Impact of data augmentation techniques on a deep learning based medical imaging task

For any deep learning (DL) based task, model generalization and prediction performance improve as a function of training data set size and variety. However, its application to medical imaging is still challenging because of the limited availability of high-quality and sufficiently diverse annotated data. Data augmentation techniques can improve the model performance when the available dataset size is limited. Anatomy region localization from medical images can be automated with deep learning and is important for tasks such as organ segmentation and lesion detection. Different data augmentation methods were compared for DL based anatomy region localization with computed tomography images. The impact of different neural network architectures was also explored. The prediction accuracy on an independent test set improved from 88% to 97% with optimal selection of data augmentation and architecture while using the same training dataset Data augmentation steps such as zoom, translation and flips had incremental effect on classifier performance whereas sample-wise mean shift appeared to degrade the classifier performance. Global average pooling improved classifier accuracy compared to fully-connected layer when limited data augmentation was used. All model architectures converged to an optimal performance with the right combination of augmentation steps. Prediction inaccuracies were mostly observed in the boundary regions between anatomies. The networks also successfully localized anatomy for Positron Emission Tomography studies reaching an accuracy of up to 97%. Similar impact of data augmentation and pooling layer was also observed.