Recognition of Knee Osteoarthritis (KOA) Using YOLOv2 and Classification Based on Convolutional Neural Network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Osteoarthritis Research

Background:

• Knee osteoarthritis (KOA) is a leading cause of disability, necessitating early diagnosis for effective treatment and to prevent knee replacement.
• Manual diagnosis of KOA from radiographic images is subjective, time-consuming, and prone to errors.
• Automated methods are essential for accurate and efficient KOA detection.

Purpose of the Study:

• To develop and validate a computer-aided framework for the classification and localization of knee osteoarthritis (KOA) using radiographic images.
• To enhance the accuracy and speed of KOA diagnosis through advanced feature extraction and machine learning techniques.

Main Methods:

• Radiographic images were converted to 3D, and Local Binary Pattern (LBP) features were extracted and reduced using Principal Component Analysis (PCA).
• Deep features were extracted using Alex-Net and Dark-net-53, with further dimensionality reduction via PCA.
• Features from LBP and deep learning models were serially fused and classified using a 10-fold cross-validation approach.
• A localization model combining Open Exchange Neural Network (ONNX) and YOLOv2 was developed and trained.

Main Results:

• The classification model achieved an accuracy of 90.6% for grading knee osteoarthritis.
• The localization model demonstrated a mean Average Precision (mAP) of 0.98 for identifying KOA in images.
• The proposed framework outperformed existing methods in experimental analyses.

Conclusions:

• The developed framework offers a robust and accurate solution for the automated classification and localization of knee osteoarthritis.
• This AI-driven approach has the potential to significantly improve early KOA diagnosis and patient management.
• The study highlights the efficacy of fused feature extraction and advanced deep learning models in medical image analysis for osteoarthritis.