5,856 pediatric chest X-ray images categorized into 'Normal' and 'Pneumonia' classes. The collection includes 1,583 normal cases and 4,273 pneumonia cases, with the latter further subdivided into bacterial and viral etiologies.
Use Cases
- Train a binary classifier to detect the presence of pneumonia using the 'Normal' and 'Pneumonia' folder labels
- Build a diagnostic tool to differentiate between bacterial and viral pneumonia using the specific metadata encoded in the image filenames
- Benchmark image augmentation techniques on medical imaging data to improve model generalization across the 624 test images
Strengths
- 5,856 JPEG images organized into train, test, and validation directories
- Labels distinguish between bacterial and viral pneumonia types within the file structure and naming conventions
- Sourced from pediatric patients at Guangzhou Women and Children’s Medical Center
- Images were screened for quality control by removing low-quality or unreadable scans prior to labeling