American Journal for Young Scientists ยท ISSN 3068-9015
Automated Detection of Pneumonia from Chest X-Ray Images Using Machine Learning
Authors: Ranganathan Srinivasan
Published: October 8, 2025
Category: Biology
Abstract
Pneumonia remains a major global health concern, and early diagnosis from chest X-ray images is critical for effective treatment. Here we evaluate machine learning methods for automated pneumonia detection using chest radiographs. Three classifiers: VGG16, Support Vector Machine (SVM), and XGBoost were implemented and compared in terms of accu- racy, precision, recall, and F1-score. VGG16 achieved the highest overall accuracy (78.3%) with balanced performance across classes, while SVM demonstrated very high sensitivity for pneumonia (recall = 0.99) but lower specificity for normal cases. XGBoost provided inter- mediate performance, favoring pneumonia detection but misclassifying some normal cases. To further interpret the learned feature space, Uniform Manifold Approximation and Pro- jection (UMAP) was applied, revealing distinct but overlapping clusters of pneumonia and control samples. These results highlight both the potential and limitations of machine learn- ing for pneumonia detection, emphasizing that while models can assist in screening, challenges in specificity and overlapping feature representations remain. Future work should focus on larger datasets, integration of clinical information, and improved interpretability to enhance real-world applicability.