Generative Style Transfer for MRI Image Segmentation: A Case of Glioma Segmentation in Sub-Saharan Africa

TL;DR

This paper develops a deep learning-based brain tumor segmentation method tailored for Sub-Saharan Africa's MRI data, demonstrating comparable performance of 2D and 3D models and proposing strategies to improve accuracy using fine-tuning and neural style transfer augmentation.

Contribution

It introduces a robust segmentation approach for SSA MRI data, analyzing domain shift effects, comparing 2D and 3D models, and proposing novel data augmentation techniques to improve model performance.

Findings

No significant domain shift effect on model efficacy.

2D and 3D models achieve similar performance with a 0.93 score.

Fine-tuning and neural style transfer improve SSA case segmentation.

Abstract

In Sub-Saharan Africa (SSA), the utilization of lower-quality Magnetic Resonance Imaging (MRI) technology raises questions about the applicability of machine learning methods for clinical tasks. This study aims to provide a robust deep learning-based brain tumor segmentation (BraTS) method tailored for the SSA population using a threefold approach. Firstly, the impact of domain shift from the SSA training data on model efficacy was examined, revealing no significant effect. Secondly, a comparative analysis of 3D and 2D full-resolution models using the nnU-Net framework indicates similar performance of both the models trained for 300 epochs achieving a five-fold cross-validation score of 0.93. Lastly, addressing the performance gap observed in SSA validation as opposed to the relatively larger BraTS glioma (GLI) validation set, two strategies are proposed: fine-tuning SSA cases using the GLI+SSA best-pretrained 2D fullres model at 300 epochs, and introducing a novel neural style transfer-based data augmentation technique for the SSA cases. This investigation underscores the potential of enhancing brain tumor prediction within SSA's unique healthcare landscape.