ProtoPatient Demo

This application demonstrates the ProtoPatient model introduced in our paper:
This Patient Looks Like That Patient: Prototypical Networks for Interpretable Diagnosis Prediction from Clinical Text by Betty van Aken, Jens-Michalis Papaioannou, Marcel G. Naik, Georgios Eleftheriadis, Wolfgang Nejdl, Felix A. Gers and Alexander Löser, published at AACL 2022.
ProtoPatient is trained on MIMIC-III data to predict diagnoses from clinical admission notes. The model outputs support medical decision making by highlighting the most relevant parts in the text for each diagnosis and presenting prototypical patients, which the model decisions are based on.

Note: Because we cannot publish clinical notes from the MIMIC-III database, we use synthetic patient examples and comparable notes from MTSamples.com for this public demo.
< >




Contact: Betty van Aken (@betty_v_a) @ DATEXIS