Knee Osteoarthritis Severity Classification Using STM32 Microcontroller

Knee osteoarthritis (KOA) is a degenerative joint illness depicted by worsening of cartilage, resulting in pain, stiffness, and reduced mobility. Early prediction of knee OA prevents knee replacement and is essential for personalized treatment. This paper presents a deep learning based system for the computerized prediction of knee OA using clinical imaging data(X-rays). Key features extracted from imaging are utilized to train and validate the models, enabling accurate prediction of knee OA progression. For this KOA datasets with 3 grade classes (normal, moderate, severe) are trained and tested with MobileNetV1, EfficientNetBO0, Xception, DenseNet201, VGG16 and finally suggesting a high efficiency algorithm, which will be deployed on STM32 microcontroller through edge impulse. A comprehensive analysis will be taken which includes latency, ROM, RAM, accuracy taken in edge impulse with stm32 and the efficient algorithm that was trained and tested using traditional method will be compared and thus we can obtain results of KOA severity level in reduced time and computation.