Chest X-Ray Images with Corrupted Data for Fault Tolerance Testing

Chest X-ray images containing corrupted or faulty data, likely intended for testing the robustness of computer vision models. The dataset's name suggests a focus on fault tolerance, possibly using a k-fold cross-validation (kcv) structure. Details on the number of images, source, and creation date are unavailable.

Use Cases

• Benchmarking model robustness by training on clean images and evaluating on corrupted x-ray images.
• Developing fault-tolerant image classification algorithms for pneumonia or other chest pathologies using corrupted inputs.
• Testing the impact of specific data corruption types (e.g., noise, artifacts) on diagnostic model accuracy.
• Creating data augmentation pipelines that simulate real-world image corruption for more resilient model training.

Strengths

• Focuses on a specific and important challenge in medical AI: model robustness to corrupted data.
• Likely provides a controlled environment for fault tolerance testing with a defined cross-validation split.

Limitations

• The exact number of images, types of corruption, and severity levels are unknown.
• Lacks metadata on patient demographics, clinical labels, or original image sources.
• Without a clear provenance, the clinical relevance and data quality are uncertain.

Provenance

Collection Method

Likely derived from an existing chest x-ray dataset with artificially introduced corruption.