eduzhai > Applied Sciences > Engineering >

A Comprehensive Study of Data Augmentation Strategies for Prostate Cancer Detection in Diffusion-weighted MRI using Convolutional Neural Networks

  • Save

... pages left unread,continue reading

Document pages: 22 pages

Abstract: Data augmentation refers to a group of techniques whose goal is to battlelimited amount of available data to improve model generalization and pushsample distribution toward the true distribution. While different augmentationstrategies and their combinations have been investigated for various computervision tasks in the context of deep learning, a specific work in the domain ofmedical imaging is rare and to the best of our knowledge, there has been nodedicated work on exploring the effects of various augmentation methods on theperformance of deep learning models in prostate cancer detection. In this work,we have statically applied five most frequently used augmentation techniques(random rotation, horizontal flip, vertical flip, random crop, and translation)to prostate Diffusion-weighted Magnetic Resonance Imaging training dataset of217 patients separately and evaluated the effect of each method on the accuracyof prostate cancer detection. The augmentation algorithms were appliedindependently to each data channel and a shallow as well as a deepConvolutional Neural Network (CNN) were trained on the five augmented setsseparately. We used Area Under Receiver Operating Characteristic (ROC) curve(AUC) to evaluate the performance of the trained CNNs on a separate test set of95 patients, using a validation set of 102 patients for finetuning. The shallownetwork outperformed the deep network with the best 2D slice-based AUC of 0.85obtained by the rotation method.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×