A Wide and Deep Transformer Neural Network for 12-Lead ECG Classification
Широкая и глубокая трансформерная нейронная сеть для классификации 12-канальной ЭКГ
2020-12-30
SCID: 54.1/d9hb7yqa
Discuss with AI
12-lead ECG classification2020 PhysioNet/CinC Challengecardiac abnormality detectiongeneralized weighted accuracywide and deep transformer
Figures from the paper
Abstract (AI)
Cardiac abnormalities are a leading cause of death and their early diagnosis are of importance for providing timely interventions.The goal of 2020 PhysioNet/CinC challenge was to develop algorithms to diagnose multiple cardiac abnormalities using 12-lead ECG data.In this work, we develop a wide and deep transformer neural network to classify each 12-lead ECG sequence into 27 cardiac abnormality classes.Our approach combines handcrafted ECG features, which were determined to be important by a random forest model, and discriminative feature representations that are automatically learned from a transformer neural network.Our entry to the 2020 Phys-ioNet/CinC challenge placed 1 st out of 41 official ranking teams (team name = prna).Using the official generalized weighted accuracy metric for evaluation, we achieved a validation score of 0.587 and top score of 0.533 on the full held-out test set.
Key Findings
1
A wide and deep transformer neural network classifies 12-lead ECG sequences into 27 cardiac abnormality classes.
2
It achieved a generalized weighted accuracy of 0.587 on validation data and 0.533 on the held-out test set.
3
The approach ranked first among 41 official teams in the 2020 PhysioNet/CinC Challenge.
4
The model combines handcrafted ECG features selected by a random forest with discriminative representations learned automatically by a transformer.
Research Object
12-lead ECG sequences
Research Subject
Classification of cardiac abnormalities into 27 classes using handcrafted and transformer-learned features
Publication Details
Publication Date
2020-12-30
Journal
Publisher
ISSN
Cited by
189
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest