ProtoPatient: Interpretable Diagnosis Prediction with Prototypical Networks

Talk-Titel: ProtoPatient: Interpretable Diagnosis Prediction with Prototypical Networks
Abstract: The use of deep neural networks for clinical outcome prediction has shown promising results. However, for application in clinical practice, such networks must not only be accurate but provide physicians with interpretable and clinically useful results. Such results include the highlighting of risk factors in a clinical note and the ability to compare against similar patients from past encounters. In this talk, I present our recent approach based on prototypical networks that provides both these abilities and further improves prediction performance compared to standard Transformer baselines.

Speaker: Betty van Aken
Bio: Betty works as an NLP researcher and PhD student at Berliner Hochschule für Technik (BHT). Her current research focuses on transfer learning for clinical NLP with the goal to develop explainable models for clinical decision support.

We are with Averbis this time but have an attendee limit of 20. So please do only confirm if you really attend.

UPDATE (17.6.2022): As there are still some spaces left for tonight, we decided to drop the registration and open up the event.

So if you don't have plans for tonight and if you are interested in NLP + meeting other AI enthusiasts in Freiburg, then you can just show up to the event. We hope to see many of you later.