Young, Julio Cristian and Suryadibrata, Alethea (2020) Applicability of various pre-trained deep convolutional neural networks for pneumonia classification based on X-Ray Images. International Journal of Advanced Trends in Computer Science and Engineering, 9 (3). ISSN 2278-3091
Full text not available from this repository.Abstract
Along with the development of machine learning methodologies, in the last few decades, a variety of machine learning techniques began to solve many problems in various fields. Encouraged by the development of computer hardware technology, deep learning has become one of machine learning methods that have become incredibly attractive due to the performance of such methods. By using deep learning, the researchers managed to find solutions to a variety of complex problems that had not yet had an optimal solution. In the field of bioinformatics and computer-aided diagnosis, previous researches show that deep learning can be utilized to predict the structure of protein compounds, detecting lumbar spinal stenosis regions, or other biomedical related prediction problems. In this study, we are experimenting with various pre-trained deep convolutional neural networks (CNN) architectures, namely, densely connected convolutional networks (DenseNet), VGG-16, Inception, and Inception-ResNet, to build another classifier on top of before-mentioned pre-trained models for classify each patient condition based on their chest X-Ray images. In this study, our resulted model needs to able to predict whether a patient’s lungs have a normal condition or have contracted either with viral or bacterial pneumonia-related conditions based on the given chest X-Ray image. Our experiment shows that by using a variation of DenseNet, DenseNet201, deep CNN is able to achieve the accuracy scores of 93.87% for the training data and 82.7% for the validation data.
| Item Type: | Article |
|---|---|
| Subjects: | 000 Computer Science, Information and General Works > 000 Computer Science, Knowledge and Systems > 006 Special Computer Methods |
| Divisions: | Faculty of Engineering & Informatics > Informatics |
| Depositing User: | Administrator UMN Library |
| Date Deposited: | 13 Oct 2021 02:36 |
| Last Modified: | 13 Oct 2021 02:36 |
| URI: | https://kc.umn.ac.id/id/eprint/18709 |
Actions (login required)
 |
View Item |
