Creation and Validation of a Chest X-Ray Dataset with Eye-tracking and Report Dictation for AI Development

We developed a rich dataset of Chest X-Ray (CXR) images to assist\ninvestigators in artificial intelligence. The data were collected using an eye\ntracking system while a radiologist reviewed and reported on 1,083 CXR images.\nThe dataset contains the following aligned data: CXR image, transcribed\nradiology report text, radiologist's dictation audio and eye gaze coordinates\ndata. We hope this dataset can contribute to various areas of research\nparticularly towards explainable and multimodal deep learning / machine\nlearning methods. Furthermore, investigators in disease classification and\nlocalization, automated radiology report generation, and human-machine\ninteraction can benefit from these data. We report deep learning experiments\nthat utilize the attention maps produced by eye gaze dataset to show the\npotential utility of this data.\n