Is Long Range Sequential Modeling Necessary For Colorectal Tumor Segmentation?

TL;DR

This paper investigates whether long-range sequential modeling is necessary for 3D colorectal tumor segmentation. It introduces CTS-204 and compares global token modeling (Mamba/Transformers) with a local-token-focused architecture, MambaOutUNet, across CTS-204 and BTCV datasets. The results show that for small, anatomically complex ROIs like colorectal tumors, robust local interactions can outperform long-range methods, though long-range techniques may still help in datasets with highly variable ROIs. The work advocates focusing on efficient local token modeling for segmentation of small tumors and positions MambaOutUNet as a valuable baseline for future 3D segmentation research, with CTS-204 enabling broader evaluation of these approaches.

Abstract

Segmentation of colorectal cancer (CRC) tumors in 3D medical imaging is both complex and clinically critical, providing vital support for effective radiation therapy planning and survival outcome assessment. Recently, 3D volumetric segmentation architectures incorporating long-range sequence modeling mechanisms, such as Transformers and Mamba, have gained attention for their capacity to achieve high accuracy in 3D medical image segmentation. In this work, we evaluate the effectiveness of these global token modeling techniques by pitting them against our proposed MambaOutUNet within the context of our newly introduced colorectal tumor segmentation dataset (CTS-204). Our findings suggest that robust local token interactions can outperform long-range modeling techniques in cases where the region of interest is small and anatomically complex, proposing a potential shift in 3D tumor segmentation research.

Figures (1)

• Figure 1: Qualitative comparison of MambaOutUNet with other established and proposed baselines.