ID | 114196 |
Title Alternative | Deep Learning for Echocardiography
|
Author |
Kusunose, Kenya
Tokushima University
Tokushima University Educator and Researcher Directory
KAKEN Search Researchers
Yamada, Hirotsugu
Tokushima University
Tokushima University Educator and Researcher Directory
KAKEN Search Researchers
Harada, Masafumi
Tokushima University
Tokushima University Educator and Researcher Directory
KAKEN Search Researchers
Sata, Masataka
Tokushima University
Tokushima University Educator and Researcher Directory
KAKEN Search Researchers
|
Keywords | echocardiography
artificial intelligence
regional wall motion abnormality
diagnostic ability
|
Content Type |
Journal Article
|
Description | Objectives: The aim of this study was to evaluate whether a deep convolutional neural network (DCNN) could detect regional wall motion abnormalities (RWMAs) and differentiate groups of coronary infarction territories from conventional 2-dimensional echocardiographic images compared with cardiologist/sonographer or resident readers.
Background: An effective intervention for reduction of misreading of RWMAs is needed. We hypothesized that a DCNN trained with echocardiographic images may provide improved detection of RWMAs in the clinical setting. Methods: A total of 300 patients with history of myocardial infarction were enrolled. In this cohort, 100 each had infarctions of the left anterior descending branch (LAD), left circumflex branch (LCX), and right coronary artery (RCA). The age-matched 100 control patients with normal wall motion were selected from our database. Each case contained cardiac ultrasound images from short axis views at end-diastolic, mid-systolic and end-systolic phases. After 100 steps of training, diagnostic accuracies were calculated on the test set. We independently trained 10 versions of the same model, and performed ensemble predictions with them. Results: For detection of the presence of wall motion abnormality, the area under the receiver-operating characteristic curve (AUC) by deep learning algorithm was similar to that by cardiologist/sonographer readers (0.99 vs. 0.98, p =0.15), and significantly higher than the AUC by resident readers (0.99 vs. 0.90, p =0.002). For detection of territories of wall motion abnormality, the AUC by the deep learning algorithm was similar to the AUC by cardiologist/sonographer readers (0.97 vs. 0.95, p =0.61) and significantly higher than the AUC by resident readers (0.97 vs. 0.83, p =0.003). In a validation group from an independent site (n=40), the AUC by the DL algorithm was 0.90. Conclusions: Our results support the possibility of DCNN use for automated diagnosis of RWMAs in the field of echocardiography. |
Journal Title |
JACC : Cardiovascular Imaging
|
ISSN | 1936878X
|
NCID | AA1234416X
AA1279525X
|
Publisher | American College of Cardiology Foundation|Elsevier
|
Volume | 13
|
Issue | 2-1
|
Start Page | 374
|
End Page | 381
|
Published Date | 2019-05-15
|
Rights | © 2019. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/
|
EDB ID | |
DOI (Published Version) | |
URL ( Publisher's Version ) | |
FullText File | |
language |
eng
|
TextVersion |
Author
|
departments |
University Hospital
Medical Sciences
|