Implicit U-Net for volumetric medical image segmentation

TL;DR

This paper introduces the Implicit U-Net, a more parameter-efficient and faster-to-train variant of the traditional U-Net for 3D medical image segmentation, maintaining comparable accuracy.

Contribution

The Implicit U-Net combines implicit representations with convolutional features, reducing parameters and computational costs while preserving segmentation performance.

Findings

40% fewer parameters than standard U-Net

30% reduction in training and inference time

Comparable segmentation accuracy on CT datasets

Abstract

U-Net has been the go-to architecture for medical image segmentation tasks, however computational challenges arise when extending the U-Net architecture to 3D images. We propose the Implicit U-Net architecture that adapts the efficient Implicit Representation paradigm to supervised image segmentation tasks. By combining a convolutional feature extractor with an implicit localization network, our implicit U-Net has 40% less parameters than the equivalent U-Net. Moreover, we propose training and inference procedures to capitalize sparse predictions. When comparing to an equivalent fully convolutional U-Net, Implicit U-Net reduces by approximately 30% inference and training time as well as training memory footprint while achieving comparable results in our experiments with two different abdominal CT scan datasets.