Home / Current Issue / Paper 1705002
Enhancing Pneumonia Detection in Chest X-rays through Transfer Learning with AlexNet and Adversarial Training
Subject area: Science,Engineering and Technology · Area of research: AlexNet and Adversarial Training
Abstract
Pneumonia poses serious health risks if not promptly diagnosed and treated. Despite the value of chest X-rays for pneumonia detection, their interpretation is complicated by intricate image details. This study proposes a dual approach for enhancing pneumonia detection in chest X-rays: implementing Transfer Learning with the AlexNet architecture and employing Adversarial Training. Transfer Learning exploits a pre-trained deep neural network, AlexNet, to boost performance with limited data. AlexNet, initialized with ImageNet weights, captures relevant features from X-ray images. The model is fine-tuned using labeled pneumonia X-ray data. Adversarial Training supplements the model's discriminative capabilities by integrating an adversarial network. By generating challenging adversarial examples, the network learns more discerning features, thus improving pneumonia detection. Experiments on a benchmark X-ray dataset and comparison with conventional methods showcase notable accuracy and precision gains with the Transfer Learning-AlexNet approach. Incorporating Adversarial Training enhances performance further, increasing sensitivity and specificity. In conclusion, coupling Transfer Learning, AlexNet, and Adversarial Training presents a promising route to significantly improve pneumonia detection accuracy in chest X-rays. This research contributes to the ongoing enhancement of medical image analysis, fostering improved clinical decision support systems.
Keywords
Pneumonia Detection, Chest X-rays, Transfer Learning, AlexNet, Adversarial Training, Medical Image Analysis
How to cite this paper
@article{1705002,
author = {Malik Jawarneh, Haydar Sabeeh Kalash, Maria AL Amri},
title = {Enhancing Pneumonia Detection in Chest X-rays through Transfer Learning with AlexNet and Adversarial Training},
journal = {Iconic Research And Engineering Journals},
year = {2023},
volume = {7},
number = {2},
pages = {318-325},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1705002.pdf},
abstract = {Pneumonia poses serious health risks if not promptly diagnosed and treated. Despite the value of chest X-rays for pneumonia detection, their interpretation is complicated by intricate image details. This study proposes a dual approach for enhancing pneumonia detection in chest X-rays: implementing Transfer Learning with the AlexNet architecture and employing Adversarial Training. Transfer Learning exploits a pre-trained deep neural network, AlexNet, to boost performance with limited data. AlexNet, initialized with ImageNet weights, captures relevant features from X-ray images. The model is fine-tuned using labeled pneumonia X-ray data. Adversarial Training supplements the model's discriminative capabilities by integrating an adversarial network. By generating challenging adversarial examples, the network learns more discerning features, thus improving pneumonia detection. Experiments on a benchmark X-ray dataset and comparison with conventional methods showcase notable accuracy and precision gains with the Transfer Learning-AlexNet approach. Incorporating Adversarial Training enhances performance further, increasing sensitivity and specificity. In conclusion, coupling Transfer Learning, AlexNet, and Adversarial Training presents a promising route to significantly improve pneumonia detection accuracy in chest X-rays. This research contributes to the ongoing enhancement of medical image analysis, fostering improved clinical decision support systems.},
keywords = {Pneumonia Detection, Chest X-rays, Transfer Learning, AlexNet, Adversarial Training, Medical Image Analysis},
month = {August},
}