Artificial intelligence (AI) algorithms are set to improve the early detection of congenital heart defects (CHD) during the prenatal stage. Identifying these issues early is crucial for a newborn’s prognosis. The Heart for Children Foundation is spearheading a program to prepare medical professionals for the integration of AI in analyzing fetal echocardiography data. Promising results already suggest that AI can effectively assist doctors in making more accurate diagnoses. Polish researchers are also developing technology that analyzes imaging data similarly to how a human specialist would.
“Unfortunately, congenital heart defects are not always identified, with an average detection rate of around 40%. If a defect is detected early and the birth takes place in a prepared facility, the prognosis for the child can be very good. However, if the birth occurs in an unprepared center without prior diagnosis, even if the child is later transferred to a specialized facility, the early postnatal outcomes may significantly influence the long-term effectiveness of treatment,” explains Dr. Natalia Mazanowska, a perinatologist from the Clinic of Obstetrics and Gynecology at the Mother and Child Institute.
Training Healthcare Professionals for AI Integration
Through the InteliCardio program, the Heart for Children Foundation is conducting training sessions aimed at equipping healthcare workers with the skills to leverage AI systems for improving CHD diagnostics. These sessions explore how AI technologies can enhance early detection, improve diagnostic accuracy, and enable faster processing of medical data.
“We are constantly seeking methods to improve our ability to detect congenital heart defects in children before birth. AI algorithms, I hope, will soon be implemented and developed in Poland. They will serve as a supportive tool for doctors during the diagnostic process. These algorithms do not independently diagnose heart defects, but they help doctors determine whether an image that appears abnormal truly deviates from the norm and whether the patient should be referred to a specialized center or undergo expert-level testing,” says Dr. Mazanowska.
Promising Research on AI-Assisted Imaging
In November 2024, the journal BMC Pregnancy and Childbirth published a study on the potential of AI in combination with ultrasound imaging for diagnosing ventricular septal defects (VSD) in fetal hearts. The study analyzed over 1,400 ultrasound images from 500 pregnancies between January 2016 and June 2022. Researchers trained an AI model using a five-fold cross-validation method, comparing its diagnostic accuracy with doctors of varying experience levels.
The findings showed that the AI tool improved the diagnostic accuracy of less-experienced doctors by 6.7% and mid-level doctors by 2.8%. For highly experienced doctors, the AI matched their diagnostic accuracy but significantly reduced the time required—from minutes to milliseconds.
“In the area of congenital heart defects, AI will have enormous applications, as our initial analyses have shown. This is particularly important because doctors in training have limited access to cases of fetuses with congenital heart defects. AI can serve as a ‘virtual assistant’ to help standardize ultrasound recordings and provide structured data for analysis. Such tools could ultimately categorize cases that require expert evaluation,” explains Prof. Marcin Wiecheć, Chair of the Ultrasound Section of the Polish Society of Gynecology and Obstetrics and an expert at Jagiellonian University’s Collegium Medicum.
Developing AI Algorithms with Local Expertise
A Polish AI company is partnering with InteliCardio to develop algorithms that collaborate with medical professionals. One such algorithm utilizes 3D imaging analysis, functioning similarly to how a cardiologist interprets imaging data.
“An expert can instantly recognize a heart defect even from a quick echocardiographic scan of a fetus and identify its category. Similarly, we can train an algorithm if provided with sufficient data. For three years, we have been working on teaching an algorithm to recognize echocardiographic images in the same way autonomous vehicles learn to navigate. We annotated heart structures layer by layer in a painstaking process, and the algorithm is now capable of independently improving its analysis as more data is provided. This ‘virtual assistant’ is continuously evolving to offer even greater precision,” adds Prof. Wiecheć.
A Growing Market for AI in Medical Imaging
According to Mordor Intelligence, the global market for AI-based medical imaging is expected to reach $7.52 billion by 2025, with projections exceeding $26 billion by 2030.
As AI continues to demonstrate its potential in prenatal diagnostics, such advancements could revolutionize the early detection and management of congenital heart defects, ultimately improving outcomes for countless newborns worldwide.
Source: Heart for Children Foundation, InteliCardio Program, BMC Pregnancy and Childbirth, Mordor Intelligence
