Automatic Childhood Pneumonia Diagnosis Based on Multi-Model Feature Fusion Using Chi-Square Feature Selection
Amira Ouerhani1, Tareq Hadidi2, Hanene Sahli3
1Research Laboratory of Biophysics and Medical Technologies, Higher Institute of Medical Technologies of Tunis, University of Tunis El Manar, Tunis 1006, Tunisia.
Journal of Imaging
|February 26, 2026
Summary
This study introduces an advanced deep learning method for detecting childhood pneumonia using fused features from CNN models. The approach significantly enhances diagnostic accuracy, aiding in timely and efficient detection.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Healthcare
- Pediatric Diagnostics
Background:
- Childhood pneumonia is a leading cause of mortality, with chest radiography (CXR) crucial for diagnosis.
- Accurate pneumonia detection in pediatric CXR is challenging due to low radiation exposure and limitations of traditional methods.
- Deep learning, specifically Convolutional Neural Networks (CNNs), shows promise in improving medical image analysis.
Purpose of the Study:
- To propose an accurate pneumonia detection method for pediatric chest X-rays.
- To leverage deep CNN architectures and optimal feature fusion for enhanced diagnostic performance.
- To improve the timely detection of childhood pneumonia, particularly in resource-limited healthcare settings.
Main Methods:
- Training enhanced VGG-19, ResNet-50, and MobileNet-V2 models on a large pneumonia dataset using transfer learning.
- Applying the Chi-Square technique to remove irrelevant features from each CNN model.
- Performing horizontal feature fusion of selected features from VGG-19 and MobileNet-V2 for binary classification using a Support Vector Machine (SVM) classifier.
Main Results:
- The proposed fused-feature approach achieved high performance metrics: 97.59% accuracy, 98.33% recall, and 98.19% F1-score on the test set.
- The method demonstrated significant performance enhancement compared to existing ensemble fusion techniques.
- The system ensures computational efficiency while providing robust feature representation.
Conclusions:
- The developed fused-feature deep CNN system offers a significant advancement in the accurate and timely detection of childhood pneumonia.
- This approach holds potential for widespread application, especially in healthcare systems with limited resources.
- The study highlights the effectiveness of combining deep learning models and feature fusion for pediatric diagnostic imaging.
