EM-Net: Efficient Channel and Frequency Learning with Mamba for 3D Medical Image Segmentation

TL;DR

EM-Net introduces a novel Mamba-based approach for 3D medical image segmentation that improves accuracy, reduces model size, and accelerates training by leveraging channel and frequency domain learning.

Contribution

The paper presents a new Mamba-based model, EM-Net, that efficiently captures regional interactions and multi-scale features in 3D medical images, outperforming existing methods.

Findings

Achieves better segmentation accuracy than SOTA models.

Uses nearly half the parameters of comparable models.

Provides 2x faster training speed.

Abstract

Convolutional neural networks have primarily led 3D medical image segmentation but may be limited by small receptive fields. Transformer models excel in capturing global relationships through self-attention but are challenged by high computational costs at high resolutions. Recently, Mamba, a state space model, has emerged as an effective approach for sequential modeling. Inspired by its success, we introduce a novel Mamba-based 3D medical image segmentation model called EM-Net. It not only efficiently captures attentive interaction between regions by integrating and selecting channels, but also effectively utilizes frequency domain to harmonize the learning of features across varying scales, while accelerating training speed. Comprehensive experiments on two challenging multi-organ datasets with other state-of-the-art (SOTA) algorithms show that our method exhibits better segmentation accuracy while requiring nearly half the parameter size of SOTA models and 2x faster training speed.