Automated Identification of Patent Ductus Arteriosus Using a Computer Vision Model.

OBJECTIVES: Patent ductus arteriosus (PDA) is a vascular defect common in preterm infants and often requires treatment to avoid associated long-term morbidities. Echocardiography is the primary tool used to diagnose and monitor PDA. We trained a deep learning model to identify PDA presence in relevant echocardiographic images. METHODS: Echocardiography video clips (n = 2527) in preterm infants were reviewed by a pediatric cardiologist and those relevant to PDA diagnosis were selected and labeled (PDA present/absent/indeterminate). We trained a convolutional neural network to classify each echocardiography frame of a clip as belonging to clips with or without PDA. A novel attention mechanism that aggregated predictions for all frames in each clip to obtain a clip-level prediction by weighting relevant frames. RESULTS: In early model iterations, we discovered training with color Doppler echocardiography clips produced the best performing classifier. For model training and validation, 1145 such clips from 66 patients (661 PDA+ clips, 484 PDA- clips) were used. Our best classifier for clip level performance obtained sensitivity of 0.80 (0.83-0.90), specificity of 0.77 (0.62-0.92) and AUC of 0.86 (0.83-0.90). Study level performance obtained sensitivity of 0.83 (0.72-0.94), specificity of 0.89 (0.79-1.0) and AUC of 0.93 (0.89-0.98). CONCLUSIONS: Our novel deep learning model demonstrated strong performance in classifying echocardiography clips with and without PDA. Further model development and external validation are warranted. Ultimately, integration of such a classifier into auto detection software could streamline PDA imaging workflow. This work is the first step toward semi-automated, bedside detection of PDA in preterm infants.

Longitudinal measures of deformation are associated with a composite measure of contractility derived from pressure-volume loop analysis in children.

Aims: The relationship between echocardiographic measures of left ventricular (LV) systolic function and reference-standard measures have not been assessed in children. The objective of this study was to assess the validity of echocardiographic indices of LV systolic function via direct comparison to a novel composite measure of contractility derived from pressure-volume loop (PVL) analysis. Methods and results: Children with normal loading conditions undergoing routine left heart catheterization were prospectively enrolled. PVLs were obtained via conductance catheters. A composite invasive composite contractility index (ICCI) was developed using data reduction strategies to combine four measures of contractility derived from PVL analysis. Echocardiograms were performed immediately after PVL analysis under the same anesthetic conditions. Conventional and speckle-tracking echocardiographic measures of systolic function were measured. Of 24 patients, 18 patients were heart transplant recipients, 6 patients had a small patent ductus arteriosus or small coronary fistula. Mean age was 9.1 ± 5.6 years. Upon multivariable regression, longitudinal strain was associated with ICCI (ß = -0.54, P = 0.02) while controlling for indices of preload, afterload, heart rate, and LV mass under baseline conditions. Ejection fraction and shortening fraction were associated with LV mass and load indices, but not contractility. Conclusion: Speckle-tracking derived longitudinal strain is associated ICCI in children with normal loading conditions. Longitudinal measures of deformation appear to accurately assess LV contractility in children.

Echocardiographic Detection of Increased Ventricular Diastolic Stiffness in Pediatric Heart Transplant Recipients: A Pilot Study.

BACKGROUND: Pediatric heart transplant recipients are at risk for increased left ventricular (LV) diastolic stiffness. However, the noninvasive evaluation of LV stiffness has remained elusive in this population. The objective of this study was to compare novel echocardiographic measures of LV diastolic stiffness versus gold-standard measures derived from pressure-volume loop (PVL) analysis in pediatric heart transplant recipients. METHODS: Patients undergoing left heart catheterization were prospectively enrolled. PVLs were obtained via conductance. The end-diastolic pressure-volume relationship was obtained via balloon occlusion. The stiffness constant, ß, was calculated. Echocardiographic measures of diastolic function were derived from spectral and tissue Doppler and two-dimensional speckle-tracking. Ventricular volumes were measured using three-dimensional echocardiography. The novel echocardiographic estimates of ventricular stiffness included E:e'/end-diastolic volume (EDV) and E:early diastolic strain rate/EDV. RESULTS: Of 24 children, 18 were heart transplant recipients. Six control patients had hemodynamically insignificant patent ductus arteriosus or coronary fistula. The mean age was 9.1 ± 5.6 years. Median end-diastolic pressure was 9 mm Hg (interquartile range, 8-13 mm Hg). Lateral E:e'/EDV (r = 0.59, P < .01), septal E:e'/EDV (r = 0.57, P < .01), and (E:circumferential early diastolic strain rate)/EDV (r = 0.54, P < .01) correlated with ß. Lateral E:e'/EDV displayed a C statistic of 0.93 in detecting patients with abnormal LV stiffness (ß > 0.015 mL-1). A lateral E:e'/EDV of >0.15 mL-1 had 89% sensitivity and 93% specificity in detecting an abnormal ß. CONCLUSIONS: Echocardiographic estimates of ventricular stiffness may be accurate compared with the gold standard in pediatric heart transplant recipients. The clinical usefulness of these noninvasive measures in assessing LV stiffness merits further study in children.

Toward computer-assisted image-guided congenital heart defect repair: an initial phantom analysis.

PURPOSE: Radiation exposure in interventional cardiology is an important consideration, due to risk of cancer and other morbidity to the patient and clinical staff. Cardiac catheterizations rely heavily on fluoroscopic imaging exposing both patient and clinician to ionizing radiation. An image-guided surgery system capable of facilitating cardiac catheterizations was developed and tested to evaluate dose reduction. METHODS: Several electromagnetically tracked tools were constructed specifically a 7-Fr catheter with five 5-degree-of-freedom magnetic seeds. Catheter guidance was accomplished using our image guidance system Kit for Navigation by Image-Focused Exploration and fluoroscopy alone. A cardiac phantom was designed and 3D printed to validate the image guidance procedure. In mock procedures, an expert clinician guided and deployed an occluder across the septal defect of the phantom heart. RESULTS: The image guidance method resulted in a dose of 1.26 mSv of radiation dose per procedure, while traditional guidance resulted in a dose of 3.33 mSv. Average overall dose savings for the image-guided method was nearly 2.07 mSv or 62 %. CONCLUSION: The work showed significant ([Formula: see text]) decrease in radiation dose with use of image guidance methods at the expense of a modest increase in procedure time. This study lays the groundwork for further exploration of image guidance applications in pediatric cardiology.

Speckle-Tracking Echocardiographic Measures of Right Ventricular Diastolic Function Correlate with Reference Standard Measures Before and After Preload Alteration in Children.

The accuracy of echocardiographic measures of right ventricular (RV) diastolic function has been sparsely studied. Our objective was to evaluate the correlation between echocardiographic and reference standard measures of RV diastolic function derived from micromanometer pressure analysis before and after preload alteration in children. Echocardiograms and micromanometer pressure analyses were prospectively performed before and after fluid bolus in children undergoing right heart catheterization. The isovolumic relaxation time constant (τ) and end-diastolic pressure (EDP) were measured. Conventional and speckle-tracking echocardiographic (STE) parameters of RV systolic and diastolic function were assessed. Normal saline bolus was given to increase RV EDP by 20 %. Twenty-eight studies were performed in 22 patients with congenital heart disease or postheart transplantation. Mean age was 8.7 ± 6.1 years. RV longitudinal early diastolic strain rate (EDSR) correlated with τ before (r = 0.57, p = 0.001) and after fluid bolus (r = 0.48, p = 0.008). No conventional echocardiographic measures correlated with τ both before and after fluid bolus. Multiple regression analysis revealed RV EDSR and LV circumferential EDSR as independent predictors of RV τ. There were no independent predictors of EDP. RV EDSR appears to correlate with the reference standard measure of early active ventricular relaxation in children at baseline and after changes in preload. Conventional echocardiographic measures of diastolic function were not predictive of diastolic function after preload alteration. Future studies should assess the prognostic significance of STE measures of diastolic function in this population.

Interpreting measurements of cardiac function using vendor-independent speckle tracking echocardiography in children: a prospective, blinded comparison with catheter-derived measurements.

BACKGROUND: Adult studies demonstrate that echocardiographic measurements of cardiac function using speckle tracking correlate with invasive measurements, but such data in the pediatric population are sparse. Our aim was to compare speckle-derived measures of cardiac function to measurements routinely obtained by cardiac catheterization in children. METHODS: Echocardiograms were performed on the day of cardiac catheterization. Using Tomtec 2D Cardiac Performance Analysis, longitudinal strain (LS), longitudinal strain rate (LSR), early diastolic LSR, and ejection fraction (EF) for the right and left ventricle (RV and LV) were calculated via speckle tracking. Global LS and LSR were calculated for the LV. These results were compared to cardiac index, maximum ventricular dp/dt (max dp/dt), ventricular end-diastolic pressure (EDP), and pulmonary capillary wedge pressure (PCWP) obtained by fluid-filled catheters. A blinded observer performed all echo measurements. RESULTS: Fifty studies were performed on 28 patients ages 4 months to 20 years old. Their diagnoses included cardiac transplant (48 studies), repaired AV septal defect (1), and dilated cardiomyopathy (1). RVEDP ranged from 2 to 22 mm Hg (median=6) and PCWP ranged from 6 to 32 mmHg (median 10). LV global LS and LV 2-chamber LSR by speckle-tracking negatively correlated with LV max dp/dt (LV global LS R=-.83, P=.001; LV 2-chamber LSR R=-.69, P=.009). RV LS weakly correlated with max dp/dt (R=.363, P=.002). Early diastolic strain rate did not correlate with EDP in either ventricle. CONCLUSION: Speckle-derived measurements of function in the LV have stronger correlation than the RV to catheter-derived measures. LV global LS has the strongest correlation with invasive function measures in children.

Validation of Noninvasive Measures of Left Ventricular Mechanics in Children: A Simultaneous Echocardiographic and Conductance Catheterization Study.

BACKGROUND: The accuracy of echocardiography in evaluating left ventricular contractility has not been validated in children. The objective of this study was to compare echocardiographic measures of contractility with those derived from pressure-volume loop (PVL) analysis in children. METHODS: Patients with relatively normal loading conditions undergoing routine left heart catheterization were prospectively enrolled. PVLs were obtained via conductance catheters. The gold-standard measure of contractility, end-systolic elastance (Ees), was obtained via balloon occlusion of one or both vena cavae. Echocardiograms were performed immediately after PVL analysis under the same anesthetic conditions. Single-beat estimations of echocardiographic Ees were calculated using four different methods. These estimates were calculated using a combination of noninvasive blood pressure readings, ventricular volumes derived from three-dimensional echocardiography, and Doppler time intervals. RESULTS: Of 24 patients, 18 patients were heart transplant recipients, and six patients had small patent ductus arteriosus or small coronary fistulae. The mean age was 9.1 ± 5.6 years. The average invasive Ees was 3.04 ± 1.65 mm Hg/mL. Invasive Ees correlated best with echocardiographic Ees by the method of Tanoue (r = 0.85, P < .01), with a mean difference of -0.07 mm Hg/mL (95% limits of agreement, -2.0 to 1.4 mm Hg/mL). CONCLUSIONS: Echocardiographic estimates of Ees correlate well with gold-standard measures obtained via conductance catheters in children with relatively normal loading conditions. The use of these noninvasive measures in accurately assessing left ventricular contractility appears promising and merits further study in children.

A multiscale model for the study of cardiac biomechanics in single-ventricle surgeries: a clinical case.

Complex congenital heart disease characterized by the underdevelopment of one ventricular chamber (single ventricle (SV) circulation) is normally treated with a three-stage surgical repair. This study aims at developing a multiscale computational framework able to couple a patient-specific three-dimensional finite-element model of the SV to a patient-specific lumped parameter (LP) model of the whole circulation, in a closed-loop fashion. A sequential approach was carried out: (i) cardiocirculatory parameters were estimated by using a fully LP model; (ii) ventricular material parameters and unloaded geometry were identified by means of the stand-alone, three-dimensional model of the SV; and (iii) the three-dimensional model of SV was coupled to the LP model of the circulation, thus closing the loop and creating a multiscale model. Once the patient-specific multiscale model was set using pre-operative clinical data, the virtual surgery was performed, and the post-operative conditions were simulated. This approach allows the analysis of local information on ventricular function as well as global parameters of the cardiovascular system. This methodology is generally applicable to patients suffering from SV disease for surgical planning at different stages of treatment. As an example, a clinical case from stage 1 to stage 2 is considered here.

Comparison of echocardiographic and pressure-volume loop indices of systolic function in patients with single ventricle physiology: a preliminary report.

BACKGROUND: Differences in ventricular geometry and physiology of patients with single ventricle anatomy complicate the application of traditional, noninvasive measurements of systolic function. We compared noninvasive measures of ventricular systolic function in single ventricle patients with invasive measures to evaluate their validity in this population. METHODS: A secondary analysis of patients with single ventricle physiology enrolled in the multi-institutional research project, "multi-scale modeling of single ventricle hearts," was performed. Pressure-volume loops (PVLs) were recorded using microconductance catheters. Transthoracic echocardiogram and cardiac magnetic resonance imaging were performed on the same day. PVL indices of systolic function including end-systolic elastance (Ees), maximal rate of pressure increase (dP/dTmax), and stroke work indexed to end-diastolic volume (SW/EDV) were compared with noninvasive measures, including echocardiographic myocardial performance index (MPI), rate of pressure rise (AV valve dP/dT), isovolumic acceleration, longitudinal shortening fraction (longSF), and fractional area change (FAC). RESULTS: Fifteen patients had PVLs available for analysis. Eleven had a dominant right ventricle, three were status poststage 1 repair, five had superior cavopulmonary anastomosis, and seven had a total cavopulmonary anastomosis. FAC correlated with Ees (r = 0.69, P < .01), SW/EDV (r = 0.64, P = .01), and dP/dTmax (r = 0.59, P = .03). LongSF correlated with dP/dTmax (r = 0.61, P = .02) MPI, AV valve dP/dT, and isovolumic acceleration did not correlate with pressure-volume loop indices of systolic function. CONCLUSIONS: Obtaining PVLs via microconductance catheters can reliably be performed in the single ventricle population and serve as a method to validate echocardiographic indices in this high-risk population. Of the echocardiographic variables, FAC showed the best correlation with PVL indices. Future studies controlling for stage of palliation should be performed to further validate echocardiographic measures of systolic function in this patient population.

Comparison of pressure-volume loop and echocardiographic measures of diastolic function in patients with a single-ventricle physiology.

Echocardiographic measurements of diastolic function have not been validated against invasive pressure-volume loop (PVL) analysis in the single-ventricle population. The authors hypothesized that echocardiographic measures of diastolic function would correlate with PVL indices of diastolic function in patients with a single-ventricle physiology. The conductance-derived PVL measures of diastolic function included the isovolumic relaxation time constant (τ), the maximum rate of ventricular pressure decline (peak -dP/dt), and a measure of passive diastolic stiffness (µ). The echocardiographic measures included Doppler inflow patterns of the dominant atrioventricular valve (DAVV), tissue Doppler velocities (TDI) at the lateral (ventricular free wall) component of the DAVV annulus, and the TDI-derived isovolumic relaxation time (IVRT'). The correlation between PVL and echocardiographic measures was examined. The study enrolled 13 patients at various stages of surgical palliation. The median age of the patients was 3 years (range 3 months to 19 years). τ correlated well with Doppler E:A (r = 0.832; p = 0.005), lateral E:E' (r = 0.747; p = 0.033), and IVRT' (r = 0.831; p = 0.001). Peak -dP/dt also was correlated with IVRT' (r = 0.609; p = 0.036), and µ also was correlated with IVRT' (r = 0.884; p = 0.001). This study represents the first-ever comparison of diastolic echocardiographic and PVL indices in a single-ventricle population. The findings show that Doppler E:A, lateral E:E', and IVRT' correlate well with PVL measures of diastolic function. This study supports further validation of echocardiographic measures of diastolic function versus PVL measures of diastolic function in the single-ventricle population.

Diagnostic accuracy of CT angiography in infants with tetralogy of Fallot with pulmonary atresia and major aortopulmonary collateral arteries.

BACKGROUND: CT angiography (CTA) is increasingly used for the evaluation of congenital heart disease. OBJECTIVE: The aim was to determine the diagnostic accuracy of CTA in infants with tetralogy of Fallot with pulmonary atresia (ToF-PA) and major aortopulmonary collateral arteries (MAPCAs). METHODS: We retrospectively evaluated 18 consecutive patients (7 girls; median age, 6 days; range, 1-334 days) with ToF-PA and MAPCAs. Findings on CTA were compared with diagnostic catheterization (n = 16) or intraoperative findings (n = 2) for the number of MAPCAs, their diameter, origin, and supplied lung lobes and for the presence and diameter of central pulmonary arteries. Spearman correlation coefficient was calculated to assess the correlation between diameter measurements on CTA and catheterization. CTA dose-length product and catheterization dose-area product were recorded, and effective radiation doses were calculated with the use of age-specific conversion factors. RESULTS: Agreement was found between CTA and catheterization or intraoperative findings for the number of MAPCAs, their origin, and supplied lung lobes in all cases. In 11 of 13 patients, CTA accurately demonstrated central pulmonary arteries. A good correlation was found between diameter measurements on CTA and catheterization for MAPCAs (r = 0.83) and central pulmonary arteries (r = 0.82). Median effective doses were 0.9 mSv for CTA and 14.4 mSv for catheterization (P < .001). CONCLUSION: CTA is accurate in the preoperative evaluation of infants with ToF-PA and MAPCAs and is associated with a substantially lower radiation dose than catheterization. Preoperative diagnostic catheterization, therefore, may only be necessary in select patients with small MAPCAs in whom the precise assessment of central pulmonary arteries is required for surgical planning.

Intervention for recoarctation in the single ventricle reconstruction trial: incidence, risk, and outcomes.

BACKGROUND: Recoarctation after the Norwood procedure increases risk for mortality. The Single Ventricle Reconstruction (SVR) trial randomized subjects with a single right ventricle undergoing a Norwood procedure to a modified Blalock-Taussig shunt or a right ventricle-pulmonary artery shunt. We sought to determine the incidence of recoarctation, risk factors, and outcomes in the SVR trial. METHODS AND RESULTS: Recoarctation was defined by intervention, either catheter based or surgical. Univariate analysis and multivariable Cox proportional hazard models were performed with adjustment for center. Of the 549 SVR subjects, 97 (18%) underwent 131 interventions (92 balloon aortoplasty, 39 surgical) for recoarctation at a median age of 4.9 months (range, 1.1-10.5 months). Intervention typically occurred at pre-stage II catheterization (n=71, 54%) or at stage II surgery (n=38, 29%). In multivariable analysis, recoarctation was associated with the shunt type in place at the end of the Norwood procedure (hazard ratio, 2.0 for right ventricle-pulmonary artery shunt versus modified Blalock-Taussig shunt; P=0.02), and Norwood discharge peak echo-Doppler arch gradient (hazard ratio, 1.07 per 1 mm Hg; P<0.01). Subjects with recoarctation demonstrated comorbidities at pre-stage II evaluation, including higher pulmonary arterial pressures (15.4±3.0 versus 14.5±3.5 mm Hg; P=0.05), higher pulmonary vascular resistance (2.6±1.6 versus 2.0±1.0 Wood units·m(2); P=0.04), and increased echocardiographic volumes (end-diastolic volume, 126±39 versus 112±33 mL/BSA(1.3), where BSA is body surface area; P=0.02). There was no difference in 12-month postrandomization transplantation-free survival between those with and without recoarctation (P=0.14). CONCLUSIONS: Recoarctation is common after Norwood and contributes to pre-stage II comorbidities. Although with intervention there is no associated increase in 1-year transplantation/mortality, further evaluation is warranted to evaluate the effects of associated morbidities.

Usefulness of live three-dimensional transesophageal echocardiography in a congenital heart disease center.

Three-dimensional (3D) transesophageal echocardiography (TEE) has numerous potential applications in the care of patients with congenital heart disease (CHD). However, there were few data examining its utility in this setting. The aim was to describe the initial experience and feasibility of this modality at a tertiary CHD center. Twenty-seven 3D-TEE studies using the x7-2t live 3D matrix TEE transducer with an iE33 echocardiographic scanner (Philips Medical Systems, Bothell, Washington) were attempted. The utility of this method was examined in the 3 settings of interventional catheterizations (n = 16), intraoperative studies (n = 4), and diagnostic studies (n = 7). The probe was successfully inserted in 26 of 27 patients (weight 22.6 to 110 kg). In conclusion, the current matrix-array 3D-TEE probe was found to have a wide variety of clinical applications in a CHD center.

Left ventricular dysfunction is associated with intraventricular dyssynchrony by 3-dimensional echocardiography in children.

INTRODUCTION: We used 3-dimensional (3D) echocardiography to identify and quantify left ventricular (LV) dyssynchrony in children with LV dysfunction compared with control subjects. METHODS: The 3D echocardiography LV full volumes were acquired in 18 children, 9 with LV dysfunction and 9 control subjects. The LV was subdivided into 16 segments (apex excluded). Time from end diastole to the minimal systolic volume for each segment was expressed as a percent of the R-R interval. The SD of these times provided a 16-segment dyssynchrony index (16-SDI). The second index (12-SDI) was similarly calculated using 6 basal and 6 mid segments. The third index consisted of 6 basal segments (6-SDI). RESULTS: The dysfunction group exhibited significantly increased 16-SDI (P = .008) and 12-SDI (P = .01). The 16-SDI was negatively correlated with 3D ejection fraction and 2-dimensional fractional shortening. CONCLUSIONS: Children with LV dysfunction demonstrate increased intraventricular LV dyssynchrony by 3D echocardiography, in a pattern that is negatively correlated with LV systolic function.

Transthoracic real-time three-dimensional echocardiography in the diagnosis and description of noncompaction of ventricular myocardium.

BACKGROUND: Numerous modalities have been used to diagnose and characterize noncompaction of ventricular myocardium (NCVM) including magnetic resonance imaging, two-dimensional echocardiography (2DE), contrast-enhanced 2DE, and angiography. The current case series examines the use of real-time three-dimensional transthoracic echocardiography (RT-3DE) in four such cases of NCVM. METHODS: From December 2003 to March 2004, we performed RT-3DE using a Philips Sonos 7500 echocardiographic scanner equipped with a 2-4 MHz 3D matrix array transthoracic probe, to evaluate four patients with NCVM. The real-time 3D transthoracic probe allows for dataset acquisition in an ultrasound wedge, which can be manipulated instantaneously. In addition, complete 3D volume rendering is acquired, allowing for volumetric imaging. RESULTS: The age range of the patients was 2 months to 42 years. One patient had the codiagnoses of coarctation of the aorta and bicuspid aortic valve. In all four patients, RT-3DE enabled diagnosis and provided detailed characterization of the affected myocardium. Entire trabecular projections and intertrabecular recesses were easily visualized simultaneously, and endocardial borders were clearly demarcated. Wall motion abnormalities of the affected myocardium were clearly visualized. The compacted and noncompacted portions of the myocardium could be differentiated well. CONCLUSIONS: Our study provides preliminary data highlighting the utility and feasibility of RT-3DE in the clinical characterization of NCVM. The complex 3D nature of this disorder and the endocardial hypertrabeculation were more readily visualized than with 2DE.