Automated Pneumonia Detection in Chest X-ray Images Using Neural Networks
Click the Poster to View Full Screen, Right click to save image
Grant: LSAMP
Keith Fernandez
College:
The Dorothy and George Hennings College of Science, Mathematics, and Technology
Major:
Computer Science
Faculty Research Advisor(s):
Daehan Kwak
Abstract:
This research aimed to develop an automated pneumonia detection system using machine learning on chest X-ray images to improve diagnostic efficiency and patient outcomes. A dataset was collected and curated, deep learning architectures were explored, and model performance was optimized, resulting in an 80.13% accuracy rate. This work has implications for medical imaging and healthcare by showcasing the potential of machine learning in clinical decision-making.
Pneumonia is a major global health concern, necessitating early and accurate diagnosis for effective patient management. Manual interpretation of chest X-rays is labor-intensive and prone to variability. Machine learning offers a solution by automating pneumonia detection, reducing errors, and expediting care decisions. This research focuses on developing an automated detection system using convolutional neural networks (CNNs) trained on a comprehensive dataset of chest X-ray images.
Data preparation involved priming and normalization, ensuring consistency across training, testing, and validation phases. Various transformations were applied to enhance model performance. The study explored deep learning architectures, leveraging CNNs to learn features from input images. Model optimization involved meticulous steps such as data preprocessing, augmentation, and hyperparameter tuning. Google Colab's resources were utilized for efficient model training and evaluation.
The designed CNN architecture comprised convolutional, pooling, and fully connected layers for classification. The model achieved an 80.13% accuracy rate, demonstrating its efficacy and potential for further improvement. Adjustments to critical hyperparameters were made to refine performance over 20 epochs.
This research addresses the need for improved diagnostic accuracy in respiratory infections. It draws from a comprehensive literature review on machine learning in healthcare and employs a methodological approach involving data collection, preprocessing, and CNN development. The study's findings highlight the model's proficiency in pneumonia detection, with implications for computer-aided diagnostic systems and patient care.
This research developed an automated pneumonia detection system using CNNs, showcasing its potential to enhance medical imaging diagnostics. The outcomes hold promise for improving patient care and healthcare efficiency. This work underscores the transformative role of machine learning in healthcare, paving the way for future advancements in clinical decision-making.