Computer aided diagnostic model for knee osteoarthritis: A multi-modal feature regression approach

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Orthopedics

Background:

• X-ray imaging is essential for diagnosing knee osteoarthritis (KOA).
• Computer-assisted diagnostic models enhance diagnostic accuracy by reducing subjectivity.
• Continuous improvement of KOA diagnostic models is crucial for effective clinical treatment.

Purpose of the Study:

• To introduce a novel computer-aided diagnostic model for KOA.
• To improve the accuracy and efficiency of KOA diagnosis using X-ray imaging.
• To integrate multi-modal features for enhanced diagnostic performance.

Main Methods:

• Developed a computer-aided diagnostic model for KOA utilizing multi-modal feature regression on X-ray images.
• Extracted image content-based features (bone gap, bone skin thickness, bone mass) and integrated medical information (age, gender, surgical history).
• Employed support vector regression to establish the relationship between diagnostic features and Kellgren-Lawrence (K-L) classification for KOA severity.

Main Results:

• Validated the model on the NDKY-N2H knee X-ray image database (1200 images).
• Achieved over 98.42% accuracy in identifying KOA and 85.06% accuracy in K-L grading for KOA severity.
• Demonstrated improved diagnostic accuracy through image preprocessing and patient information integration.

Conclusions:

• Accurate KOA diagnosis is vital for improving patient health outcomes.
• The proposed multi-modal feature regression model shows significant promise for reliable and efficient KOA diagnosis.
• This innovative approach leverages both image content and medical information for enhanced diagnostic capabilities in X-ray imaging.