Adversarial Evolving Neural Network for Longitudinal Knee Osteoarthritis Prediction.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Knee osteoarthritis (KOA) prevalence has doubled, necessitating early diagnosis for effective management.
• Longitudinal KOA grading remains understudied, with limited exploration of disease progression knowledge.
• Accurate monitoring requires robust methods for tracking KOA severity over time.

Purpose of the Study:

• To propose a novel deep learning architecture, the adversarial evolving neural network (A-ENN), for longitudinal KOA grading.
• To effectively characterize KOA disease progression by incorporating domain knowledge.
• To improve the accuracy and reliability of long-term KOA severity assessment.

Main Methods:

• Developed a novel deep learning architecture: adversarial evolving neural network (A-ENN).
• Utilized convolution and deconvolution computations to capture disease progression patterns.
• Integrated an adversarial training scheme with a discriminator to generate evolution traces.
• Fused evolution traces (domain knowledge) with convolutional image representations for grading.

Main Results:

• Achieved an overall accuracy of 62.7% on the Osteoarthritis Initiative (OAI) dataset.
• Demonstrated performance across different time points: baseline (64.6%), 12-month (63.9%), 24-month (63.2%), 36-month (61.8%), and 48-month (60.2%).
• The proposed A-ENN method effectively learns and utilizes disease progression patterns for longitudinal grading.

Conclusions:

• The adversarial evolving neural network (A-ENN) shows promise for longitudinal knee osteoarthritis grading.
• Incorporating domain knowledge through evolution traces enhances grading accuracy.
• The A-ENN architecture offers a potential advancement in monitoring and managing KOA progression.