Network Structure Optimization Using Compound Scaling in 3D MRI Structural Image Segmentation

Abstract

In recent years, AI technology has been utilized in the medical field for diagnostic support and health management. This study aims to use Convolutional Neural Networks (CNN) for brain MRI image segmentation, optimizing the network based on Compound Scaling in EfficientNet. Using a Residual Network (ResNet) as the CNN model, we explored the optimal depth and width, finding that a structure with a slight emphasis on depth was the most effective.