A Transfer Learning Assisted 3D Deep Learning Method For Knee Osteoarthritis Detection

The invention introduces a method for detecting and determining knee osteoarthritis (OA) from medical imaging, specifically using a transfer learning assisted 3D deep learning approach. The process involves obtaining magnetic resonance imaging (MRI) volumes, converting them to the Neuroimaging Informatics Technology Initiative (NIfTI) standard, and transforming them to a standardized format through resizing, Z-normalizing, and standardizing. Feature extraction is then performed using different neural networks. The method trains and compares MRI volumes using both pretrained and randomly assigned initial weights, utilizing pretrained weights as the initial weights for the training process of 3D models. The performance of these models is assessed based on classification metrics to effectively detect knee osteoarthritis.