Skip to main content

AI-assisted Diagnostic Tagging, Captioning and Authoring

Save to calendar

Jun 25

Title: AI-assisted Diagnostic Tagging, Captioning and Authoring
Date and time: Thursday 25 June 2020, 12pm – 1pm
Speaker: Ioannis Pavlopoulos

Abstract: In this talk Ioannis Pavlopoulos will present two frameworks, designed to assist medical experts to better manage their workload. During a pandemic, when the workload can be overwhelming, this assistance might be even more valuable. The first framework is based on a neural clinical language model, which can decrease the time of medical report authoring. The second framework employs a deep learning system that automatically assigns abnormalities to X-Rays, but also generates explanatory diagnostic texts in natural language. These two systems form an artificial assistant that can be very useful in days/situations of heavy workload. Consider for example a radiologist who is about to examine hundreds or thousands of patients. Then, by not performing any screening, important medical cases will not be served faster. An alternative scenario, however, would be the following:
a) The assistant assigns automatically (in seconds) abnormality tags to all X-Rays.
b) The assistant selects the top K cases with the most (or most significant) abnormalities.
c) The radiologist examines the selected cases, not only with their automatically suggested tags, but also with an automatically generated explanatory diagnostic text.
d) Finally, when the radiologist starts authoring the medical report, a clinical predictivekeyboard reduces the authoring time and saves from potential language errors.

Bio: Dr. John (Ioannis) Pavlopoulos is a postdoctoral researcher at the Stockholm University and an adjunct professor of Language Technology at the Athens University of Economics and Business, Greece. After completing his PhD (2015), Dr. Pavlopoulos has been researching Natural Language Processing tasks, such as emotion analysis, abusive language detection, medical image captioning, and language generation.