Dilated deeply supervised networks for hippocampus segmentation in MRI

TL;DR

This paper introduces a modified 3D U-Net with dilated convolutions and deep supervision for hippocampus segmentation in MRI, improving accuracy in early Alzheimer's Disease diagnosis.

Contribution

It presents a novel variant of 3D U-Net that incorporates dilated convolutions and deep supervision for enhanced multi-scale feature integration.

Findings

Outperforms conventional methods in segmentation accuracy

Effective in hippocampus head and body segmentation

Validated on MICCAI 2018 dataset

Abstract

Tissue loss in the hippocampi has been heavily correlated with the progression of Alzheimer's Disease (AD). The shape and structure of the hippocampus are important factors in terms of early AD diagnosis and prognosis by clinicians. However, manual segmentation of such subcortical structures in MR studies is a challenging and subjective task. In this paper, we investigate variants of the well known 3D U-Net, a type of convolution neural network (CNN) for semantic segmentation tasks. We propose an alternative form of the 3D U-Net, which uses dilated convolutions and deep supervision to incorporate multi-scale information into the model. The proposed method is evaluated on the task of hippocampus head and body segmentation in an MRI dataset, provided as part of the MICCAI 2018 segmentation decathlon challenge. The experimental results show that our approach outperforms other conventional methods in terms of different segmentation accuracy metrics.