Pediatric pneumonia X-ray image classification: predictive model development with DenseNet-169 transfer learning
Journal of Medical Artificial Intelligence · 2025-12 · 8 · 37-37
DOI: 10.21037/jmai-24-356 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Medical Artificial Intelligence, 2025-12. Verfasst von Sasanka Katreddi und 4 weiteren Autor:innen. Die Arbeit wurde bisher 9-mal zitiert (Crossref).
Abstract (Originalsprache)
Background
Pneumonia is a leading cause of global mortality, particularly affecting infants, the elderly, and immunocompromised individuals. Timely and accurate diagnosis is essential for effective treatment, yet traditional methods often lack reliability and efficiency. Recent developments in medical imaging, especially through the application of deep learning (DL) and machine learning (ML) algorithms which are key components of artificial intelligence (AI), have demonstrated significant promise in improving the detection of pneumonia in chest X-ray images. This prospective study investigates the application of DL techniques, particularly convolutional neural networks (CNNs) to enhance the accuracy and efficiency of pediatric pneumonia diagnosis.
Methods
The eligibility criteria included pediatric patients aged 1–5 years, and the dataset was sourced from Kaggle. The dataset underwent preprocessing to enhance its suitability for transfer learning applications. The sample was split into training 85.88%, validation 4.2%, and test 9.92% sets. Various pre-trained CNN models were utilized, including VGG16, DenseNet-121, and DenseNet-169. Furthermore, ensemble methods that integrated these architectures were created to boost the overall performance of the models. The effectiveness of the models was assessed using standard evaluation metrics such as accuracy, recall, precision, and learning curves.
Results
Among the evaluated architectures, the DenseNet-169 model achieved the highest levels of accuracy and precision, recording 91.66% (89.37–93.95%) and 90.99% (88.62–93.36%), respectively with 95% confidence interval (CI). These results underscore the potential of DL methodologies, particularly those that leverage transfer learning and model ensembles, in advancing the detection of pneumonia from chest X-ray images. The statistical analysis was conducted using Python and other python libraries such as SciPy, Statsmodels.
Conclusions
The findings of our study indicate that DL, particularly using transfer learning and model ensembles, offers substantial potential for enhancing the accuracy of pneumonia detection. Notably, the DenseNet-169 model exhibited outstanding performance, which may facilitate improved early diagnosis and treatment of pneumonia, ultimately leading to better patient outcomes and reduced mortality rates.
Zitieren
APA
Katreddi, S., Midatani, A., Roy, A. P., Velpuri, U., & Kasani, S. (2025). Pediatric pneumonia X-ray image classification: predictive model development with DenseNet-169 transfer learning. Journal of Medical Artificial Intelligence, 8, 37-37. https://doi.org/10.21037/jmai-24-356
Vancouver
Katreddi S, Midatani A, Roy AP, Velpuri U, Kasani S. Pediatric pneumonia X-ray image classification: predictive model development with DenseNet-169 transfer learning. Journal of Medical Artificial Intelligence. 2025;8:37-37. doi:10.21037/jmai-24-356
BibTeX
@article{Katreddi2025,
author = {Katreddi, Sasanka and Midatani, Avinash and Roy, Akshara P. and Velpuri, Usha and Kasani, Sujan},
title = {Pediatric pneumonia X-ray image classification: predictive model development with DenseNet-169 transfer learning},
journal = {Journal of Medical Artificial Intelligence},
year = {2025},
volume = {8},
pages = {37--37},
doi = {10.21037/jmai-24-356}
}
EUVIMED