Deteksi dan Klasifikasi Osteoartritis Lutut pada Citra X-ray Menggunakan ResNet-50 dan DenseNet-121 dengan Visualisasi Berbasis Web

Abstract

Knee osteoarthritis (OA) is one of the most common types of joint disorders in the elderly, characterized by pain and limited mobility. This study aims to develop a hybrid deep learning model to detect and classify the severity of knee OA in X-ray images using a combination of ResNet-50 and DenseNet-121 architecture. The severity is determined based on the Kellgren-Lawrence (KL) grading system, which consists of five grades (0-4). The dataset used was obtained from Roboflow and Dr. Pirngadi General Hospital Medan (RSUD Dr. Pirngadi Medan), with a total of 5,353 images after preprocessing. The model was trained using transfer learning and fine-tuning approaches, then integrated into a Streamlit-based web application. The model performance was evaluated using accuracy, precision, recall, and f1-score metrics. The evaluation resulted in an accuracy value of 60.1%, an average precision of 61.7%, an average recall of 59.5%, and an average f1-score of 60.2%. The model experienced overfitting and bias toward grade 4 (severe), and was unable to accurately classify other grades. This issue was caused by the limited size of the dataset and the model being too complex for a small dataset. Although the accuracy obtained was not optimal, the system has potential as an early-stage tool for knee OA detection and classification, and can be optimized through model training with more balanced and diverse dataset.