Spectral U-Net: Enhancing Medical Image Segmentation via Spectral Decomposition

TL;DR

Spectral U-Net introduces spectral decomposition using DTCWT and iDTCWT within a U-Net architecture to improve medical image segmentation by reducing information loss and enhancing detail reconstruction, showing superior results on multiple datasets.

Contribution

The paper presents a novel spectral decomposition-based U-Net architecture with Wave-Blocks, improving segmentation accuracy by better preserving information during down- and up-sampling.

Findings

Outperforms standard U-Net on Retina Fluid, Brain Tumor, and Liver Tumor datasets.

Effectively mitigates information loss during down-sampling.

Enhances detail reconstruction during up-sampling.

Abstract

This paper introduces Spectral U-Net, a novel deep learning network based on spectral decomposition, by exploiting Dual Tree Complex Wavelet Transform (DTCWT) for down-sampling and inverse Dual Tree Complex Wavelet Transform (iDTCWT) for up-sampling. We devise the corresponding Wave-Block and iWave-Block, integrated into the U-Net architecture, aiming at mitigating information loss during down-sampling and enhancing detail reconstruction during up-sampling. In the encoder, we first decompose the feature map into high and low-frequency components using DTCWT, enabling down-sampling while mitigating information loss. In the decoder, we utilize iDTCWT to reconstruct higher-resolution feature maps from down-sampled features. Evaluations on the Retina Fluid, Brain Tumor, and Liver Tumor segmentation datasets with the nnU-Net framework demonstrate the superiority of the proposed Spectral U-Net.