Can a single image processing algorithm work equally well across all phases of DCE-MRI?

TL;DR

This study investigates whether a single image processing algorithm can effectively handle all phases of DCE-MRI, revealing that pretraining with contrast-enhanced data and fine-tuning with non-contrast data improves generalization.

Contribution

It quantifies the impact of contrast changes on segmentation and registration tasks and proposes a pretraining and fine-tuning strategy for better generalization across DCE-MRI phases.

Findings

Pretraining with CE data and fine tuning with non-CE data yields the best generalization.

Contrast changes significantly affect segmentation and registration performance.

The approach can potentially improve deep learning models for DCE-MRI analysis.

Abstract

Image segmentation and registration are said to be challenging when applied to dynamic contrast enhanced MRI sequences (DCE-MRI). The contrast agent causes rapid changes in intensity in the region of interest and elsewhere, which can lead to false positive predictions for segmentation tasks and confound the image registration similarity metric. While it is widely assumed that contrast changes increase the difficulty of these tasks, to our knowledge no work has quantified these effects. In this paper we examine the effect of training with different ratios of contrast enhanced (CE) data on two popular tasks: segmentation with nnU-Net and Mask R-CNN and registration using VoxelMorph and VTN. We experimented further by strategically using the available datasets through pretraining and fine tuning with different splits of data. We found that to create a generalisable model, pretraining with CE data and fine tuning with non-CE data gave the best result. This interesting find could be expanded to other deep learning based image processing tasks with DCE-MRI and provide significant improvements to the models performance.