You are here
Assessment of Data Augmentation Strategies Toward Performance Improvement of Abnormality Classification in Chest Radiographs.
Image augmentation is a commonly performed technique to prevent class imbalance in datasets to compensate for insufﬁcient training samples, or to prevent model overﬁtting. Traditional augmentation (TA) techniques include various image transformations, such as rotation, translation, channel splitting, etc. Alternatively, Generative Adversarial Network (GAN), due to its proven ability to synthesize convincinglyrealistic images, has been used to perform image augmentation as well. However, it is unclear whether GAN augmentation (GA) strategy provides an advantage over TA for medical image classiﬁcation tasks. In this paper, we study the usefulness of TA and GA for classifying abnormal chest X-ray (CXR) images. We ﬁrst trained a progressive-growing GAN (PGGAN) to synthesize high-resolution CXRs for performing GA. Then, we trained an abnormality classiﬁer using three training sets individually – training set with TA, with GA and with no augmentation (NA). Finally, we analyzed the abnormality classiﬁer’s performance for the three training cases, which led to the following conclusions: (1) GAN strategy is not always superior to TA for improving the classiﬁer’s performance; (2) in comparison to NA, however, both TA and GA leads to a signiﬁcant performance improvement; and, (3) increasing the quantity of images in TA and GA strategies also improves the classiﬁer’s performance.