Case Study: Explaining Diabetic Retinopathy Detection Deep CNNs via Integrated Gradients

TL;DR

This paper applies integrated gradients to explain a deep CNN for diabetic retinopathy detection, enhancing interpretability by analyzing intermediate layers, filtering units, and generating counterexamples, thus aiding lesion identification.

Contribution

It introduces novel applications of integrated gradients for explaining intermediate layers and generating contrary samples in diabetic retinopathy detection models.

Findings

Enhanced interpretability of CNN via integrated gradients

Ability to identify potential lesions through visualization

New methods for filtering unimportant units and generating counterexamples

Abstract

In this report, we applied integrated gradients to explaining a neural network for diabetic retinopathy detection. The integrated gradient is an attribution method which measures the contributions of input to the quantity of interest. We explored some new ways for applying this method such as explaining intermediate layers, filtering out unimportant units by their attribution value and generating contrary samples. Moreover, the visualization results extend the use of diabetic retinopathy detection model from merely predicting to assisting finding potential lesions.