3D Deformable Convolutions for MRI classification

TL;DR

This paper introduces 3D deformable convolutional layers integrated into VoxResNet for MRI classification, demonstrating improved performance and robustness over standard convolutions in analyzing unprocessed 3D MRI data.

Contribution

The paper proposes a novel 3D deformable convolutional layer and integrates it into VoxResNet, showing enhanced MRI classification capabilities.

Findings

3D deformable convolutions outperform standard convolutions.

The proposed dVoxResNet is effective for unprocessed 3D MRI data.

The method is robust to geometrical variations in MRI images.

Abstract

Deep learning convolutional neural networks have proved to be a powerful tool for MRI analysis. In current work, we explore the potential of the deformable convolutional deep neural network layers for MRI data classification. We propose new 3D deformable convolutions(d-convolutions), implement them in VoxResNet architecture and apply for structural MRI data classification. We show that 3D d-convolutions outperform standard ones and are effective for unprocessed 3D MR images being robust to particular geometrical properties of the data. Firstly proposed dVoxResNet architecture exhibits high potential for the use in MRI data classification.