Region Attention Transformer for Medical Image Restoration

TL;DR

This work introduces Region Attention Transformer (RAT) for medical image restoration by partitioning inputs into semantic regions with SAM and performing region-restricted attention (R-MSA), thereby reducing interference from irrelevant content and preserving continuous structures. A cross-region coupling via windowed attention complements region-specific processing, while a focal region loss concentrates learning on the most challenging regions. The approach achieves state-of-the-art results across PET image synthesis, CT denoising, and pathological image super-resolution, validated by extensive experiments and ablations. The combination of SAM-guided region partitioning, region-confined attention, and adaptive loss demonstrates a practical, interpretable, and effective framework for MedIR tasks with potential for broad clinical impact.

Abstract

Transformer-based methods have demonstrated impressive results in medical image restoration, attributed to the multi-head self-attention (MSA) mechanism in the spatial dimension. However, the majority of existing Transformers conduct attention within fixed and coarsely partitioned regions (\text{e.g.} the entire image or fixed patches), resulting in interference from irrelevant regions and fragmentation of continuous image content. To overcome these challenges, we introduce a novel Region Attention Transformer (RAT) that utilizes a region-based multi-head self-attention mechanism (R-MSA). The R-MSA dynamically partitions the input image into non-overlapping semantic regions using the robust Segment Anything Model (SAM) and then performs self-attention within these regions. This region partitioning is more flexible and interpretable, ensuring that only pixels from similar semantic regions complement each other, thereby eliminating interference from irrelevant regions. Moreover, we introduce a focal region loss to guide our model to adaptively focus on recovering high-difficulty regions. Extensive experiments demonstrate the effectiveness of RAT in various medical image restoration tasks, including PET image synthesis, CT image denoising, and pathological image super-resolution. Code is available at \href{https://github.com/Yaziwel/Region-Attention-Transformer-for-Medical-Image-Restoration.git}{https://github.com/RAT}.

Figures (4)

• Figure 1: Overall architecture of the proposed region attention transformer (RAT).
• Figure 2: Visual comparison among methods on PET image synthesis.
• Figure 3: Visual comparison among methods on CT image denoising.
• Figure 4: Visual comparison among methods on pathological image super-resolution.