Echocardiography in multicenter clinical trials: experience from the Treatment of Mild Hypertension Study

Automated Left Ventricular Dimension Assessment Using Artificial Intelligence Developed and Validated by a UK-Wide Collaborative

BACKGROUND

requires training and validation to standards expected of humans. We developed an online platform and established the Unity Collaborative to build a dataset of expertise from 17 hospitals for training, validation, and standardization of such techniques.

METHODS

The training dataset consisted of 2056 individual frames drawn at random from 1265 parasternal long-axis video-loops of patients undergoing clinical echocardiography in 2015 to 2016. Nine experts labeled these images using our online platform. From this, we trained a convolutional neural network to identify keypoints. Subsequently, 13 experts labeled a validation dataset of the end-systolic and end-diastolic frame from 100 new video-loops, twice each. The 26-opinion consensus was used as the reference standard. The primary outcome was precision SD, the SD of the differences between AI measurement and expert consensus.

RESULTS

In the validation dataset, the AI's precision SD for left ventricular internal dimension was 3.5 mm. For context, precision SD of individual expert measurements against the expert consensus was 4.4 mm. Intraclass correlation coefficient between AI and expert consensus was 0.926 (95% CI, 0.904-0.944), compared with 0.817 (0.778-0.954) between individual experts and expert consensus. For interventricular septum thickness, precision SD was 1.8 mm for AI (intraclass correlation coefficient, 0.809; 0.729-0.967), versus 2.0 mm for individuals (intraclass correlation coefficient, 0.641; 0.568-0.716). For posterior wall thickness, precision SD was 1.4 mm for AI (intraclass correlation coefficient, 0.535 [95% CI, 0.379-0.661]), versus 2.2 mm for individuals (0.366 [0.288-0.462]). We present all images and annotations. This highlights challenging cases, including poor image quality and tapered ventricles.

CONCLUSIONS

Experts at multiple institutions successfully cooperated to build a collaborative AI. This performed as well as individual experts. Future echocardiographic AI research should use a consensus of experts as a reference. Our collaborative welcomes new partners who share our commitment to publish all methods, code, annotations, and results openly.

Reproducibility in echocardiographic assessment of diastolic function in a population based study (the STANISLAS Cohort study)

INTRODUCTION

There is limited evidence regarding intra-observer and inter-observer variations in echocardiographic measurements of diastolic function. This study aimed to assess this reproducibly within a population-based cohort study.

METHODS

Sixty subjects in sinus rhythm were randomly selected among 4th visit participants of the STANISLAS Cohort (Lorraine region, France). This 4th examination systematically included M-mode, 2-dimensional, DTI and pulsed-wave Doppler echocardiograms. Reproducibility of variables was studied by intra-class correlation coefficients (ICC) and Bland Altman plots.

RESULTS

Our population was on average middle-aged (50 ± 14 y), overweight (BMI = 26 ± 6 kg/m2) and non-smoking (87%) with a quarter of the participants having self-declared hypertension or treated with anti-hypertensive medication(s). Intra-observer ICC were > 0.90 for all analyzed parameters except for left ventricular ejection fraction (LVEF) which was 0.89 (0.81-0.93). The mean relative intra-observer differences were small and limits of agreement of relative differences were narrow for all considered parameters (<5% and <15% respectively). Inter-observer ICC were > 0.90 for all analyzed parameters except for LVEF (ICC = 0.87) and both mitral and pulmonary A wave duration (0.83 and 0.73 respectively). The mean relative inter-observer differences were <5% for all parameters except for pulmonary A wave duration (mean difference = 6.5%). Limits of agreement of relative differences were narrow (<15%), except for mitral A wave duration and velocity (both <20%) as well as left ventricular mass and pulmonary A wave duration (both <30%). Intra-observer agreements with regard to the presence and severity of diastolic dysfunction were excellent (Kappa = 0.93 (0.83-1.00) and 0.88 (0.75-0.99), respectively).

CONCLUSION

In this validation study within the STANISLAS cohort, diastolic function echocardiographic parameters were found to be highly reproducible. Diastolic dysfunction consequently appears as a highly effective clinical and research tool.

Low agreement between cardiologists diagnosing left ventricular hypertrophy in children with end-stage renal disease

Background

Monitoring of the appearance of left ventricular hypertrophy (LVH) by echocardiography is currently recommended for in the management of children with End-stage renal disease (ESRD). In order to investigate the validity of this method in ESRD children, we assessed the intra- and inter-observer reproducibility of the diagnosis LVH.

Methods

Echocardiographic measurements in 92 children (0–18 years) with ESRD, made by original analysists, were reassessed offline, twice, by 3 independent observers. Smallest detectable changes (SDC) were calculated for continuous measurements of diastolic interventricular septum (IVSd), Left ventricle posterior wall thickness (LVPWd), Left ventricle end-diastolic diameter (LVEDd), and Left ventricle mass index (LVMI). Cohen’s kappa was calculated to assess the reproducibility of LVH defined in two different ways. LVH_WT was defined as Z-value of IVSd and/or LVPWd>2 and LVH_MI was defined as LVMI> 103 g/m² for boys and >84 g/m² for girls.

Results

The intra-observer SDCs ranged from 1.6 to 1.7 mm, 2.0 to 2.6 mm and 17.7 to 30.5 g/m² for IVSd, LVPWd and LVMI, respectively. The inter-observer SDCs were 2.6 mm, 2.9 mm and 24.6 g/m² for IVSd, LVPWd and LVMI, respectively. Depending on the observer, the prevalence of LVH_WT and LVH_MI ranged from 2 to 30% and from 8 to 25%, respectively. Kappas ranged from 0.4 to 1.0 and from 0.1 to 0.5, for intra-and inter- observer reproducibility, respectively.

Conclusions

Changes in diastolic wall thickness of less than 1.6 mm or LVMI less than 17.7 g/m² cannot be distinguished from measurement error in individual children, even when measured by the same observer. This limits the use of echocardiography to detect changes in wall thickness in children with ESRD in routine practice.

Cardiovascular reactivity and left ventricular mass: An integrative review

Left ventricular hypertrophy has been shown to be an independent predictor of risk for cardiovascular morbidity and mortality. Behavioral scientists have focused on how hemodynamic factors influenced by psychosocial stress may be associated with left ventricular mass (LVM). We reviewed existing studies examining stress-related cardiovascular reactivity (CVR) and LVM, with a goal of examining the moderating role of population (age and hypertensive status) and methodological factors (task type, sample size, and study design) explaining the observed results. Twenty-one studies met the criteria for this review. Results showed only a modestly consistent relationship between CVR and LVM. Forty-three percent of the studies reported 1 or more significant results linking systolic blood pressure reactivity with LVM, and 14% of the studies showed that diastolic blood pressure reactivity was significantly related to LVM. Hypertensive status, task type, and sample size did not play a major role in moderating the relationship between LVM and CVR. A somewhat larger percentage of positive results was shown in prospective and adult studies. The association between CVR and LVM may be real, although the effect size is modest, and we discuss methodological strategies for enhancing statistical power in future investigations. Additional sampling factors (e.g., race, gender) may also impact this relationship. Finally, greater attention is warranted to the role of the psychosocial environment, as this may interact with reactivity to influence LVM.

Comparison of m-mode echocardiographic left ventricular mass measured using digital and strip chart readings: the Atherosclerosis Risk in Communities (ARIC) study

BACKGROUND

Epidemiological and clinical studies frequently use echocardiography to measure LV wall thicknesses and chamber dimension for estimating quantitative measures of LV mass. While echocardiographic M-mode LV images have traditionally been measured using hand-held calipers and strip-chart paper tracings, digitized M-mode LV image measurements made directly on the computer screen using electronic calipers have become standard practice. We sought to determine if systematic differences in LV mass occur between the two methods by comparing LV mass measured from simultaneous M-mode strip chart recordings and digitized recordings.

METHODS

The Atherosclerosis Risk in Communities study applied the latter method. To determine if systematic differences in LV mass occur between the two methods, LV mass was measured from simultaneous M-mode strip chart recordings and digitized recordings.

RESULTS

We found no difference in LV mass (p > .25) and a strong correlation in LV mass between the two methods (r = 0.97). Neither age, sex, nor hypertension status affected the correlation of LV mass between the two methods.

CONCLUSIONS

We conclude that digital estimates of LV mass provide unbiased estimates comparable to the strip-chart method.

Reliability of multicenter pediatric echocardiographic measurements of left ventricular structure and function: the prospective P(2)C(2) HIV study

BACKGROUND

To assess the reliability of pediatric echocardiographic measurements, we compared local measurements with those made at a central facility.

METHODS AND RESULTS

The comparison was based on the first echocardiographic recording obtained on 735 children of HIV-infected mothers at 10 clinical sites focusing on measurements of left ventricular (LV) dimension, wall thicknesses, and fractional shortening. The recordings were measured locally and then remeasured at a central facility. The highest agreement expressed as an intraclass correlation coefficient (ICC=0.97) was noted for LV dimension, with much lower agreement for posterior wall thickness (ICC=0.65), fractional shortening (ICC=0.64), and septal wall thickness (ICC=0.50). The mean dimension was 0.03 cm smaller in central measurements (95% prediction interval [PI], -0.32 to 0.25 cm) for which 95% PI reflects the magnitude of differences between local and central measurements. Mean posterior wall thickness was 0.02 cm larger in central measurements (95% PI, -0.18 to 0.22 cm). Mean fractional shortening was 1% smaller in central measurements. However, the 95% PI was -10% to 8%, indicating that a fractional shortening of 32% measured centrally could be anywhere between 22% and 40% when measured locally. Central measurements of mean septal thickness were approximately 0.1 cm thicker than local ones (95% PI, -0.18 to 0.34 cm). Centrally measured wall thickness was more closely related to mortality and possibly was more valid than local measurements.

CONCLUSIONS

Although LV dimension was reliably measured, local measurements of LV wall thickness and fractional shortening differed from central measurements.

Reversal of pathophysiologic changes with long-term lisinopril treatment in isolated systolic hypertension

The purpose of this study was to evaluate in a prospective, double-blind, placebo-controlled study the effect of long-term (2-year) lisinopril treatment on cardiovascular end-organ damage in patients with previously untreated isolated systolic hypertension (ISH). All patients with ISH were derived from a population screening program. End-organ damage measurements, done initially and after 6 and 24 months of treatment, included measurements of aortic distensibility and echocardiographic left ventricular mass index (LVMI) and diastolic function. Blood pressure was measured by office and ambulatory measurements. Of the 97 subjects with ISH selected from the screening, 62 (30 lisinopril) completed the study according to protocol. Office blood pressure decreased in both groups, but ambulatory results significantly decreased with lisinopril-treatment only. Aortic distensibility increased significantly with lisinopril, as opposed to a decrease in placebo-treated subjects. The main effect of increased distensibility occurred between 6 and 24 months, whereas ambulatory blood pressure changed mainly in the first 6 months of treatment. LVMI decreased in both treatment groups, with a significantly higher reduction in lisinopril-treated subjects. Left ventricular diastolic function showed no significant changes in either group. The vascular pathophysiologic alterations of ISH-a decreased aortic distensibility-can be improved with long-term lisinopril treatment, whereas values deteriorate further in placebo-treated subjects. These results, in one of the first studies including subjects with previously untreated ISH only, indicate that lisinopril treatment might favorably influence the cardiovascular risk of ISH.

Validity of echocardiographic measurement in an epidemiological study. Project HeartBeat!

In Project HeartBeat!, a longitudinal study of cardiovascular disease risk factors in healthy children and adolescents, 3 samples of 40, 80, and 182 echocardiograms, respectively, were randomly selected and reread to evaluate intraobserver and interobserver variabilities and comparability between measurements of field echocardiographic technicians and reference readings at Texas Children's Hospital. Included in the evaluation were 8 M-mode echocardiographic measurements, ie, aortic root diameter, left atrial diameter, and end-diastolic and end-systolic measurements of interventricular septal thickness, left ventricular (LV) diameter, and LV posterior wall thickness; 8 Doppler measurements; and a calculated LV mass. Means and SDs of the differences of the paired measurements were used to assess the relative bias and random error of the measurements. For the intraobserver comparison, means and SDs of the differences were very small, indicating that the echo measurements were performed consistently by each project echo technician. Interobserver comparison showed statistically but not clinically significant differences between the paired readings of end-diastolic septal thickness, end-systolic LV posterior wall thickness, and 5 Doppler measurements. Comparison with reference readings at Texas Children's Hospital showed significant differences in diastolic LV diameter, systolic septal thickness, and right ventricular ejection time. These differences, however, were minimal with limited clinical significance. Mean differences in LV mass for the corresponding comparisons were -1.82, 4.50, and 0.0013 g, and the SDs were 18.79, 24.16, and 12.35 g, respectively. We conclude that the echocardiographic measurements taken from healthy children in a longitudinal study can be made accurately with acceptable reproducibility.

What determines echogenicity in a general population? The Tromsø Study

BACKGROUND

It is widely recognized that in some people it is difficult or impossible to acquire adequate measurements of cardiac performance and anatomy by any echocardiographic technique. We used our population-based screening to determine the characteristics of such unmeasurable subjects.

METHOD

In a sample of 3287 subjects aged 25 to 85 years, we used standard 2-dimensional guided M-mode echocardiography and pulsed and color Doppler to assess left ventricular (LV) structure and function.

RESULTS

Of 3287 subjects only 0.4% could not be measured by any technique. In 2794 subjects M-mode registrations of good quality were obtained, which allowed calculation of LV mass and LV ejection fraction. Those in whom measurements could not be obtained had a significantly higher age, body mass index, blood pressure, waist/hip ratio, and were more likely to smoke, be a man, be taking antihypertensive medication, have a history of ischemic heart disease, and have a low level of physical activity.

CONCLUSION

Because subjects with high cardiovascular risk factor levels are less likely to be measurable with echocardiography, a need exists for other noninvasive diagnostic methods in these persons.

Ambulatory blood pressure monitoring and echocardiography--noninvasive techniques for evaluation of the hypertensive patient

Clinic blood pressure measurements have only limited ability to determine which hypertensive patients are at greatest risk of cardiovascular events. Ambulatory blood pressure monitoring allows for noninvasive measurement of blood pressure throughout the 24-hour period. This may help to clarify discrepancies between blood pressure values obtained in and out of the clinic and confirm the presence of white-coat hypertension, broadly defined as an elevated clinic blood pressure but a normal ambulatory blood pressure. Ambulatory blood pressure values have been shown to have a better relationship to cardiovascular morbidity and mortality and end-organ damage than clinic blood pressure values. Further, patients with white-coat hypertension appear to be at greater risk of cardiovascular morbidity and end-organ damage than a normotensive population, although they are at less overall risk than a hypertensive population. Hypertensive heart disease is characterized by diastolic dysfunction, increased left ventricular mass, and coronary flow abnormalities. Left ventricular hypertrophy increases the risk of coronary heart disease, congestive heart failure, stroke, ventricular arrhythmias, and sudden death. A variety of invasive and noninvasive techniques are described herein that measure left ventricular mass, diastolic function, and coronary blood flow abnormalities. Most antihypertensive treatments promote regression of left ventricular hypertrophy and reversal of diastolic dysfunction, which may decrease symptoms of congestive heart failure and improve survival.

Interobserver Variability of Left Ventricular Measurements in a Population of Predominantly Obese Hypertensives Using Simultaneously Acquired and Displayed M-Mode and 2-D Cine Echocardiography

Achieving low interobserver variability is a goal of echocardiographic determined left ventricular (LV) mass measurements. In a group of hypertensives, we evaluated interobserver variation using a method of simultaneously acquired two-dimensional (2-D) cine and M-mode images displayed in a split screen format. Sixty echocardiographic images from ongoing hypertension trials, including serial studies of the same patients, were obtained with an UltraMarktrade mark 6 Ultrasound System (Advanced Technology Laboratories, Inc., Bothell, WA, USA). Images were digitized online, coded with a random number, and pooled prior to the analysis to minimize observer bias. Studies were read by two independent observers in a blinded fashion and in a random order using a Color Vue II Nova MicroSonicstrade mark analyzer (Nova Microsonics, Mahwah, NJ, USA). The M-mode tracing and cine of three consecutive frames of 2-D parasternal short-axis views of the LV were simultaneously displayed in a split screen format. The 2-D cine was used as a reference image during M-mode measurements of LV dimensions. Measurements were obtained and the LV mass estimated according to the Penn convention. Interobserver variability for left ventricular internal dimension (LVID), interventricular septal thickness (IVS), posterior wall thickness (PWT), and left ventricular mass (LV mass) is low when either correlated (Pearson correlation coefficients of 0.94, 0.82, 0.75, and 0.93, respectively) or expressed as a percent of the mean (3.0%, 10.0%, 10.2%, and 8.9%, respectively). When read in a blinded fashion, interobserver variability (especially for LV mass) is small using digitized, simultaneously acquired and displayed cines of 2-D and M-mode echocardiograms. This is likely due to the ability to discriminate myocardial wall edges (endocardium) from other associated structures such as tricuspid and mitral apparatus. This method may prove useful in studies of LV mass.

Comparison of five antihypertensive monotherapies and placebo for change in left ventricular mass in patients receiving nutritional-hygienic therapy in the Treatment of Mild Hypertension Study (TOMHS)

BACKGROUND

Increased left ventricular mass (LVM) by echocardiography is associated with increased risk of cardiovascular disease. Thus, it is of interest to compare the effects of both pharmacological and nonpharmacological approaches to the treatment of hypertension on reduction of LVM.

METHODS AND RESULTS

Changes in LV structure were assessed by M-mode echocardiograms in a double-blind, placebo-controlled clinical trial of 844 mild hypertensive participants randomized to nutritional-hygienic (NH) intervention plus placebo or NH plus one of five classes of antihypertensive agents: (1) diuretic (chlorthalidone), (2) beta-blocker (acebutolol), (3) alpha-antagonist (doxazosin mesylate), (4) calcium antagonist (amlodipine maleate), or (5) angiotensin-converting enzyme inhibitor (enalapril maleate). Echocardiograms were performed at baseline, at 3 months, and annually for 4 years. Changes in blood pressure averaged 16/12 mm Hg in the active treatment groups and 9/9 mm Hg in the NH only group. All groups showed significant decreases (10% to 15%) in LVM from baseline that appeared at 3 months and continued for 48 months. The chlorthalidone group experienced the greatest decrease at each follow-up visit (average decrease, 34 g), although the differences from other groups were modest (average decrease among 5 other groups, 24 to 27 g). Participants randomized to NH intervention only had mean changes in LVM similar to those in the participants randomized to NH intervention plus pharmacological treatment. The greatest difference between groups was seen at 12 months, with mean decreases ranging from 35 g (chlorthalidone group) to 17 g (acebutolol group) (P = .001 comparing all groups). Within-group analysis showed that changes in weight, urinary sodium excretion, and systolic BP were moderately correlated with changes in LVM, being statistically significant in most analyses.

CONCLUSIONS

NH intervention with emphasis on weight loss and reduction of dietary sodium is as effective as NH intervention plus pharmacological treatment in reducing echocardiographically determined LVM, despite a smaller decrease in blood pressure in the NH intervention only group. A possible exception is that the addition of diuretic (chlorthalidone) may have a modest additional effect on reducing LVM.