Ferdinand Dhombres1, Nathalie Roux2, Stéphanie Friszer2, Roger Bessis3, Babak Khoshnood4, Jean-Marie Jouannic3. 1. Sorbonne Universités, UPMC Univ Paris 06, Fetal Medicine Department in Armand Trousseau University Hospital, APHP, Paris, France; French College of Fetal Echography (CFEF), France. Electronic address: ferdinand.dhombres@aphp.fr. 2. Sorbonne Universités, UPMC Univ Paris 06, Fetal Medicine Department in Armand Trousseau University Hospital, APHP, Paris, France. 3. Sorbonne Universités, UPMC Univ Paris 06, Fetal Medicine Department in Armand Trousseau University Hospital, APHP, Paris, France; French College of Fetal Echography (CFEF), France. 4. INSERM UMR 1153, Obstetrical, Perinatal and Pediatric Epidemiology Research Team (EPOPé), Center for Epidemiology and Statistics, Sorbonne Paris Cité (CRESS), DHU Risks in Pregnancy, Paris Descartes University, Paris, France.
Abstract
OBJECTIVE: To assess the extent to which the distribution of crown-rump length (CRL) values may be correlated with different criteria for the quality of the CRL images. STUDY DESIGN: This is a retrospective analysis of a series of 977 CRL images, by two independent observers, for the presence or the absence of 14 quality hallmarks. Inter-observer agreement for the hallmarks was assessed by the proportion of agreement and Cohen's kappa. The association between the quantiles of the CRL distribution and the presence or absence of the 14 quality hallmarks was modeled using quantile regression. RESULTS: The overall inter-observer agreement across the 14 hallmarks was 91.7%, kappa=0.81, 95% CI [0.80-0.82]. Distribution of CRL measurements varied considerably as a function of image quality: when the fetus was in extension, the mean CRL was +5.7mm (vs. not in extension, p<0.001), when the fetus was in flexion (vs. not), the mean CRL was -4.7mm (p<0.001) and when the image magnification was <65% (vs. >65%), the mean CRL was -4.2mm (p<0.001). There was a global trend to over-estimate the CRL for the higher deciles and to under-estimate the CRL for the lower deciles when the sagittal quality hallmarks were absent. No significant impact on CRL distribution was observed in association with the precise placement of the calipers nor with the horizontal orientation of the fetus. CONCLUSION: Distribution of CRL measurements was influenced by the quality of CRL images. In particular, inadequate position of the fetus (flexion/extension) and insufficient image magnification were associated with systematic changes in the values of CRL. Our results show that as the quality of CRL images decreases, the associated variations in the distribution of CRL can have an impact on the chromosomal risk assessment and may lead to inappropriate obstetrical decisions.
OBJECTIVE: To assess the extent to which the distribution of crown-rump length (CRL) values may be correlated with different criteria for the quality of the CRL images. STUDY DESIGN: This is a retrospective analysis of a series of 977 CRL images, by two independent observers, for the presence or the absence of 14 quality hallmarks. Inter-observer agreement for the hallmarks was assessed by the proportion of agreement and Cohen's kappa. The association between the quantiles of the CRL distribution and the presence or absence of the 14 quality hallmarks was modeled using quantile regression. RESULTS: The overall inter-observer agreement across the 14 hallmarks was 91.7%, kappa=0.81, 95% CI [0.80-0.82]. Distribution of CRL measurements varied considerably as a function of image quality: when the fetus was in extension, the mean CRL was +5.7mm (vs. not in extension, p<0.001), when the fetus was in flexion (vs. not), the mean CRL was -4.7mm (p<0.001) and when the image magnification was <65% (vs. >65%), the mean CRL was -4.2mm (p<0.001). There was a global trend to over-estimate the CRL for the higher deciles and to under-estimate the CRL for the lower deciles when the sagittal quality hallmarks were absent. No significant impact on CRL distribution was observed in association with the precise placement of the calipers nor with the horizontal orientation of the fetus. CONCLUSION: Distribution of CRL measurements was influenced by the quality of CRL images. In particular, inadequate position of the fetus (flexion/extension) and insufficient image magnification were associated with systematic changes in the values of CRL. Our results show that as the quality of CRL images decreases, the associated variations in the distribution of CRL can have an impact on the chromosomal risk assessment and may lead to inappropriate obstetrical decisions.
Authors: N Fries; F Dhombres; M Massoud; J J Stirnemann; R Bessis; G Haddad; L J Salomon Journal: BMC Pregnancy Childbirth Date: 2021-02-27 Impact factor: 3.007