Performance Comparison for Different Neural Network Architectures for chest X-Ray Image Classification

Nowadays, the diagnose of diseases or abnormally clinical conditions of the chest can be done by reading chest X-Ray images by radiologists. The diagnosis from radiologist is considered qualitative analysis, which collects and analyzes the non-numerical data, which relates to the individual's e...

Full description

Saved in:

Bibliographic Details
Published in:	2021 7th International Conference on Engineering, Applied Sciences and Technology (ICEAST) pp. 49 - 53
Main Authors:	Rajadanuraks, Pinyada, Suranuntchai, Sarapom, Pechprasarn, Suejit, Treebupachatsakul, Treesukon
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01-04-2021
Subjects:	artificial intelligence chest X-Ray images Computer architecture convolutional neural network Deep learning deep learning neural network Mathematical model medical image processing Network architecture Neural networks Training X-ray imaging
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Nowadays, the diagnose of diseases or abnormally clinical conditions of the chest can be done by reading chest X-Ray images by radiologists. The diagnosis from radiologist is considered qualitative analysis, which collects and analyzes the non-numerical data, which relates to the individual's experience. Consequently, errors may occur during the diagnosis, such as misdiagnosis. Therefore, this research has a purpose for developing the software program which gathers Medical Image Processing and Artificial Intelligence focuses on Deep Learning Neural Network, implementing on MATLAB program by training and classifying on labeled and non-labeled chest X-Ray images between normal condition and 13 abnormal cases. It is also stated that the trained model results are considered quantitative analysis, which can be confirmed or rejected by testing the causal relationship between variables. In this research, we apply several pre-trained architectures to our network, which, finally, the ResNet-50 architecture gave the highest percentage of accuracy compared with other architectures. Moreover, it is the network that appropriates for our image dataset.
DOI:	10.1109/ICEAST52143.2021.9426289