Automated analysis of fetal cardiac function using color tissue Doppler imaging

Automated quantitative evaluation of fetal atrioventricular annular plane systolic excursion

OBJECTIVES

The primary aim of this study was to evaluate the feasibility of automated measurement of fetal atrioventricular (AV) plane displacement (AVPD) over several cardiac cycles using myocardial velocity traces obtained by color tissue Doppler imaging (cTDI). The secondary objectives were to establish reference ranges for AVPD during the second half of normal pregnancy, to assess fetal AVPD in prolonged pregnancy in relation to adverse perinatal outcome and to evaluate AVPD in fetuses with a suspicion of intrauterine growth restriction (IUGR).

METHODS

The population used to develop the reference ranges consisted of women with an uncomplicated singleton pregnancy at 18-42 weeks of gestation (n = 201). The prolonged-pregnancy group comprised women with an uncomplicated singleton pregnancy at ≥ 41 + 0 weeks of gestation (n = 107). The third study cohort comprised women with a singleton pregnancy and suspicion of IUGR, defined as an estimated fetal weight < 2.5^th centile or an estimated fetal weight < 10^th centile and umbilical artery pulsatility index > 97.5^th centile (n = 35). Cineloops of the four-chamber view of the fetal heart were recorded using cTDI. Regions of interest were placed at the AV plane in the left and right ventricular walls and the interventricular septum, and myocardial velocity traces were integrated and analyzed using an automated algorithm developed in-house to obtain mitral (MAPSE), tricuspid (TAPSE) and septal (SAPSE) annular plane systolic excursion. Gestational-age specific reference ranges were constructed and normalized for cardiac size. The correlation between AVPD measurements obtained using cTDI and those obtained by anatomic M-mode were evaluated, and agreement between these two methods was assessed using Bland-Altman analysis. The mean Z-scores of fetal AVPD in the cohort of prolonged pregnancies were compared between cases with normal and those with adverse outcome using Mann-Whitney U-test. The mean Z-scores of fetal AVPD in IUGR fetuses were compared with those in the normal reference population using Mann-Whitney U-test. Inter- and intraobserver variability for acquisition of cTDI recordings and offline analysis was assessed by calculating coefficients of variation (CV) using the root mean square method.

RESULTS

Fetal MAPSE, SAPSE and TAPSE increased with gestational age but did not change significantly when normalized for cardiac size. The fitted mean was highest for TAPSE throughout the second half of gestation, followed by SAPSE and MAPSE. There was a significant correlation between MAPSE (r = 0.64; P < 0.001), SAPSE (r = 0.72; P < 0.001) and TAPSE (r = 0.84; P < 0.001) measurements obtained by M-mode and those obtained by cTDI. The geometric means of ratios between AVPD measured by cTDI and by M-mode were 1.38 (95% limits of agreement (LoA), 0.84-2.25) for MAPSE, 1.00 (95% LoA, 0.72-1.40) for SAPSE and 1.20 (95% LoA, 0.92-1.57) for TAPSE. In the prolonged-pregnancy group, the mean ± SD Z-scores for MAPSE (0.14 ± 0.97), SAPSE (0.09 ± 1.02) and TAPSE (0.15 ± 0.90) did not show any significant difference compared to the reference ranges. Twenty-one of the 107 (19.6%) prolonged pregnancies had adverse perinatal outcome. The AVPD Z-scores were not significantly different between pregnancies with normal and those with adverse outcome in the prolonged-pregnancy cohort. The mean ± SD Z-scores for SAPSE (-0.62 ± 1.07; P = 0.006) and TAPSE (-0.60 ± 0.89; P = 0.002) were significantly lower in the IUGR group compared to those in the normal reference population, but the differences were not significant when the values were corrected for cardiac size. The interobserver CVs for the automated measurement of MAPSE, SAPSE and TAPSE were 28.1%, 17.7% and 15.3%, respectively, and the respective intraobserver CVs were 33.5%, 15.0% and 17.9%.

CONCLUSIONS

This study showed that fetal AVPD can be measured automatically by integrating cTDI velocities over several cardiac cycles. Automated analysis of AVPD could potentially help gather larger datasets to facilitate use of machine-learning models to study fetal cardiac function. The gestational-age associated increase in AVPD is most likely a result of increasing cardiac size, as the AVPD normalized for cardiac size did not change significantly between 18 and 42 weeks. A decrease was seen in TAPSE and SAPSE in IUGR fetuses, but not after correction for cardiac size. © 2021 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Fetal cardiac function at intrauterine transfusion assessed by automated analysis of color tissue Doppler recordings

BACKGROUND

Fetal anemia is associated with a hyperdynamic circulation and cardiac remodeling. Rapid intrauterine transfusion (IUT) of blood with high hematocrit and viscosity into the umbilical vein used to treat this condition can temporarily further affect fetal heart function. The aim of this study was to evaluate the short-term changes in fetal myocardial function caused by IUT using automated analysis of cine-loops of the fetal heart obtained by color tissue Doppler imaging (cTDI).

METHODS

Fetal echocardiography was performed before and after IUT. cTDI recordings were obtained in a four-chamber view and regions of interest were placed at the atrioventricular plane in the left ventricular (LV), right ventricular (RV) and septal walls. Myocardial velocities were analyzed by an automated analysis software to obtain peak myocardial velocities during atrial contraction (Am), ventricular ejection (Sm), rapid ventricular filling (Em) and Em/Am ratio was calculated. Myocardial velocities were converted to z-scores using published reference ranges. Delta z-scores (after minus before IUT) were calculated. Correlations were assessed between variables and hemoglobin before IUT.

RESULTS

Thirty-two fetuses underwent 70 IUTs. Fourteen were first time transfusions. In the LV and septal walls, all myocardial velocities were significantly increased compared to normal values, whereas in the RV only Sm was increased before IUT (z-scores 0.26-0.52). In first time IUTs, there was a negative correlation between LV Em (rho = - 0.61, p = 0.036) and LV Em/Am (rho = - 0.82, p = 0.001) z-scores and hemoglobin before IUT. The peak myocardial velocities that were increased before IUT decreased, whereas LV Em/Am increased significantly after IUT.

CONCLUSIONS

This study showed that peak myocardial velocities assessed by cTDI are increased in fetuses before IUT reflecting the physiology of hyperdynamic circulation. In these fetuses, the fetal heart is able to adapt and efficiently handle the volume load caused by IUT by altering its myocardial function.

Automated analysis of fetal cardiac function using color tissue Doppler imaging in second half of normal pregnancy

OBJECTIVES

Color tissue Doppler imaging (cTDI) is a promising tool for the assessment of fetal cardiac function. However, the analysis of myocardial velocity traces is cumbersome and time-consuming, limiting its application in clinical practice. The aim of this study was to evaluate fetal cardiac function during the second half of pregnancy and to develop reference ranges using an automated method to analyze cTDI recordings from a cardiac four-chamber view.

METHODS

This was a cross-sectional study including 201 normal singleton pregnancies between 18 and 42 weeks of gestation. During fetal echocardiography, a four-chamber view of the heart was visualized and cTDI was performed. Regions of interest were positioned at the level of the atrioventricular plane in the left ventricular (LV), right ventricular (RV) and septal walls of the fetal heart, to obtain myocardial velocity traces that were analyzed offline using the automated algorithm. Peak myocardial velocities during atrial contraction (Am), ventricular ejection (Sm) and rapid ventricular filling, i.e. early diastole (Em), as well as the Em/Am ratio, mechanical cardiac time intervals and myocardial performance index (cMPI) were evaluated, and gestational age-specific reference ranges were constructed.

RESULTS

At 18 weeks of gestation, the peak myocardial velocities, presented as fitted mean with 95% CI, were: LV Am, 3.39 (3.09-3.70) cm/s; LV Sm, 1.62 (1.46-1.79) cm/s; LV Em, 1.95 (1.75-2.15) cm/s; septal Am, 3.07 (2.80-3.36) cm/s; septal Sm, 1.93 (1.81-2.06) cm/s; septal Em, 2.57 (2.32-2.84) cm/s; RV Am, 4.89 (4.59-5.20) cm/s; RV Sm, 2.31 (2.16-2.46) cm/s; and RV Em, 2.94 (2.69-3.21) cm/s. At 42 weeks of gestation, the peak myocardial velocities had increased to: LV Am, 4.25 (3.87-4.65) cm/s; LV Sm, 3.53 (3.19-3.89) cm/s; LV Em, 4.55 (4.18-4.94) cm/s; septal Am, 4.49 (4.17-4.82) cm/s; septal Sm, 3.36 (3.17-3.55) cm/s; septal Em, 3.76 (3.51-4.03) cm/s; RV Am, 6.52 (6.09-6.96) cm/s; RV Sm, 4.95 (4.59-5.32) cm/s; and RV Em, 5.42 (4.99-5.88) cm/s. The mechanical cardiac time intervals generally remained more stable throughout the second half of pregnancy, although, with increased gestational age, there was an increase in duration of septal and RV atrial contraction, LV pre-ejection and septal and RV ventricular ejection, while there was a decrease in duration of septal postejection. Regression equations used for the construction of gestational age-specific reference ranges for peak myocardial velocities, Em/Am ratios, mechanical cardiac time intervals and cMPI are presented.

CONCLUSION

Peak myocardial velocities increase with gestational age, while the mechanical time intervals remain more stable throughout the second half of pregnancy. Using an automated method to analyze cTDI-derived myocardial velocity traces, it was possible to construct reference ranges, which could be used in distinguishing between normal and abnormal fetal cardiac function. Copyright © 2018 ISUOG. Published by John Wiley & Sons Ltd.

The Incremental Benefit of Color Tissue Doppler in Fetal Arrhythmia Assessment

BACKGROUND

Accurate fetal arrhythmia (FA) diagnosis is key for effective management. Currently, FA assessment relies on standard echocardiography-based techniques (M mode and spectral Doppler), which require adequate fetal position and cursor alignment to define temporal relationships of mechanical events. Few data exist on the application of color Doppler tissue imaging (c-DTI) in FA assessment. The aim of this study was to examine the feasibility and clinical applicability of c-DTI in FA assessment in comparison with standard techniques.

METHODS

Pregnancies with diagnosed FA were prospectively recruited to undergo c-DTI following fetal echocardiography. Multiple-cycle four-chamber clips in any orientation were recorded (mean frame rate, 180 ± 16 frames/sec). With offline analysis, sample volumes were placed on atrial (A) and ventricular (V) free walls for simultaneous recordings. Atrial and ventricular rates, intervals (for atrial-ventricular conduction and tachyarrhythmia mechanism), and relationships were assessed to decipher FA mechanism. FA diagnosis by c-DTI, conventional echocardiographic techniques, and postnatal electrocardiography and/or Holter monitoring were compared.

RESULTS

FA was assessed by c-DTI in 45 pregnancies at 15 to 39 weeks, including 16 with atrial and/or ventricular ectopic beats; 18 with supraventricular tachyarrhythmias, including ectopic atrial tachycardia in 11, atrioventricular reentrant tachycardia in four, atrial flutter in two, and intermittent atrial flutter and junctional ectopic rhythm in one; three with ventricular tachycardias; and eight with bradycardias or atrioventricular conduction pathology, including five with complete atrioventricular block (AVB), one with first-degree AVB evolving into complete AVB, one with second-degree AVB, and one with sinus bradycardia. After training, FA diagnosis by c-DTI could be made irrespective of fetal orientation within 10 to 15 min. FA diagnosis by c-DTI concurred with standard techniques in 41 cases (91%), with additional findings identified by c-DTI in 10. c-DTI led to new FA diagnoses in four cases (9%) not definable by standard techniques. FA diagnosis by c-DTI was confirmed in all 20 with persistent arrhythmias after birth, including three with new diagnoses defined by c-DTI. c-DTI was particularly helpful in deciphering SVT mechanism (long vs short ventricular-atrial interval) in all 18 cases, whereas standard techniques permitted definition in only half.

CONCLUSIONS

c-DTI with offline analysis permits rapid and accurate definition of FA mechanism, providing new information in nearly one-third of affected pregnancies.

Intraoperative Renal Resistive Index as an Acute Kidney Injury Biomarker: Development and Validation of an Automated Analysis Algorithm

OBJECTIVE

Intraoperative Doppler-determined renal resistive index (RRI) is a promising early acute kidney injury (AKI) biomarker. As RRI continues to be studied, its clinical usefulness and robustness in research settings will be linked to the ease, efficiency, and precision with which it can be interpreted. Therefore, the authors assessed the usefulness of computer vision technology as an approach to developing an automated RRI-estimating algorithm with equivalent reliability and reproducibility to human experts.

DESIGN

Retrospective.

SETTING

Single-center, university hospital.

PARTICIPANTS

Adult cardiac surgery patients from 7/1/2013 to 7/10/2014 with intraoperative transesophageal echocardiography-determined renal blood flow measurements.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

Renal Doppler waveforms were obtained retrospectively and assessed by blinded human expert raters. Images (430) were divided evenly into development and validation cohorts. An algorithm for automated RRI analysis was built using computer vision techniques and tuned for alignment with experts using bootstrap resampling in the development cohort. This algorithm then was applied to the validation cohort for an unbiased assessment of agreement with human experts. Waveform analysis time per image averaged 0.144 seconds. Agreement was excellent by intraclass correlation coefficient (0.939; 95% confidence interval [CI] 0.921 to 0.953) and in Bland-Altman analysis (mean difference [human-algorithm] -0.0015; 95% CI -0.0054 to 0.0024), without evidence of systematic bias.

CONCLUSION

The authors confirmed the value of computer vision technology to develop an algorithm for RRI estimation from automatically processed intraoperative renal Doppler waveforms. This simple-to-use and efficient tool further adds to the clinical and research value of RRI, already the "earliest" among several early AKI biomarkers being assessed.