ResNet-based ECG Diagnosis of Myocardial Infarction in the Emergency Department

Abstract

Abstract In this half time seminar, we focus on the modeling of temporal data with deep learning approaches. We describe two distinct views of modeling temporal data: First, a dynamical system view using input-output time series, where we look at the example of system identification from a deep learning perspective. Second, a prediction view, where the aim is to obtain one point prediction from a single input time series. We concentrate on the application of electrocardiography and the specific challenges we face with this type of data in the real world. We explain how to solve multiple challenges in data processing, problem formulation, modeling and model analysis in this setting.

Links: Slides

Back to page: home.