3.3 Artificial intelligence detects myocardial infarctions in the ECG more reliably than cardiologistsn

The Deutsche Gesellschaft für Kardiologie - Herz- und Kreislaufforschung e.V.(DKG) developed an algorithm for machine learning that, with the help of ECG interpretation, can detect a myocardial infarction that has occurred within the last 3 months more reliably than a cardiologist.

Figure from (Makimoto 2019)

Project details:

  • Used deep learning
  • Train to answer the following question:
    • Is this an ECG of a patient who’s had an Myocardial infarction in the last three months?
  • Compared with 9 cardiologist
    • algorithm is more reliable
  • Explainability of ML prediction was addressed

To better understand how the ML algorithm deducted the prediction a heatmap of the ECG was produced. The red areas show the focus of the ML algorithm.

Heatmap of ECG shows the focus of ML algorithm, figure from (Makimoto 2019)

3.3.1 First in Germany: Artificial intelligence recognizes COVID-19 in clinical routine

Another application using ML in for diagnostic is to detect COVID-19, details can be found at https://www.uniklinikum-jena.de/Uniklinikum+Jena/Aktuelles/Pressemitteilungen/Erstmals+in+Deutschland_+Künstliche+Intelligenz+erkennt+COVID_19+in+der+klinischen+Routine-pos-3-p-24500.html

References

Makimoto, Hisaki. 2019. “Künstliche Intelligenz Erkennt Myokardinfarkte Im EKG Zuverlässiger Als Kardiologen.”