112,120 frontal-view X-ray images from 30,805 unique patients categorized into 14 distinct thoracic pathology labels. The dataset includes metadata such as patient age, gender, and view position (AP/PA) alongside multi-label disease classifications.
Use Cases
- Train a multi-label classification model to detect thoracic diseases using the Finding Labels column and image files
- Analyze demographic trends in pathology prevalence by correlating Patient Age and Patient Gender with specific disease labels
- Evaluate the impact of radiographic projection on diagnostic accuracy by comparing images labeled with AP versus PA in the View Position column
- Develop patient-specific progression models by grouping records by Patient ID and ordering them by the Follow-up # column
Strengths
- 112,120 PNG images annotated with 14 pathology classes including Atelectasis, Cardiomegaly, and Pneumothorax
- Includes metadata for Patient Age, Patient Gender, and View Position (AP or PA)
- Contains 30,805 unique Patient IDs allowing for longitudinal study of disease progression
- Multi-label classification format within the Finding Labels column allows for identifying co-occurring conditions