Comparative Evaluation of Transfer Learning Models for Knee Osteoarthritis Severity Classification Using KL Grading Scale

Knee Osteoarthritis (KOA) is a prevalent and debilitating joint condition that affects millions of individuals worldwide. Early and accurate diagnosis of KOA severity is crucial for effective treatment planning and patient management. One widely accepted method for assessing KOA severity is the Kellgren-Lawrence (KL) grading scale, which categorizes KOA into different stages based on radiographic evidence from knee X-ray images. This study addresses the critical need for an efficient and accurate classification system for KOA severity, leveraging the power of deep learning and transfer learning techniques. The primary objective of this research work is to identify the most effective model for classifying KOA severity levels from knee X-ray images using the KL grading scale. This study has encountered significant challenges related to preprocessing of knee images and system computation time. Initially, knee images required intricate preprocessing steps to enhance data quality, and the computational time posed constraints on real-time applications. To overcome these challenges, this study has employed advanced techniques and conducted a comprehensive performance analysis of various deep learning models, including NasNetLarge, AlexNet, Inception V3, DenseNet, Mobile Net, and ResNet. These models have shown significant promise in various image classification tasks, making them compelling candidates for the task of KOA severity classification.