Abstract

Knee osteoarthritis, affecting 365 million people globally, demands innovative solutions for early diagnosis and effective management. This study develops a generative AI-supported system to enhance clinical care, leveraging a Retrieval-Augmented Generation framework. The system integrates a curated corpus of knee osteoarthritis-specific guidelines with the DeepSeek large language model to provide context-aware diagnostic support and personalized rehabilitation guidance. A web-based interface, the Knee Therapy Research Assistant, delivers intuitive access for clinicians and patients, offering KL grading interpretations and exercise protocols like quadriceps-strengthening routines with animated diagrams. Preliminary demonstrations showcase its usability and low-latency response (1.2 seconds), driven by hybrid search capabilities. The system addresses gaps in patient self-management and diagnostic accuracy, but requires clinical validation. This work highlights the potential of AI as a co-pilot in orthopedic care, with future efforts aimed at real-world testing and expansion to other musculoskeletal conditions.

References

[1] Musumeci G. Functional anatomy in knee osteoarthritis: patellofemoral joint and tibiofemoral joint. J Funct Morphol Kinesiol 2017;2:8.

[2] World Health Organization. Osteoarthritis. 2023. Available from: https://www.who.int/news-room/fact-sheets/detail/osteoarthritis [Accessed 18 April 2025].

[3] Dantas LO, Salvini TF, McAlindon TE. Knee osteoarthritis: key treatments and implications for physical therapy. Braz J Phys Ther 2021;25:135-46.

[4] Roos EM, Arden NK. Strategies for the prevention of knee osteoarthritis. Nat Rev Rheumatol 2016;12:92-101.

[5] Lee LS, Chan PK, Wen C, Fung WC, Cheung A, Chan VWK, et al. Artificial intelligence in diagnosis of knee osteoarthritis and prediction of arthroplasty outcomes: a review. Arthroplasty 2022;4:16.

[6] Han T, Kather JN, Pedersoli F, Zimmermann M, Keil S, Schulze-Hagen M, et al. Predicting osteoarthritis progression in radiographs via unsupervised representation learning. arXiv preprint arXiv:2104.05923, 2021.

[7] Touahema S, Zaimi I, Zrira N, Ngote MN. How can artificial intelligence identify knee osteoarthritis from radiographic images with satisfactory accuracy?: a literature review for 2018–2024. Appl Sci 2024;14:6333.

[8] Du K, Li A, Zuo QH, Zhang CY, Guo R, Chen P, et al. Comparing AI-generated and clinician-created personalized self-management guidance for knee osteoarthritis patients: a blinded observational study (preprint). JMIR Preprints 2024. doi:10.2196/preprints.67830.

[9] Zhang C, Liu S, Zhou X, Zhou S, Tian Y, Wang S, et al. Examining the role of large language models in orthopedics: systematic review. J Med Internet Res 2024;26:e59607.

[10] Kuroiwa T, Sarcon A, Ibara T, Yamada E, Yamamoto A, Tsukamoto K, et al. The potential of ChatGPT as a self-diagnostic tool in common orthopedic diseases: exploratory study. J Med Internet Res 2023;25:e47621.

[11] Yang J, Ardavanis KS, Slack KE, Fernando ND, Della Valle CJ, Hernandez NM. Chat generative pretrained transformer (ChatGPT) and Bard: artificial intelligence does not yet provide clinically supported answers for hip and knee osteoarthritis. J Arthroplasty 2024;39:1184-90.

[12] Li J, Gao X, Dou T, Gao Y, Zhu W. Assessing the performance of GPT-4 in the field of osteoarthritis and orthopaedic case consultation. medRxiv preprint 2023. doi:10.1101/2023.08.01.23293571.

[13] Lewis P, Perez E, Piktus A, Petroni F, Karpukhin V, Goyal N, et al. Retrieval-augmented generation for knowledge-intensive NLP tasks. Adv Neural Inf Process Syst 2020;33:9459-74.

[14] Jin Q, Kim W, Chen Q, Comeau DC, Yeganova L, Wilbur WJ, et al. MedCPT: contrastive pre-trained transformers with large-scale PubMed search logs for zero-shot biomedical information retrieval. arXiv preprint arXiv:2307.00589, 2023.

[15] Zilliz Blog. Review of hybrid search in Milvus. 2024. Available from: https://zilliz.com/blog/hybrid-search-in-milvus [Accessed 18 April 2025].

[16] Chen T, Rocktäschel T, Riedel S. Out-of-domain semantics to the rescue! Zero-shot hybrid retrieval models. arXiv preprint arXiv:2201.10582, 2022.

[17] DeepSeek-VL Team. DeepSeek LLM: a scalable and open-source language model series. arXiv preprint arXiv:2401.10660, 2024