A Critical Appraisal of Data Augmentation Methods for Imaging-Based Medical Diagnosis Applications

TL;DR

This paper critically evaluates current data augmentation techniques for medical imaging, highlighting their limitations and potential negative impacts on diagnostic accuracy due to improper application and lack of optimization.

Contribution

It provides an analysis of how existing augmentation methods can distort medical images and emphasizes the need for tailored approaches in medical imaging applications.

Findings

Intensity-based augmentations distort MRI scans

Common augmentations require manual tuning

Augmentation can negatively impact classification performance

Abstract

Current data augmentation techniques and transformations are well suited for improving the size and quality of natural image datasets but are not yet optimized for medical imaging. We hypothesize that sub-optimal data augmentations can easily distort or occlude medical images, leading to false positives or negatives during patient diagnosis, prediction, or therapy/surgery evaluation. In our experimental results, we found that utilizing commonly used intensity-based data augmentation distorts the MRI scans and leads to texture information loss, thus negatively affecting the overall performance of classification. Additionally, we observed that commonly used data augmentation methods cannot be used with a plug-and-play approach in medical imaging, and requires manual tuning and adjustment.