Visual Interpretable and Explainable Deep Learning Models for Brain Tumor MRI and COVID-19 Chest X-ray Images

TL;DR

This paper evaluates attribution methods to improve interpretability of deep learning models in medical imaging, highlighting biomarkers and biases in brain tumor MRI and COVID-19 chest X-ray analysis to foster trust among healthcare professionals.

Contribution

It introduces an adaptive path-based gradient integration technique for explaining deep neural network decisions in medical image analysis.

Findings

The method effectively highlights relevant biomarkers.

It exposes potential model biases.

It enhances transparency and trust in AI predictions.

Abstract

Deep learning shows promise for medical image analysis but lacks interpretability, hindering adoption in healthcare. Attribution techniques that explain model reasoning may increase trust in deep learning among clinical stakeholders. This paper aimed to evaluate attribution methods for illuminating how deep neural networks analyze medical images. Using adaptive path-based gradient integration, we attributed predictions from brain tumor MRI and COVID-19 chest X-ray datasets made by recent deep convolutional neural network models. The technique highlighted possible biomarkers, exposed model biases, and offered insights into the links between input and prediction. Our analysis demonstrates the method's ability to elucidate model reasoning on these datasets. The resulting attributions show promise for improving deep learning transparency for domain experts by revealing the rationale behind predictions. This study advances model interpretability to increase trust in deep learning among healthcare stakeholders.