Enhanced x-ray knee osteoarthritis classification: a multi-classification approach using MambaOut and latent...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthopedics

Background:

• Knee Osteoarthritis (KOA) is a widespread degenerative joint disease.
• Accurate KOA severity classification is vital for patient diagnosis and treatment.
• Current classification methods face challenges with sample imbalance across different grades.

Purpose of the Study:

• To develop a novel multiclassification algorithm for X-ray KOA severity.
• To enhance KOA classification performance using MambaOut and Latent Diffusion Models (LDM).
• To address and mitigate sample imbalance issues in KOA datasets.

Main Methods:

• Implemented a novel multiclassification algorithm integrating MambaOut and LDM.
• Utilized LDM for AI-generated data synthesis to augment minority classes.
• Optimized the autoencoder's loss function and incorporated pathological labels within the LDM framework.

Main Results:

• Achieved an average accuracy of 86.3% in the four-classification task.
• Obtained an average precision of 85.3% and an F1 score of 0.855.
• Reduced mean absolute error to 14.7%, outperforming existing advanced methods.

Conclusions:

• The proposed MambaOut and LDM-based algorithm significantly advances KOA classification.
• AI-generated data synthesis effectively addresses sample imbalance in medical imaging.
• This integration of advanced neural networks and generative models shows great potential for medical image analysis.