Forskning
Udskriv Udskriv
Switch language
Region Hovedstaden - en del af Københavns Universitetshospital
Udgivet

Explaining deep neural networks for knowledge discovery in electrocardiogram analysis

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

DOI

  1. Infants with congenital heart defects have reduced brain volumes

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  2. Extrarenal expression of α-klotho, the kidney related longevity gene, in Heterocephalus glaber, the long living Naked Mole Rat

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  3. In vitro efficacy of artemisinin-based treatments against SARS-CoV-2

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  4. Low-dose aspirin for primary and secondary prevention of cardiovascular events in Denmark 1998-2018

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  5. Acute and persistent symptoms in non-hospitalized PCR-confirmed COVID-19 patients

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  1. FGL1 as a modulator of plasma D-dimer levels: exome-wide marker analysis of plasma tPA, PAI-1 and D-dimer

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  2. Response to Roldan et al

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  3. Atrial fibrillation-a complex polygenetic disease

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  4. Determinants of penetrance and variable expressivity in monogenic metabolic conditions across 77,184 exomes

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

  5. The Association of Low Vitamin K Status with Mortality in a Cohort of 138 Hospitalized Patients with COVID-19

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

Vis graf over relationer

Deep learning-based tools may annotate and interpret medical data more quickly, consistently, and accurately than medical doctors. However, as medical doctors are ultimately responsible for clinical decision-making, any deep learning-based prediction should be accompanied by an explanation that a human can understand. We present an approach called electrocardiogram gradient class activation map (ECGradCAM), which is used to generate attention maps and explain the reasoning behind deep learning-based decision-making in ECG analysis. Attention maps may be used in the clinic to aid diagnosis, discover new medical knowledge, and identify novel features and characteristics of medical tests. In this paper, we showcase how ECGradCAM attention maps can unmask how a novel deep learning model measures both amplitudes and intervals in 12-lead electrocardiograms, and we show an example of how attention maps may be used to develop novel ECG features.

OriginalsprogEngelsk
Artikelnummer10949
TidsskriftScientific Reports
Vol/bind11
Udgave nummer1
Sider (fra-til)10949
ISSN2045-2322
DOI
StatusUdgivet - 26 maj 2021
Eksternt udgivetJa

Bibliografisk note

Funding Information:
This work is funded in part by Novo Nordisk Foundation project number NNF18CC0034900.

ID: 65837298