![]() ![]() However, the use of data augmentation to improve the robustness and generalizability of the model has been less explored and is the primary motivation for this paper. Algorithms such as SMOTE ( 4) are often used to augment the minority class by intelligently synthesizing new data without overfitting. It should be noted that data augmentation is commonly used in binary classification in cases where a large imbalance exists between the size of the two classes being used in a machine learning model. One of the techniques used during the training and testing phases is data augmentation, which is used to make the deep learning model more robust to different types of noise, as well as increase the training dataset, which is typically needed in clinical applications. Deep learning provides the ability to learn and nonlinearly associate high-dimensional features in X-ray images that feature COVID-19 ( 3). It is therefore expected that this area of research will attract more researchers and that more papers will be published on this topic. Recent research ( 2) has shown that the combination of deep learning and chest X-rays could be faster and less expensive than the gold standard for COVID-19 diagnosis, which is a laboratory technique called reverse transcriptase polymerase chain reaction (PCR). More people are being infected with COVID-19 every day ( 1) therefore there is a need for a quick and reliable technology to help with the screening and management of the virus. We also provide clinical perspectives on geometric augmentation to consider regarding the development of a robust COVID-19 X-ray-based detector. This is an interesting finding that may improve current deep learning algorithms using geometrical augmentations for detecting COVID-19. When we retrained a recently published deep learning without augmentation on the same dataset, the detection accuracy significantly increased, with a χ McNema r ′ s statistic 2 = 163. The MCC without augmentation (MCC = 0.51) outperformed four recent geometrical augmentations (MCC = 0.47 for Data Augmentation 1, MCC = 0.44 for Data Augmentation 2, MCC = 0.48 for Data Augmentation 3, and MCC = 0.49 for Data Augmentation 4). Contrary to expectation, our results show that the removal of recently used geometrical augmentation steps actually improved the Matthews correlation coefficient (MCC) of 17 models. We empirically examined the influence of augmentation with respect to detection accuracy, dataset diversity, augmentation methodology, and network size. We compared the performance of 17 deep learning algorithms with and without different geometric augmentations. In this paper, we examined the impact of geometric augmentations as implemented in several recent publications for detecting COVID-19. In the image processing field, an image augmentation step (i.e., adding more training data) is often used to reduce overfitting on the training dataset, and improve prediction accuracy on the testing dataset. While overfitting can result in perfect modeling on the original training dataset, on a new testing dataset it can fail to achieve high accuracy. Deep learning relies on a large amount of data to avoid overfitting. Additionally, deep learning methods have recently shown remarkable results in detecting COVID-19 on chest X-rays, making it a promising screening technology for COVID-19. 10Department of Radiology, Faculty of Medicine, University of British Columbia, Vancouver, BC, CanadaĬhest X-ray imaging technology used for the early detection and screening of COVID-19 pneumonia is both accessible worldwide and affordable compared to other non-invasive technologies.9D-Lab, Massachusetts Institute of Technology, Cambridge, MA, United States.8Biomedical and Mobile Health Technology Laboratory, Department of Health Sciences and Technology, ETH Zurich, Zurich, Switzerland.7Department of Physics & Astronomy, Louisiana State University, Baton Rouge, LA, United States.6Department of Radiology, Louisiana State University Health Sciences Center, New Orleans, LA, United States.5Department of Emergency and Trauma Radiology, Vancouver General Hospital, Vancouver, BC, Canada.4School of Electrical and Computer Engineering, University of British Columbia, Vancouver, BC, Canada.3Nuffield Department of Surgical Sciences, University of Oxford, Oxford, United Kingdom.2School of Mechatronic Systems Engineering, Simon Fraser University, Burnaby, BC, Canada.1Rady Faculty of Health Sciences, University of Manitoba, Winnipeg, MB, Canada.Mohamed Elgendi 1,2,3,4 *, Muhammad Umer Nasir 5, Qunfeng Tang 4, David Smith 6, John-Paul Grenier 6, Catherine Batte 7, Bradley Spieler 6, William Donald Leslie 1, Carlo Menon 2,8, Richard Ribbon Fletcher 9, Newton Howard 3, Rabab Ward 4, William Parker 5 and Savvas Nicolaou 5,10 ![]()
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |