Beyond the algorithms: Challenges in predicting knee osteoarthritis pain from a machine learning point of view.
Authors
Affiliations (5)
Affiliations (5)
- Research Unit of Health Sciences and Technology, Faculty of Medicine, University of Oulu, Finland. Electronic address: [email protected].
- Orthopaedics, Department of Clinical Sciences Lund Faculty of Medicine, Lund University, Sweden.
- Center for Treatment of Rheumatic and Musculoskeletal Diseases (REMEDY), Diakonhjemmet Hospital, Oslo, Norway; Faculty of Medicine, University of Oslo, Norway.
- Department of Orthopaedics, Faculty of Medicine, Juntendo University, Japan.
- Research Unit of Health Sciences and Technology, Faculty of Medicine, University of Oulu, Finland; Department of Diagnostic Radiology, Oulu University Hospital, Oulu, Finland.
Abstract
Knee osteoarthritis (OA) presents complex pain experiences often poorly correlated with structural imaging findings. Machine learning (ML) and deep learning (DL) hold promise for predicting and identifying both current and future pain and personalizing interventions, but translating these insights into clinical practice remains challenging. This study highlights key challenges in applying ML and DL to predict knee OA pain, with a focus on the complex relationship between imaging features and subjective pain experiences. While not based on a systematic search or formal data extraction, the review aims to provide an informed overview of current issues and research directions. Key challenges include the limited availability of longitudinal data, small and imbalanced datasets, and the difficulty of aligning multimodal data over time. Existing pain assessment tools often fail to capture the dynamic and multidimensional nature of pain. In addition, many studies underrepresent diverse populations and neglect psychosocial factors, which weakens model robustness and limits generalizability. Future research should integrate longitudinal and real-time data (e.g., wearables) and promote diverse, collaborative datasets to bridge the gap between ML-based predictions and clinical applicability in knee OA pain management.