Background: Knee osteoarthritis (OA) is a chronic degenerative disease that originates in the synovial joint and progressively affects surrounding tissues, leading to biomechanical alterations in the lower extremities. Individuals with knee OA exhibit significant changes in gait parameters and plantar pressure distribution. The early diagnosis of OA through these altered gait parameters is of critical importance for implementing preventive treatment approaches. In recent years, machine learning techniques have helped to develop clinical prediction models. These algorithms are able to analyze complex datasets, identify key determinants of disease progression and aid in the diagnosis of various medical conditions.

Objectives: The aim of this study was to predict the presence of OA and the stage of OA with the help of machine learning by examining the clinical features and pedobarographic analysis of individuals.

Methods: 50 patients with knee OA (mean age= 53.35±1.21 years, mean BMI=31.69±0.93 kg/m2) and 50 healthy controls (mean age= 51.05±0.63 years, mean BMI=29.28±1.15 kg/m2) were included in the study. The clinical characteristics of individuals, such as age, height, weight, pain, and foot posture index, were evaluated. Plantar pressure distribution was evaluated using the Digital Biometry Scanning System and Milletrix software (DIASU, Italy). For each task, three basic steps were followed when designing the machine learning pipeline: - Normalization - Dimension reduction(K-best feature selection, Principal component analysis, Random projections) – Classifier (Random forest classification, K-Nearest neighbours classifier, Support Vector Machine, Gaussian processes, Multi-level perceptrons, Quadratic discriminant analysis, Gradient Boosting, Logistic regression). Each classification pipeline was trained and tested by splitting the dataset into training and test sets. Specifically, 70% of the subjects were included in the training set, while 30% of the data was reserved for the test set.

Results: The relationship between clinical findings and pedobarographic characteristics with the presence of OA and OA stages is given in the heat map (Figure 1). Machine learning highly predicts both the presence and staging of OA (Table 1).

Conclusion: OA diagnosis can be predicted by machine learning parameters using pedobarographic data frequently used in clinics. Machine learning is an assistive tool that can be used in clinical prediction model for OA diagnosis. It is important to detect knee OA from gait parameters and to take preventive measures before the disease progresses.

REFERENCES: [1] Chen J, Zheng Q, Lan Y, Li M, Lin L. Development and validation of a new nomogram for self-reported OA based on machine learning: a cross-sectional study. Sci Rep. 2025;15(1):827. Published 2025 Jan 4. doi:10.1038/s41598-024-83524-y.

[2] Tayfur B, Ritsche P, Sunderlik O, et al. Automatic Segmentation of Quadriceps Femoris Cross-Sectional Area in Ultrasound Images: Development and Validation of Convolutional Neural Networks in People With Anterior Cruciate Ligament Injury and Surgery. Ultrasound Med Biol. 2025;51(2):364-372. doi:10.1016/j.ultrasmedbio.2024.11.004.

Figure 1.

Correlation Between Input and Output Variables

AS: Affected side OS: Other side

The features in red depict the clinical data, while the features in blue indicate the pedobarographic data

*Pearson correlation test

Acknowledgements: NIL.

Disclosure of Interests: None declared.

© The Authors 2025. This abstract is an open access article published in Annals of Rheumatic Diseases under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ). Neither EULAR nor the publisher make any representation as to the accuracy of the content. The authors are solely responsible for the content in their abstract including accuracy of the facts, statements, results, conclusion, citing resources etc.