U-KAN Makes Strong Backbone for Medical Image Segmentation and Generation

TL;DR

This paper introduces U-KAN, a novel backbone for medical image segmentation and generation that integrates Kolmogorov-Arnold Networks (KANs) into U-Net, achieving higher accuracy and efficiency in benchmarks.

Contribution

The paper presents U-KAN, a redesigned U-Net architecture incorporating KAN layers, enhancing interpretability and performance in medical image tasks and diffusion models.

Findings

U-KAN outperforms traditional U-Net in accuracy on medical segmentation benchmarks.

U-KAN achieves comparable or better results with less computational cost.

U-KAN demonstrates effectiveness as a noise predictor in diffusion models.

Abstract

U-Net has become a cornerstone in various visual applications such as image segmentation and diffusion probability models. While numerous innovative designs and improvements have been introduced by incorporating transformers or MLPs, the networks are still limited to linearly modeling patterns as well as the deficient interpretability. To address these challenges, our intuition is inspired by the impressive results of the Kolmogorov-Arnold Networks (KANs) in terms of accuracy and interpretability, which reshape the neural network learning via the stack of non-linear learnable activation functions derived from the Kolmogorov-Anold representation theorem. Specifically, in this paper, we explore the untapped potential of KANs in improving backbones for vision tasks. We investigate, modify and re-design the established U-Net pipeline by integrating the dedicated KAN layers on the tokenized intermediate representation, termed U-KAN. Rigorous medical image segmentation benchmarks verify the superiority of U-KAN by higher accuracy even with less computation cost. We further delved into the potential of U-KAN as an alternative U-Net noise predictor in diffusion models, demonstrating its applicability in generating task-oriented model architectures. These endeavours unveil valuable insights and sheds light on the prospect that with U-KAN, you can make strong backbone for medical image segmentation and generation. Project page:\url{https://yes-u-kan.github.io/}.