Eye-tracking Annotations for Deep Learning Radiology Applications
This project is NIH-funded, started in 2020, and aims to collect eye-tracking data as a non-intrusive way of providing localization information on labels associated with each chest x-ray. The data collection will include timestamped transcriptions of the dictations o radiological reports, making the association between region looked and word said at a given moment possible. The project's steps include:
- providing all collected data as an open dataset,
- evaluating interpretability methods for deep learning methods when compared to the attention from radiologists, and
- using collected data as supervision for novel deep learning models.