A Novel Hybrid Approach Based on Deep CNN Features to Detect Knee Osteoarthritis.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthopedics

Background:

• Knee Osteoarthritis (KOA) significantly impacts adult quality of life due to decreased articular cartilage.
• KOA leads to joint pain, effusion, limited movement, and gait abnormalities.

Purpose of the Study:

• To develop a novel deep learning framework for early-stage KOA detection and classification.
• To accurately classify KOA into Kellgren-Lawrence (KL) grades I-IV.

Main Methods:

• Preprocessing of X-ray images and Region of Interest (ROI) extraction via segmentation.
• Hybrid feature extraction using Convolutional Neural Network (CNN) with Local Binary Patterns (LBP) and Histogram of Oriented Gradients (HOG).
• Classification using Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbour (KNN) on extracted features.

Main Results:

• Histogram of Oriented Gradients (HOG) features achieved approximately 97% accuracy for KOA detection and classification.
• The framework demonstrated effectiveness across all four KL grades of Knee Osteoarthritis.

Conclusions:

• Deep learning-based feature extraction and classification offer a promising approach for early and accurate KOA diagnosis.
• The proposed method, particularly using HOG features, can aid in timely intervention and management of Knee Osteoarthritis.