A machine learning approach on chest X-rays for pediatric pneumonia detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Pediatric Pneumonia Diagnosis

Background:

• Pneumonia is a leading infectious cause of death in children under 5.
• Early detection of pediatric pneumonia is critical for reducing mortality.
• Chest X-ray interpretation for pediatric pneumonia has poor interobserver agreement, necessitating automated detection.

Purpose of the Study:

• To automate the early detection of pediatric pneumonia.
• To utilize machine learning (ML) for its lower computational demand and higher interpretability compared to deep learning (DL).

Main Methods:

• Data augmentation was used to balance the dataset classes.
• Feature extraction schemes were optimized.
• Multiple ML models were evaluated and compared against a transfer learning (TL) benchmark.

Main Results:

• The Quadratic Support Vector Machine (SVM) model achieved 97.58% accuracy, outperforming existing ML literature.
• The ML model demonstrated significantly faster classification times than the TL benchmark.

Conclusions:

• The proposed ML approach is highly effective for reliable pediatric pneumonia detection.
• This method offers a computationally efficient and interpretable alternative for automated pneumonia diagnosis.