Lightweight Deep Learning for Knee Osteoarthritis Analysis: A Mobilenet Perspective

Most of the people globally suffer from osteoarthritis of the knees. It is a common degenerative joint disorder that requires early identification and precise grading to inform effective therapy plans. A frequent aspect of traditional medical procedures has manual radiography analysis, which is laborious and prone to errors. This paper offers a portable and effective deep learning-based method for KOA the identification and assessment that makes use of MobileNet. MobileNet, a type of convolutional neural network specifically intended for mobile or embedded devices, is used to computationally effectively extract discriminative features from knee images. The proposed model was tested and trained using a publicly available OAI dataset, exhibiting competitive accuracy in KOA classification as well as adaptability and severity evaluation. Through the use of techniques like data enhancement and transfer learning, the model maintains minimal latency and resource requirements while demonstrating great performance. The results demonstrate the promise of MobileNet as a scalable real-time KOA diagnostic tool, particularly in resource-constrained environments like as distant clinics or mobile medical units. For multiclass classification across several KOA severity categories, the MobileNet model obtains an accuracy of 85.8%, while for binary classification of KOA presence, it reaches 93.9%. Future we will focus on improving the model's performance by including more datasets, sophisticated preprocessing methods, and hybrid architectures.