Performance Evaluation of Vanilla, Residual, and Dense 2D U-Net Architectures for Skull Stripping of Augmented 3D T1-weighted MRI Head Scans

TL;DR

This paper compares different 2D U-Net architectures for skull stripping in MRI scans, demonstrating that Dense U-Net achieves superior accuracy by enhancing feature reuse and enabling shallower models.

Contribution

It introduces a comparative analysis of Vanilla, Residual, and Dense 2D U-Net architectures specifically for skull stripping, highlighting the advantages of Dense U-Net.

Findings

Dense U-Net achieves 99.75% accuracy on test data.

Dense interconnections promote feature reuse and allow shallower models.

Dense U-Net outperforms Vanilla and Residual architectures in accuracy.

Abstract

Skull Stripping is a requisite preliminary step in most diagnostic neuroimaging applications. Manual Skull Stripping methods define the gold standard for the domain but are time-consuming and challenging to integrate into processing pipelines with a high number of data samples. Automated methods are an active area of research for head MRI segmentation, especially deep learning methods such as U-Net architecture implementations. This study compares Vanilla, Residual, and Dense 2D U-Net architectures for Skull Stripping. The Dense 2D U-Net architecture outperforms the Vanilla and Residual counterparts by achieving an accuracy of 99.75% on a test dataset. It is observed that dense interconnections in a U-Net encourage feature reuse across layers of the architecture and allow for shallower models with the strengths of a deeper network.