Automated Ensemble Deep Learning for Chest X-Ray Covid-19 Image Classification Using Multiple Hyperparameter Optimizations
Conference proceedings article
Authors/Editors
Strategic Research Themes
Publication Details
Author list: Mudaser, Muhibullah; Krathu, Worarat; Jaiyen, Saichon;
Publisher: Institute of Electrical and Electronics Engineers Inc.
Publication year: 2023
Start page: 348
End page: 353
Number of pages: 6
ISBN: 979-835034210-9
Languages: English-Great Britain (EN-GB)
Abstract
The COVID-19 pandemic has created an urgent need for reliable methods to detect the virus. Chest X-ray images have emerged as a valuable diagnostic tool due to their accessibility and potential for identifying COVID-19 patterns. A review of the existing literature in this area reveals a gap in the research related to the use of automated ensemble deep learning techniques for Covid-19 image classification, particularly in the context of hyperparameter optimization. Therefore, the current study seeks to fill this research gap by conducting experiments to determine the effectiveness of various ensemble deep-learning approaches and exploring the impact of multiple hyperparameter optimizations on the accuracy of the classification model. The methodology involves collecting a large dataset of COVID and normal chest X-ray images, preprocessing the data, and developing individual deep-learning models such as DenseNet201, DenseNet121, and MobileNetV2. Hyperparameter optimization techniques, including BayesSearchCV, GridSearchCV, and RandomSearchCV, are applied to fine-tune the models. The results demonstrate high accuracy, specificity, and sensitivity values for the selected models. Ensemble method used, such as stacking effectively combined the strengths of different models, achieved results an accuracy of 95.98%, sensitivity of 95.39%, and specificity of 96.57%. ฉ 2023 IEEE.
Keywords
BayesSearchCV, Chest X-ray Images, Deep Learning, Ensemble Model, Gridsearchcv, Randomsearchcv
Contact Information