TY - JOUR
T1 - Fetal electrocardiography and artificial intelligence for prenatal detection of congenital heart disease
AU - de Vries, Ivar R.
AU - van Laar, Judith O. E. H.
AU - van der Hout-van der Jagt, Marieke B.
AU - Clur, Sally-Ann B.
AU - Vullings, Rik
N1 - Funding Information: This research was partly funded by ZonMW as part of the Topspecialistische Zorg en Onderzoek (TZO) project “BIRTHSCREEN” (project number 10070022010001). Publisher Copyright: © 2023 The Authors. Acta Obstetricia et Gynecologica Scandinavica published by John Wiley & Sons Ltd on behalf of Nordic Federation of Societies of Obstetrics and Gynecology (NFOG).
PY - 2023/11
Y1 - 2023/11
N2 - Introduction: This study aims to investigate non-invasive electrocardiography as a method for the detection of congenital heart disease (CHD) with the help of artificial intelligence. Material and methods: An artificial neural network was trained for the identification of CHD using non-invasively obtained fetal electrocardiograms. With the help of a Bayesian updating rule, multiple electrocardiographs were used to increase the algorithm's performance. Results: Using 122 measurements containing 65 healthy and 57 CHD cases, the accuracy, sensitivity, and specificity were found to be 71%, 63%, and 77%, respectively. The sensitivity was however 75% and 69% for CHD cases requiring an intervention in the neonatal period and first year of life, respectively. Furthermore, a positive effect of measurement length on the detection performance was observed, reaching optimal performance when using 14 electrocardiography segments (37.5 min) or more. A small negative trend between gestational age and accuracy was found. Conclusions: The proposed method combining recent advances in obtaining non-invasive fetal electrocardiography with artificial intelligence for the automatic detection of CHD achieved a detection rate of 63% for all CHD and 75% for critical CHD. This feasibility study shows that detection rates of CHD might improve by using electrocardiography-based screening complementary to the standard ultrasound-based screening. More research is required to improve performance and determine the benefits to clinical practice.
AB - Introduction: This study aims to investigate non-invasive electrocardiography as a method for the detection of congenital heart disease (CHD) with the help of artificial intelligence. Material and methods: An artificial neural network was trained for the identification of CHD using non-invasively obtained fetal electrocardiograms. With the help of a Bayesian updating rule, multiple electrocardiographs were used to increase the algorithm's performance. Results: Using 122 measurements containing 65 healthy and 57 CHD cases, the accuracy, sensitivity, and specificity were found to be 71%, 63%, and 77%, respectively. The sensitivity was however 75% and 69% for CHD cases requiring an intervention in the neonatal period and first year of life, respectively. Furthermore, a positive effect of measurement length on the detection performance was observed, reaching optimal performance when using 14 electrocardiography segments (37.5 min) or more. A small negative trend between gestational age and accuracy was found. Conclusions: The proposed method combining recent advances in obtaining non-invasive fetal electrocardiography with artificial intelligence for the automatic detection of CHD achieved a detection rate of 63% for all CHD and 75% for critical CHD. This feasibility study shows that detection rates of CHD might improve by using electrocardiography-based screening complementary to the standard ultrasound-based screening. More research is required to improve performance and determine the benefits to clinical practice.
KW - artificial intelligence
KW - congenital heart disease
KW - fetal electrocardiography
KW - fetal heart
KW - prenatal diagnosis
UR - http://www.scopus.com/inward/record.url?scp=85167652005&partnerID=8YFLogxK
U2 - https://doi.org/10.1111/aogs.14623
DO - https://doi.org/10.1111/aogs.14623
M3 - Article
C2 - 37563851
SN - 0001-6349
VL - 102
SP - 1511
EP - 1520
JO - Acta obstetricia et gynecologica Scandinavica
JF - Acta obstetricia et gynecologica Scandinavica
IS - 11
ER -