Resources

This dataset includes bounding box-statement pairs for chest X-ray images, derived from radiologists’ eye-tracking data (for explainability) and annotations, for local visual-language models.

This dataset includes bounding box-statement pairs for chest X-ray images, derived from radiologists’ eye-tracking data (for explainability) and annotations, for local visual-language models.

This dataset contains 3032 cases of eye-tracking data collected while five radiologists dictated reports for frontal chest x-rays, synchronized timestamped dictation transcription, and manual labels for validation of localization of abnormalities.

We present a large-scale CXR VQA dataset derived from MIMIC-CXR with 42M QA pairs, featuring hierarchical answers, bounding boxes, and structured tags. We generated QA-pairs using LLM-based extraction from radiology reports and localization models.

We present a large-scale CXR VQA dataset derived from MIMIC-CXR with 42M QA pairs, featuring hierarchical answers, bounding boxes, and structured tags. We generated QA-pairs using LLM-based extraction from radiology reports and localization models.

A pressure and flow dataset was collected from a trial of 30 adults at the University of Canterbury undergoing CPAP therapy for a variety of instructed breath rates at PEEP levels of 4cmH2O and 7cmH2O.

The Chest ImaGenome dataset is a scene graph dataset with additional chronological comparison relations for chest X-rays. It is automatically derived from the MIMIC-CXR dataset. A manually annotated gold standard is also available for 500 patients.

MS-CXR is a new dataset containing 1162 chest X-ray bounding box labels paired with radiology text descriptions, annotated and verified by two board-certified radiologists.

MS-CXR is a new dataset containing 1162 chest X-ray bounding box labels paired with radiology text descriptions, annotated and verified by two board-certified radiologists.

We created 23 questions resembling eligibility criteria from the apixaban clinical trial and evaluated them on a random sample of 100 patient notes from MIMIC-IV. We release the 2300 total question-answer pairs as a dataset here.