Optimized Osteoarthritis Detection Using MobileNetV3: Advancing Medical Imaging Analysis

Affected millions of individuals globally, osteoarthritis (OA) is a degenerative joint condition defined by pain, stiffness, and reduced mobility. Early diagnosis is essential for both effective intervention and treatment to delay or avoid significant effects including joint replacement. This paper explores the use of medical imaging the relevance of a tailored MobileNetV3 model for automatic osteoarthritis diagnosis. Model training and evaluation were conducted using a dataset comprising 768 tagged X-ray pictures from Kaggle including 451 cases of osteoarthritis and 297 normal cases. To improve model generalization, the dataset was pre-processed including augmentation, normalizing, and resizing. By means of dense and dropout layers, the MobileNetV3 architecture attained a classification accuracy of 83%. Performance measures for the diagnosis of osteoarthritis with an F1-score of 0.86 and for normal cases with an F1-score of 0.79 revealed excellent results. Emphasizing the reliability of the model, stable accuracy graphs coupled with consistent decline in training and validation losses show. Providing possible application on restricted resources, the proposed approach presents a reasonable, scalable, and affordable approach for early osteoarthritis detection. This work underlines the huge potential of deep learning in medical diagnostics by providing a strong diagnostic tool, therefore paving the road for improved patient outcomes and additional clinical usage in automated illness classification. Future projects will focus on raising sensitivity to normal cases to raise diagnosis accuracy.