Artificial Intelligence in Pediatric and Fetal Echocardiography: Looking Beyond the Now

Perioperative Echocardiography During the Coronavirus Crisis: Considerations in Pediatrics and Congenital Heart Disease

Echocardiography is the main modality in diagnosing acquired and congenital heart disease (CHD) in fetal and pediatric patients. However, operator variability, complex image interpretation, and lack of experienced sonographers and cardiologists in certain regions are the main limitations existing in fetal and pediatric echocardiography. Advances in artificial intelligence (AI), including machine learning (ML) and deep learning (DL), offer significant potential to overcome these challenges by automating image acquisition, image segmentation, CHD detection, and measurements. Despite these promising advancements, challenges such as small number of datasets, algorithm transparency, physician comfort with AI, and accessibility must be addressed to fully integrate AI into practice. This review highlights AI's current applications, challenges, and future directions in fetal and pediatric echocardiography.

We sought to determine the incremental diagnostic utility of pediatric cardiac assessment in the offspring of women with congenital heart disease who have had previous fetal echocardiography. We prospectively followed pregnant women with congenital heart disease who were receiving care at 2 obstetric and cardiac centers and identified 276 infants who underwent both fetal echocardiography and pediatric cardiac assessment. All of the infants with abnormal fetal echocardiography findings or abnormal pediatric cardiac assessments underwent subsequent confirmatory pediatric echocardiography. In this cohort, congenital heart disease was detected in 22 (8%) of 276 offspring born to women with congenital heart disease. There was concordance between the results of fetal echocardiography and pediatric cardiac assessment in 235 (85%) of 276 offspring (231, both normal; 4, both abnormal) and discordance between the results of fetal echocardiography and pediatric cardiac assessment in 41 (15%) of 276 infants. In the 41 subjects with discordant results, there were normal fetal echocardiography findings but abnormal pediatric cardiac assessments in 35 of 41 (pediatric echocardiography revealed congenital heart disease in 18 of 35 and normal anatomy in 17 of 35) and abnormal fetal echocardiography findings but normal pediatric cardiac assessments in 6 of 41 (pediatric echocardiography findings normal in all 6 of the infants). Fetal echocardiography detected all of the major forms of congenital heart disease. Lesions missed by fetal echocardiography but detected on pediatric cardiac assessment included shunt lesions and minor valvular abnormalities. Although fetal echocardiography can reliably exclude major forms of congenital heart disease, minor congenital heart disease lesions can be missed on fetal echocardiography; however, these can be diagnosed with careful pediatric cardiac assessment. Postnatal pediatric cardiac assessment has incremental diagnostic utility for the detection of congenital heart disease in the offspring of women with congenital heart disease and previous fetal echocardiography.

2021 Refocus - The Revolution of Imaging in Pediatric and Congenital Heart Disease

ACC/AHA/AAP RECOMMENDATIONS FOR TRAINING IN PEDIATRIC CARDIOLOGY

Artificial intelligence in echocardiography to diagnose congenital heart disease and fetal echocardiography

Artificial Intelligence (AI) models have the potential to revolutionize fetal echocardiography, enhancing image processing and diagnostic capabilities. There is promise in AI's potential to improve the accuracy and efficiency of congenital heart defect detection in utero. There is significant interest towards Machine learning’s enhanced capabilities in automated segmentation of fetal cardiac structures, reducing inter-observer variability, enhancing image quality and early detection of congenital heart diseases (CHD). Deep learning models have exhibited high sensitivity and specificity in identifying various cardiac anomalies. AI-assisted fetal echocardiography has also demonstrated potential in reducing false-positive rates and minimizing unnecessary interventions, improving fetal survival rates. Traditional fetal echocardiogram has been shown to struggle from limitations such as fetal, maternal and equipment factors, which paves the way for advanced AI models to step-in. This review was conducted in accordance with the “Preferred Reporting Items for Systematic Reviews and Meta-analysis” (PRISMA) guidelines. An extensive search was conducted in all the major medical databases for relevant articles concerning machine learning algorithms and its application in fetal echocardiography. This was followed by a detailed systematic review to derive and analyse the relevant qualitative and quantitative data. The statistical analysis was performed in R-Studio. Pooled accuracy was assessed using the metaprop function. The heterogeneity of the papers was assessed using Higgins' I^2 test. The meta-analysis included a total of 9 papers, amounting to 106275 quantitative elements concerning image processing and 6088 elements on the diagnostic and predictive faculties of the AI algorithms, resulting in a pooled accuracy of 0.90 [0.84;0.93, 95% CI, p<0.0001] and 0.99 [0.66;1.00, 95% CI, p<0.0001] respectively. The highest individual accuracy amounted to 93.32% by Zhang et al for image processing, while an accuracy of 100% by Nurmaini et al was the highest for diagnostic prediction. The random effects model was used to compare and assess the quantitative data. The review also highlighted the individual qualitative characteristics of the AI models. AI's utility in terms of fortifying fetal echocardiography has been statistically shown to be clinically significant, as evident by its high accuracy. AI's adoption into various clinical aspects has been shown to be promising.

Chapter 14 - Artificial intelligence in three-dimensional and fetal echocardiography

As a cardiologist in a group practice or hospital settings, many a times you are requested to comment on fetal Echocardiography performed by radiologist or fetal medicine specialist. With overconfidence, we perform or comment on fetal echocardiography without knowing critical differences between adult and fetal echocardiography. First, fetal heart at 20 weeks of gestation is of the size of an almond. Due to the limitation of resolution of ultrasound machines, small size of fetal heart makes it difficult to visualize many direct signs of congenital heart defects such as total anomalous pulmonary venous connection (TAPVC), transposition of the great arteries (TGA), coarctation, and double outlet right ventricle (RV). Hence, in fetal echo, we use indirect signs and clues to diagnose these disorders. Second, there are two physiological shunts in fetal circulation; fossa ovalis and ductus arteriosus which make its very different from adult circulation. A cardiologist needs to know about indirect signs, special views, peculiar fetal cardiac defects, and hemodynamics of fetal circulation before attempting fetal echocardiography. Many of us have an impression that fetal heart is just a miniature form of adult heart. Fetal echocardiography is a lot different from adult or pediatric echocardiography because there are many structural and functional differences in fetal circulation. Moreover, many congenital heart defects such as TGA, TAPVC, and ventricular septal defect, can present with only subtle findings, so we need to be more vigilant while performing fetal echocardiography.

Congenital heart defects (CHDs) are the most common congenital defect, occurring in approximately 1 in 100 live births and being a leading cause of perinatal morbidity and mortality. Of note, approximately 25% of these defects are classified as critical, requiring immediate postnatal care by pediatric cardiology and neonatal cardiac surgery teams. Consequently, early and accurate diagnosis of CHD is key to proper prenatal and postnatal monitoring in a tertiary care setting. In this scenario, fetal echocardiography is considered the gold standard imaging ultrasound method for the diagnosis of CHD. However, the availability of this examination in clinical practice remains limited due to the need for a qualified specialist in pediatric cardiology. Moreover, in light of the relatively low prevalence of CHD among at-risk populations (approximately 10%), ultrasound cardiac screening for potential cardiac anomalies during routine second-trimester obstetric ultrasound scans represents a pivotal aspect of diagnosing CHD. In order to maximize the accuracy of CHD diagnoses, the views of the ventricular outflow tract and the superior mediastinum were added to the four-chamber view of the fetal heart for routine ultrasound screening according to international guidelines. In this context, four-dimensional spatio-temporal image correlation software (STIC) was developed in the early 2000s. Some of the advantages of STIC in fetal cardiac evaluation include the enrichment of anatomical details of fetal cardiac images in the absence of the pregnant woman and the ability to send volumes for analysis by an expert in fetal cardiology by an internet link. Sequentially, new technologies have been developed, such as fetal intelligent navigation echocardiography (FINE), also known as "5D heart", in which the nine fetal cardiac views recommended during a fetal echocardiogram are automatically generated from the acquisition of a cardiac volume. Furthermore, artificial intelligence (AI) has recently emerged as a promising technological innovation, offering the potential to warn of possible cardiac anomalies and thus increase the ability of non-cardiology specialists to diagnose CHD. In the early 2010s, the advent of 3D reconstruction software combined with high-definition printers enabled the virtual and 3D physical reconstruction of the fetal heart. The 3D physical models may improve parental counseling of fetal CHD, maternal-fetal interaction in cases of blind pregnant women, and interactive discussions among multidisciplinary health teams. In addition, the 3D physical and virtual models can be an useful tool for teaching cardiovascular anatomy and to optimize surgical planning, enabling simulation rooms for surgical procedures. Therefore, in this review, the authors discuss advanced image technologies that may optimize prenatal diagnoses of CHDs.

Guidelines for Physician Training in Fetal Echocardiography: Recommendations of the Society of Pediatric Echocardiography Committee on Physician Training

ASE Statement on the Reintroduction of Echocardiographic Services during the COVID-19 Pandemic

Detailed follow-up in a series of 276 mothers with congenital heart disease--whose children are at increased risk of heart disease--demonstrated that, although a fetal echocardiography is reliable in detecting major disease, some minor lesions are undetectable until after birth. Most of these minor defects will be diagnosed by routine clinical cardiac examination and can be confirmed by pediatric echocardiography.

The past year was dominated by developments in echocardiographic assessment of myocardial function in children, and this paper reviews some of these major advancements in pediatric echocardiography as a guide to those interested in imaging of the pediatric heart and vascular system. Characterization of myocardial function using Doppler tissue imaging, both in the normal child and in the child with congenital or acquired heart disease, was a primary focus of pediatric echocardiographic investigation. Other new technologies, including integrated backscatter analysis and three-dimensional echocardiography, appear to hold significant promise as tools to improve myocardial assessment echocardiographically. Three-dimensional echocardiography also is developing into a powerful technique in fetal echocardiography, allowing rapid data acquisition and extensive image postprocessing with opportunities for both anatomic and functional assessment. Childhood fitness and diseases have a significant impact on the heart and vascular bed, and descriptions of echocardiographic findings in obese children, children engaged in athletic activities, children with renal disease, children who have undergone cardiac transplantation, and those with aortic valve disease were better characterized by investigations published over the past year. Pediatric echocardiography has clearly expanded from a diagnostic tool used to describe anatomic abnormalities associated with congenital heart disease to a noninvasive myocardial monitoring tool that allows serial assessment of the pathologic effects of both cardiac and noncardiac disease.

Advances in the Application of Artificial Intelligence in Fetal Echocardiography