies.
Results
UQ approaches encompass multi‐pass methods (e.g., test‐time augmentation, Monte Carlo dropout, and model ensembling) that infer uncertainty from variability across repeated inferences; single‐pass methods (e.g., conformal prediction, and evidential deep learning) that augment each individual prediction with uncertainty metrics; and other techniques that leverage auxiliary information, such as inter‐rater variability, hidden‐layer activations, or generative reconstruction errors, to estimate confidence. Applications in MSK imaging, include highlighting uncertain areas in cartilage segmentation and identifying uncertain predictions in joint implant design detections; downstream applications include enhanced clinical utility and more efficient data annotation pipelines.
Conclusion
Embedding UQ into DL workflows is essential for translating high‐performance models into clinical practice. Future research should prioritize robust out‐of‐distribution handling, computational efficiency, and standardized evaluation metrics to accelerate the adoption of trustworthy AI in MSK medicine. Level of Evidence Not applicable.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Tehran University of Medical Sciences, School of Medicine, Tehran, Iran.
- Department of Biomedical Informatics, Harvard Medical School, Boston, Massachusetts, USA.
- Hospital for Special Surgery, New York, New York, USA.
ORCID-Profile: Amir M Vahdani, Moein Shariatnia, Pranav Rajpurkar, Ayoosh Pareek
Zitieren
APA
Vahdani, A. M., Shariatnia, M., Rajpurkar, P., & Pareek, A. (2025). Towards trustworthy artificial intelligence in musculoskeletal medicine: A narrative review on uncertainty quantification. Knee Surgery, Sports Traumatology, Arthroscopy. https://doi.org/10.1002/ksa.12737
Vancouver
Vahdani AM, Shariatnia M, Rajpurkar P, Pareek A. Towards trustworthy artificial intelligence in musculoskeletal medicine: A narrative review on uncertainty quantification. Knee Surgery, Sports Traumatology, Arthroscopy. 2025. doi:10.1002/ksa.12737
BibTeX
@article{Vahdani2025,
author = {Vahdani, Amir M and Shariatnia, Moein and Rajpurkar, Pranav and Pareek, Ayoosh},
title = {Towards trustworthy artificial intelligence in musculoskeletal medicine: A narrative review on uncertainty quantification},
journal = {Knee Surgery, Sports Traumatology, Arthroscopy},
year = {2025},
doi = {10.1002/ksa.12737}
}
EUVIMED