Knee Arthritis Classification Using Attention Mechanism

Abstract

Arthritis is the generic term for a group of inflammatory diseases that cause pain, stiffness,and swelling in the bones, muscles, and joints. Especially in the major joints like the knees,arthritis can be highly dangerous. Rheumatoid arthritis (RA), Osteoarthritis (OA), andOther types of arthritis can considerably affect people's way of daily life. Because ofsymptom similarities and uncertainties in diagnosis, detecting and classifying these typesusing X-ray images is a challenging task. There is a significant lack of knowledge regardingthe detection and classification of many types of arthritis because prior research hasprimarily concentrated on detecting individual arthritis diseases and they faces challengesin capturing fine-grained disease features and multiscale discriminative features. A lack ofqualified radiologists complicates the problem, especially in developing countries likeEthiopia, where it burdens healthcare providers and delays diagnosis. To overcome thesechallenges, in this study, proposed a multiscale attention deep learning approach thatincorporates attention mechanisms and multiscale feature extraction to enhance arthritisdetection and classification in X-ray images. With a dataset of X-ray images from anEthiopian local hospital from 2018 to 2024, evaluate the proposed model's effectiveness incomparison to other pretrained models. The proposed model's remarkable 0.995 accuracywas attained along with metrics for precision, recall, and F1-score that were all similarlyhigh. The results demonstrate that this approach outperforms from other models in terms ofarthritis detection and classification accuracy. By incorporating attention mechanisms, thisproposed method effectively captures multiscale fine-grained disease features present in X-ray images. This improvement in arthritis detection and classification can significantlycontribute to diagnosis and appropriate treatment decisions. These findings highlight themodel's effectiveness in correctly classifying multiple types of arthritis, addressing theurgent demand for advanced imaging techniques in areas where radiologists with thenecessary training are scarce. This study eventually showed that automated detection andclassification of arthritis was improved by integrating methods of attention and multi-scalefeature extraction, ultimately leading to better patient outcomes through prompt andaccurate diagnosis.