Use of Artificial Intelligence in Ultrasound Diagnosis of Fetal Central Nervous System Anomalies Between 19 and 22 Weeks’ Gestation

• Published in: Innovacionnaâ medicina Kubani (Online) no. 2; pp. 42 - 47
• Main Authors: Pomortsev, A. V., Redko, A. N., Barsukova, E. A., Matosyan, M. A., Dyachenko, J. Yu, Dyachenko, R. A., Beloglyadova, I. A., Yanaeva, M. V., Babayan, V. T.
• Format: Journal Article
• Language: English; Russian
• Published: Scientific Research Institute, Ochapovsky Regional Clinical Hospital no. 1 28-04-2024
• Subjects: artificial intelligence; fetal anomalies; health care digitalization; neural network; ultrasound diagnosis
• ISSN: 2541-9897; 2541-9897
• DOI: 10.35401/2541-9897-2024-9-2-42-47

Introduction: The use of modern technologies, including artificial intelligence (AI), in medical imaging is a current hot topic. Standardization of radiologic and ultrasound scans of the studied area is a prerequisite for implementation of this computer system. It is important to create and apply AI in ultrasound diagnosis of fetal central nervous system (CNS) anomalies in order to improve the quality of differential diagnosis.
Objective
: To evaluate diagnostic accuracy of AI in detecting fetal CNS anomalies between 19 and 22 weeks’ gestation. Materials and methods: We conducted a multicenter 2-stage study to evaluate AI effectiveness in detecting fetal CNS anomalies between 19 and 22 weeks’ gestation. At stage I, more than 1500 pregnant women underwent sonographic examination of the fetal head in the axial plane with 5 anatomical landmarks, and we recorded a 15-second video in the MP4 format (video sequence). At stage II, we tested “Decision-Making System for Detecting Fetal Central Nervous System Anomalies” to determine its diagnostic accuracy.
Results
: The diagnostic accuracy of the developed software (“Formulation of an Imaging-Based Diagnosis of Fetal Central Nervous System Anomalies”) in regard to such parameters as “normal findings” and “abnormal findings” was 78.9%. The diagnostic accuracy for formulation of a specific imaging-based diagnosis was 74.4%.
Conclusions
: The AI implemented into modern ultrasound differential diagnosis of fetal CNS anomalies between 19 and 22 weeks’ gestation will make it possible to formulate an imaging-based diagnosis (“normal findings”/“abnormal findings”) with high accuracy and can be used as an additional computer technology in the primary screening of pregnant women.