AAI_2025_Capstone_Chronicles_Combined

Abstract

Data collection and storage in healthcare present significant challenges due to high costs

and resource limitations. Many hospitals struggle to gather and maintain large datasets

necessary for training machine learning models that could otherwise enhance diagnostic

accuracy. Additionally, healthcare providers often face overwhelming workloads, complex cases,

and limited resources, increasing the potential for diagnostic errors. This project addresses

these issues by employing Generative Adversarial Networks (GANs) to generate synthetic X-ray

images of patients with pneumonia. The synthetic images aim to enrich the dataset for training

a Convolutional Neural Network (CNN) classifier to detect pneumonia from X-ray samples. The

primary objective is to increase data availability through synthetic generation of realistic images.

By supplementing the dataset with realistic synthetic images, diagnostic models such as CNN

classifiers are expected to achieve higher accuracy, ultimately enhancing diagnostic precision.

The dataset utilized is the Labeled Optical Coherence Tomography (OCT) and Chest X-ray

Images for Classification from Kermany et al. (2018), publicly available under a Creative

Commons Attribution 4.0 International (CC BY 4.0) license. It contains 5,863 images labeled as

either pneumonia or normal, originating from retrospective cohorts of pediatric patients aged

one to five years at Guangzhou Women and Children's Medical Center, where chest X-rays were

collected as part of routine clinical care (Kermany, Zhang, & Goldbaum, 2018). In a live system,

data would be sourced from hospital imaging databases, with synthetic augmentation

continuously refining the dataset during training.

95