Generate a report that provides diagnostic inferences by identifying hidden causalities in MIMIC CXR Reports. The MIMIC CXR Reports and (optional) DICOM Chest X-ray images of each case are analyzed by a task participant's model to create a causality report, which is then validated against the correct answer data. The goal is to produce a causality report that reflects the way that a medical professional would diagnose.
This task involves generating a causal explanation report for chest X-ray findings based on structured input derived from radiologists’ responses to a questionnaire. Unlike Task 1, Task 2 does not require the MIMIC license, as it relies solely on the answers to a specific set of questions provided by radiologists through crowdsourcing. This approach aims to enable participants without MIMIC data access to engage with the shared task while still emphasizing the clinical interpretative process.
Contribution to explainable CAD technology.
Contribution to causal interpretation techniques in chest radiology report generation.
Contribution to automatic annotation technology for generating datasets for causal interpretation of chest radiology images.
Contribution to a technology that achieves equivalent performance with fewer resources than GPT.