Microscopic fine-grained instance classification through deep attention

TL;DR

This paper introduces a lightweight deep learning model with attention mechanisms for fine-grained microscopic image classification, achieving state-of-the-art results and interpretability in biomedical datasets.

Contribution

It presents a novel end-to-end deep network that combines gated attention and feature fusion for high-resolution, fine-grained biomedical image classification without extra annotations.

Findings

State-of-the-art accuracy on breast cancer histology dataset

Effective classification on fungi mycology dataset

Model interpretability aligned with clinical features

Abstract

Fine-grained classification of microscopic image data with limited samples is an open problem in computer vision and biomedical imaging. Deep learning based vision systems mostly deal with high number of low-resolution images, whereas subtle detail in biomedical images require higher resolution. To bridge this gap, we propose a simple yet effective deep network that performs two tasks simultaneously in an end-to-end manner. First, it utilises a gated attention module that can focus on multiple key instances at high resolution without extra annotations or region proposals. Second, the global structural features and local instance features are fused for final image level classification. The result is a robust but lightweight end-to-end trainable deep network that yields state-of-the-art results in two separate fine-grained multi-instance biomedical image classification tasks: a benchmark breast cancer histology dataset and our new fungi species mycology dataset. In addition, we demonstrate the interpretability of the proposed model by visualising the concordance of the learned features with clinically relevant features.