RadFormer: Transformers with Global-Local Attention for Interpretable and Accurate Gallbladder Cancer Detection

TL;DR

RadFormer is a transformer-based neural network that combines global and local attention mechanisms to achieve highly accurate and interpretable detection of Gallbladder Cancer from ultrasound images, outperforming radiologists.

Contribution

The paper introduces a novel architecture integrating global and local attention with bag of words embeddings for interpretable and accurate GBC detection from ultrasound images.

Findings

Model outperforms radiologists in GBC detection accuracy.

Provides interpretable explanations aligned with medical literature.

Facilitates discovery of new visual features for GBC diagnosis.

Abstract

We propose a novel deep neural network architecture to learn interpretable representation for medical image analysis. Our architecture generates a global attention for region of interest, and then learns bag of words style deep feature embeddings with local attention. The global, and local feature maps are combined using a contemporary transformer architecture for highly accurate Gallbladder Cancer (GBC) detection from Ultrasound (USG) images. Our experiments indicate that the detection accuracy of our model beats even human radiologists, and advocates its use as the second reader for GBC diagnosis. Bag of words embeddings allow our model to be probed for generating interpretable explanations for GBC detection consistent with the ones reported in medical literature. We show that the proposed model not only helps understand decisions of neural network models but also aids in discovery of new visual features relevant to the diagnosis of GBC. Source-code and model will be available at https://github.com/sbasu276/RadFormer