Comparing Deep Learning strategies for paired but unregistered multimodal segmentation of the liver in T1 and T2-weighted MRI

TL;DR

This paper evaluates various deep learning strategies for liver segmentation in paired T1 and T2 MRI images, focusing on registration, multi-task training, and loss functions, revealing that most methods perform similarly except for multi-task approaches.

Contribution

The study systematically compares multiple deep learning strategies for multimodal liver segmentation, highlighting the impact of registration, multi-task training, and loss functions.

Findings

Most methods achieved similar performance levels.

Multi-task training performed poorly compared to other strategies.

Choice of loss function influenced segmentation quality.

Abstract

We address the problem of multimodal liver segmentation in paired but unregistered T1 and T2-weighted MR images. We compare several strategies described in the literature, with or without multi-task training, with or without pre-registration. We also compare different loss functions (cross-entropy, Dice loss, and three adversarial losses). All methods achieved comparable performances with the exception of a multi-task setting that performs both segmentations at once, which performed poorly.