Bridging Clinical Expertise and AI: Deep Learning for Automated Knee Osteoarthritis Diagnosis

Abstract

Knee osteoarthritis affects the knee sections, and its severity gets worse over the course of 10 to 15 years. Knee osteoarthritis makes it tougher for older people to stay active and do what they need to do. The skill level of doctors has a big effect on how they analyze visual exams. Finding problems early on requires methods that provide quick operation and accurate detection while yet being cost- effective. Deep learning (DL) uses convolutional neural networks (CNN) to find the best way to diagnose knee osteoarthritis. Convolutional neural networks need a lot of input data to learn. An automated X-ray analysis that uses deep learning technology makes it easier to extract features and shortens the time it takes to train models. Researchers have found that a deep learning system that uses a trained VGG16 network on five-class knee X-rays can use SoftMax to accurately tell the difference between normal, questionable, mild, moderate, and serious issues. Two deep convolutional neural networks (CNNs) are used by an automated system to grade patients with knee osteoarthritis (OA) using the Kellgren-Lawrence (KL) criteria. The study included 1650 different knee X-ray scans, which were split into two separate groups. The dataset has 514 normal cases, 477 questionable cases, 232 mild cases, 221 moderate cases, and 206 severe cases of knee osteoarthritis. The suggested model for finding knee osteoarthritis (OA) using X-ray radiographs has a classification accuracy of over 95%. The training results were 95% accurate, and the validation results were 87% accurate.

Ähnliche Arbeiten