Eye-tracking Annotations for Deep Learning Radiology Applications

In this ongoing project, funded by NIBIB/NIH, we collected eye-tracking data, as a proof-of-concept of a non-intrusive way of providing localization information on labels associated with each chest x-ray read by radiologists.

The REFLACX (Reports and Eye-Tracking Data for Localization of Abnormalities in Chest X-rays) dataset contains 3,032 cases of eye-tracking data collected while a radiologist dictated reports for images from the MIMIC-CXR dataset, paired and synchronized with timestamped transcriptions of the dictations. It also contains manual labels for each image, including bounding boxes localizing the lungs and heart and validation labels of image-level abnormalities and localization of abnormalities through drawn ellipses.

Code used for collecting, processing and analyzing the data: https://github.com/ricbl/eyetracking
Link to the dataset: https://physionet.org/content/reflacx-xray-localization/1.0.0/
Dataset description paper: https://arxiv.org/abs/2109.14187

Contacts

PI: Tolga Tasdizen, This email address is being protected from spambots. You need JavaScript enabled to view it.
Research assistant: Ricardo Bigolin Lanfredi, This email address is being protected from spambots. You need JavaScript enabled to view it.
Lead radiologist: Joyce D. Schroeder, This email address is being protected from spambots. You need JavaScript enabled to view it.