Discussion

The operators appear to have systematically included LVOT images in their screening practices, as images of the LVOT were found in 95% of the files studied. With regards to image quality, practices have improved since 2017, but there was no difference over the study period (first quarter of 2020) since the introduction of the quality criteria. Practitioners who were already aware of the use of LVOT images before the new recommendations continued to use them systematically, and probably more accurately after 2016. Scores were not significantly higher in the second trimester relative to the third trimester. Given the results of the Var center, echocardiography screening activity in the primary center is just as satisfactory as in the tertiary center. The scores obtained according to cardiac position show a significant improvement in average scores for Positions 2 and 3 when the back of the foetus is in the posterior position between 3 and 9 o’clock. It was noted that the Quality Criteria 5, 6 and 7 were very poorly considered over the three periods, regardless of cardiac position.
In addition to obtaining the required images, it is relevant to carry out a “quality” screening examination. This notion of quality presupposes that the images obtained are as informative as possible. As early as 2003, Chaoui advocated quality control of foetal echocardiograms in order to avoid inadequate examinations (7). Certification, similar to that required for first-trimester assessments, could be considered in an attempt to improve screening. Indeed, in some countries, regular audits for maintaining certification or quality control have been successful (12)(13)(14). Since then, the notion of quality control and auditing has developed in the field of echocardiography, with first-trimester cardiac screening examinations in low-risk populations (15), as well as cardio-thoracic ratio measurement (16), the use of four-chamber views (17)(18) and the outflow tracts (17)(19). With regards to the 4-chamber view, apical or basal views are recommended in order to respect the quality criteria proposed(18). However, to a lesser extent, side views are also recommended. For the LVOT, one of the most important criteria is visualization of the interventricular septum and, in particular, its relationship to the possible overlying vessel. This area of interest is best visualized in 2D, when the ultrasound beam approaches it perpendicularly and not, as is too often the case, by taking a perpendicular approach to the lower part of the interventricular septum (Figure 11 ), with the risk of misinterpreting interventricular communication of the conical region, as shown by our example in Figures 12 and 13 . This principle of taking a perpendicular approach to the anatomical region on either side of the aortic valves is fundamental, whether the approach is anterior or posterior. It is therefore preferable to use a lateral approach to the LVOT. This underlines the importance of a dynamic examination by means of scanning and angulation methods for probes producing the ultrasound beam (5)(11)(20) and this point was also already highlighted by Sklansky and De Vore (8) .
The detection of conotruncal anomalies should and could be improved (21). This leads us to question the causes of such errors during ultrasound screening. Several factors, unrelated to the operator, have already been highlighted by different teams: gestational age at the time of ultrasound, maternal adipose tissue, history of abdominal surgery, foetal position, quantity of amniotic fluid, position of the placenta,etc. (22)(23). Other limiting factors directly attributable to sonographers have also been demonstrated: the experience of the operator (23), the number of ultrasound scans carried out annually, the inability to obtain the required images or so-called “quality” images and the inability of the operator to recognize unusual images despite obtaining quality anatomical images. The latter concepts have again been reported recently by Sun et al. (24) and Van Nisselrooij et al.(9), who also confirm that the main current limitation of CHD screening is linked to in poor image quality for other outflow tracts.
In the light of these observations and the relative consistency in reported CHD screening, it would appear to be appropriate to accept the extent of training programmes for sonographers, however, it is essential to propose solutions which should accompany these programmes. It is with this in mind that Artificial Intelligence (AI) has been developed. Currently, the most relevant model in automated learning of image analysis is the convolutional neural network (25)(26). AI could thus help the practitioner by carrying out a real-time audit of the images provided (presence or absence and quality), helping to obtain recommended quality anatomical images (27), but also in terms of improving image resolution (28)(29)(30). AI could thus be of great help, not from a diagnostic perspective, but rather as a support tool for the practitioner and a guarantee of quality, analogous to a seatbelt in a car, as it may be associated with a potentially high negative predictive value regarding the risk of major CHD.

The strengths of our study

. This was a multicenter study based in different centers with practitioners from different backgrounds and with different levels of experience, providing an example of current screening practices in the same country in tertiary and primary centers.
. This work focused on the quality of echocardiographic screening and the causes associated with the current limitations of screening for CHD.
. Inter- and intra-operator variability regarding the proposed criteria was low, as the kappa coefficients were all reported at between 0.601 and 1. Concordance ranged from good to very good, thus the scoring of the images was carried out correctly and the proposed criteria were easy to use.
. The evaluation of the quality criteria proposed was carried out in routine practice by operators who probably have different degrees of motivation in their desire to use new tools aimed at improving screening practices. Indeed, most of the studies relating to the study of quality criteria were carried out in tertiary centers with operators working in close contact with the person(s) capable of ascertaining the quality criteria. We thus acknowledge that the reality may be quite different, and this sheds new light on what underlies the current limitations of CHD screening.
. Quality criteria for LVOT screening should be distinguished from normality criteria. For example, for Criterion 6, the quality criterion is not the presence of septo-aortic continuity but the visualization of the relationship between the interventricular septum and the possible overlying vessel, whether or not there is continuity between the interventricular septum and this possible vessel.

Limitations of the study

. It is not possible to know whether the lack of images in the files reflects a lack of investigation of the area in question, or whether the area was investigated but no image was available and stored to demonstrate this.
. Does a significant difference in quality score truly reflect a lack of, or a real change in practice, with a real clinical impact? Indeed, assessment of the only relevant difference would be based on rates of detection of heart defects recorded in registers.
. The lack of significant improvement in scores after the introduction of the quality criteria raises questions about their relevance and use, but also about the techniques associated with guidance on how to implement them. Practitioners may have been made aware of the contribution of the quality criteria even if this was not reflected in the analysis. It is clear that we should try to understand the causes behind the non-compliance of these criteria: difficulty in understanding them, in remembering to use them and/or being aware of them.