Advanced search


Search results        Search results      Copy URL to E-Mail

Supporting Early Osteopathic Treatment Through Automated Grading of Knee Osteoarthritis
Lois, D. [1]

Journal: Journal of Osteopathic Medicine Date: 2025/12, 125(12):Pages: A714–715. doi: Subito , type of study: descriptive study

Full text    (https://www.degruyterbrill.com/document/doi/10.1515/jom-2025-2000/html)

Keywords:

descriptive study [65]
diagnosis [382]
grading [1]
knee osteoarthritis [6]
machine learning [1]
models [54]

Abstract:

Context: Knee osteoarthritis (OA) is a progressive joint condition that leads to chronic pain, limited mobility, and decreased quality of life. The Kellgren-Lawrence (K&L) grading system is widely used to assess OA severity on radiographs, but manual grading is time-consuming and may vary between clinicians. In osteopathic medicine, early diagnosis is essential to guide personalized interventions, including osteopathic manipulative treatment (OMT), which may improve joint function and reduce long-term complications. This study addresses the need for a reliable, automated tool to assist osteopathic physicians in quickly identifying OA severity and initiating timely care. Objective: To develop and validate a deep learning model using EfficientNet-B0 to automatically classify knee OA severity from radiographic images and support early osteopathic evaluation and intervention. Methods: This machine learning study used a public, de-identified dataset of 1,650 knee radiographs labeled by K&L grade (0–4). No inclusion or exclusion criteria were applied due to the dataset’s pre-labeled nature. Images were preprocessed with grayscale channel replication, resizing (224×224 pixels), normalization, and data augmentation (horizontal flipping and 5° rotation). An EfficientNet-B0 model pretrained on ImageNet was fine-tuned for five-class classification using cross-entropy loss, the Adam optimizer, and a One Cycle learning rate scheduler. Model evaluation was performed using three repetitions of five-fold stratified cross-validation (15 total runs), which preserved class balance in each split. Each fold was trained for up to 20 epochs with early stopping based on validation accuracy. Accuracy and confusion matrix analysis were used to assess performance. This model is designed to support osteopathic physicians by enabling earlier and more consistent detection of OA severity, which may inform timely use of osteopathic manipulative treatment and other interventions. Results: The model achieved a mean validation accuracy of 87.39% (standard deviation 1.36%) in Repetition 1, and 86.18% (standard deviations 1.44% and 1.50%) in Repetitions 2 and 3, respectively. The overall average validation accuracy across all folds and repetitions was 86.59%, with a global standard deviation of 1.55%. Performance was consistent across trials, indicating strong generalizability. Confusion matrix analysis revealed the highest accuracy in identifying Grade 0 (normal) and Grade 4 (severe) OA cases. Intermediate grades, particularly Grades 1 and 2, showed more overlap and were more difficult to distinguish. Training and validation accuracy curves showed steady learning, while training loss consistently declined across epochs. Conclusion: This study demonstrates that a deep learning model can accurately and reliably classify knee osteoarthritis severity on radiographs using the Kellgren-Lawrence grading scale. The tool’s consistent performance suggests it could assist osteopathic physicians in identifying OA earlier and more objectively, enabling earlier use of osteopathic manipulative treatment (OMT) and other care strategies. By facilitating faster and more reliable grading, this model can support timely referrals for OMT, particularly in primary care or triage settings where radiographic access precedes specialist consultation. Early intervention with OMT may improve joint mobility, reduce chronic pain progression, and potentially delay or reduce the need for pharmacologic or surgical interventions. In the long term, this approach could enhance patient functional outcomes and support a preventive, whole-person osteopathic philosophy of care. Future research should explore integration into clinical workflows, assess patient-reported outcomes following AI-guided treatment initiation, and validate performance in more diverse populations.

Search results      Copy URL to E-Mail

 
 
 






  • Impressum • Legal notice • Datenschutz •


ostlib.de/data_gzsydhwkapuqmjecftxb



Supported by

OSTLIB recommends