MEDUSA: Multi-scale Encoder-Decoder Self-Attention Deep Neural Network Architecture for Medical Image Analysis

TL;DR

MEDUSA introduces a novel multi-scale encoder-decoder self-attention architecture for medical image analysis, achieving state-of-the-art results by unifying global and local attention mechanisms in a single model.

Contribution

This work presents the first unified, multi-scale self-attention mechanism with multiple heads feeding into different network levels for medical imaging tasks.

Findings

Achieved state-of-the-art performance on COVIDx, RSNA RICORD, and RSNA Pneumonia datasets.

Introduced a single, high-capacity self-attention mechanism with multiple heads at different scales.

Demonstrated the effectiveness of global and local attention integration in medical image analysis.

Abstract

Medical image analysis continues to hold interesting challenges given the subtle characteristics of certain diseases and the significant overlap in appearance between diseases. In this work, we explore the concept of self-attention for tackling such subtleties in and between diseases. To this end, we introduce MEDUSA, a multi-scale encoder-decoder self-attention mechanism tailored for medical image analysis. While self-attention deep convolutional neural network architectures in existing literature center around the notion of multiple isolated lightweight attention mechanisms with limited individual capacities being incorporated at different points in the network architecture, MEDUSA takes a significant departure from this notion by possessing a single, unified self-attention mechanism with significantly higher capacity with multiple attention heads feeding into different scales in the network architecture. To the best of the authors' knowledge, this is the first "single body, multi-scale heads" realization of self-attention and enables explicit global context amongst selective attention at different levels of representational abstractions while still enabling differing local attention context at individual levels of abstractions. With MEDUSA, we obtain state-of-the-art performance on multiple challenging medical image analysis benchmarks including COVIDx, RSNA RICORD, and RSNA Pneumonia Challenge when compared to previous work. Our MEDUSA model is publicly available.