Expanding the donor pool for congenital heart disease transplant candidates by implementing 3D imaging-derived total cardiac volumes

Deep Learning for Automated Measurement of Total Cardiac Volume for Heart Transplantation Size Matching

Total Cardiac Volume (TCV)-based size matching using Computed Tomography (CT) is a novel technique to compare donor and recipient heart size in pediatric heart transplant that may increase overall utilization of available grafts. TCV requires manual segmentation, which limits its widespread use due to time and specialized software and training needed for segmentation. This study aims to determine the accuracy of a Deep Learning (DL) approach using 3-dimensional Convolutional Neural Networks (3D-CNN) to calculate TCV, with the clinical aim of enabling fast and accurate TCV use at all transplant centers. Ground truth TCV was segmented on CT scans of subjects aged 0-30 years, identified retrospectively. Ground truth segmentation masks were used to train and test a custom 3D-CNN model consisting of a DenseNet architecture in combination with residual blocks of ResNet architecture. The model was trained on a cohort of 270 subjects and a validation cohort of 44 subjects (36 normal, 8 heart disease retained for model testing). The average Dice similarity coefficient of the validation cohort was 0.94 ± 0.03 (range 0.84-0.97). The mean absolute percent error of TCV estimation was 5.5%. There is no significant association between model accuracy and subject age, weight, or height. DL-TCV was on average more accurate for normal hearts than those listed for transplant (mean absolute percent error 4.5 ± 3.9 vs. 10.5 ± 8.5, p = 0.08). A deep learning-based 3D-CNN model can provide accurate automatic measurement of TCV from CT images. This initial study is limited as a single-center study, though future multicenter studies may enable generalizable and more accurate TCV measurement by inclusion of more diverse cardiac pathology and increasing the training data.

Deep Learning for Automated Measurement of Total Cardiac Volume for Heart Transplantation Size Matching

Background

Total Cardiac Volume (TCV) based size matching using Computed Tomography (CT) is a novel technique to compare donor and recipient heart size in pediatric heart transplant that may increase overall utilization of available grafts. TCV requires manual segmentation, which limits its widespread use due to time and specialized software and training needed for segmentation.

Objective

This study aims to determine the accuracy of a Deep Learning (DL) approach using 3-dimensional Convolutional Neural Networks (3D-CNN) to calculate TCV, with the clinical aim of enabling fast and accurate TCV use at all transplant centers.

Materials and Methods

Ground truth TCV was segmented on CT scans of subjects aged 0-30 years, identified retrospectively. Ground truth segmentation masks were used to train and test a custom 3D-CNN model consisting of a Dense-Net architecture in combination with residual blocks of ResNet architecture.

Results

The model was trained on a cohort of 270 subjects and a validation cohort of 44 subjects (36 normal, 8 heart disease retained for model testing). The average Dice similarity coefficient of the validation cohort was 0.94 ± 0.03 (range 0.84-0.97). The mean absolute percent error of TCV estimation was 5.5%. There is no significant association between model accuracy and subject age, weight, or height. DL-TCV was on average more accurate for normal hearts than those listed for transplant (mean absolute percent error 4.5 ± 3.9 vs. 10.5 ± 8.5, p = 0.08).

Conclusion

A deep learning based 3D-CNN model can provide accurate automatic measurement of TCV from CT images.

Echocardiography Provides a Reliable Estimate of Total Cardiac Volume for Pediatric Heart Transplantation

BACKGROUND

Donor-to-recipient size matching for heart transplantation typically involves comparing donor and recipient body weight; however, weight is not linearly related to cardiac size. Attention has shifted toward the use of computed tomography- (CT-) derived total cardiac volume (TCV), that is, CT-TCV, to compare donor and recipient heart organ size. At this time, TCV size matching is near impossible for most centers due to logistical limitations. To overcome this impediment, echocardiogram-derived TCV (ECHO-TCV) is an attractive, alternative option to estimate CT-TCV. The goal of this study is to test whether ECHO-TCV is an accurate and reliable surrogate for TCV measurement compared with the gold standard CT-TCV.

METHODS

ECHO-TCV and CT-TCV were measured in a cohort spanning the neonatal to young adult age range with the intention to simulate the pediatric heart transplant donor pool. ECHO-TCV was measured using a modified Simpson's summation-of-discs method from the apical 4-chamber (A4C) view. The gold standard of CT-TCV was measured from CT scans using three-dimensional reconstruction software. The relationship between ECHO-TCV and CT-TCV was evaluated and compared with other anthropometric and image-based markers that may predict CT-TCV. Inter-rater reliability of ECHO-TCV was tested among 4 independent observers. Subanalyses were performed to identify imaging views and timing that enable greater accuracy of ECHO-TCV.

RESULTS

Banked imaging data of 136 subjects with both echocardiogram and CT were identified. ECHO-TCV demonstrated a linear relationship to CT-TCV with a Pearson correlation coefficient of r = 0.96 (95% CI, 0.95-0.97; P < .0001) and mean absolute percent error of 8.6%. ECHO-TCV correlated most strongly with CT-TCV in the subset of subjects <4 years of age (n = 33; r = 0.98; 95% CI, 0.96-0.99; P < .0001). The single-score intraclass correlation coefficient across all 4 raters is 0.96 (interquartile range, 0.93-0.98). ECHO-TCV measured from a standard A4C view at end diastole with the atria in the plane of view had the strongest correlation to CT-TCV.

CONCLUSIONS

ECHO-TCV by the A4C view was found to be both an accurate and reliable alternative measurement of CT-TCV and is derived from readily available donor ECHO images. The ECHO-TCV findings in this study make the ECHO method an attractive means of direct donor-to-recipient TCV size matching in pediatric heart transplantation.

Comparing donor and recipient total cardiac volume predicts risk of short-term adverse outcomes following heart transplantation

INTRODUCTION

In pediatric heart transplantation, donor: recipient weight ratio (DRWR) has long been the sole metric for size matching. Total cardiac volume (TCV)-based size matching has emerged as a novel method to precisely identify an upper limit of donor organ size of a heart transplant recipient while minimizing the risk of complications from oversizing. The clinical adoption of donor: recipient volume ratio (DRVR) to prevent short-term adverse outcomes of oversizing is unknown. The purpose of this single-center study is to determine the relationship of DRWR and DRVR to the risk of post-operative complications from allograft oversizing.

METHODS

Recipient TCV was measured from imaging studies and donor TCV was calculated from published TCV prediction models. DRVR was defined as donor TCV divided by recipient TCV. The primary outcome was short-term post-transplant complications (SPTC), a composite outcome of delayed chest closure and prolonged intubation > 7 days. A multivariable logistic regression model of DRWR (cubic spline), DRVR (linear) and linear interaction between DRWR and DRVR was used to examine the probability of experiencing a SPTC over follow-up as a function of DRWR and DRVR.

RESULTS

A total of 106 transplant patients' records were reviewed. Risk of the SPTC increased as DRVR increased. Both low and high DRWR was associated with the SPTC. A logistic regression model including DRWR and DRVR predicted SPTC with an AUROC curve of 0.74. [95% CI 0.62 0.85]. The predictive model identified a "low-risk zone" of donor-recipient size match between a weight ratio of 0.8 and 2.0 and a TCV ratio less than 1.0.

CONCLUSION

DRVR in combination with DRWR predicts short-term post-transplant adverse events. Accepting donors with high DRWR may be safely performed when DRVR is considered.

Pediatric heart transplantation: The past, the present, and the future

Heart transplantation (HTx) has a storied past, with origins dating back to the early twentieth century and the first pediatric orthotopic heart transplant performed in 1967 on a neonate with Ebstein abnormality. Today, approximately 500 pediatric HTx are performed annually, with survival times now measured in decades rather than days or weeks. In large part, advances in immunosuppression, critical care, dedicated transplant teams and mechanical circulatory support have paved the way for improvements in waitlist mortality and post-transplant survival, with future directions including the development of intracorporeal ventricular assist devices (VADs) for small children, expanding/standardizing donor criteria, and xenotransplantation.

Donor-Recipient Weight Match in Pediatric Heart Transplantation: Liberalizing Weight Matching with Caution

(1) Background: To expand the donor pool, greater donor hearts tended to be used in heart transplantation. However, the data about the feasibility of expanding the donor and recipient weight ratios (DRWRs. All donor and recipient weight ratio (DRWR) in this study or cited from other articles were converted to the DRWR calculated by ((donor weight-recipient weight)/recipient weight) × 100%.) to >30% was still scant in China’s pediatric heart transplantation (HTx). The potential risk increased along with the further expansion of the appropriate range of DRWR to >30% and its upper limit was still in debate. (2) Methods: Seventy-eight pediatric patients (age < 18 years) undergoing HTx between 2015 and 2020 at our center were divided into two groups based on the DRWR (>30% and ≤30%). Variables were summarized and analyzed via univariate analyses and multivariate analyses. A Kaplan-Meier methodology was used to calculate survival and conditional survival. (3) Results: No significant difference was found in one-year, three-year or five-year survival between the two groups. (4) Conclusions: The expansion of DRWR to >30% was acceptable for China’s pediatric HTx. Notably, continuously liberalizing of the upper DRWR boundary to more than 200% could be used as a stop-loss option but should be applied with caution.

Predictors and clinical significance of pericardial effusions after pediatric heart transplantation

BACKGROUND

We aimed to describe the incidence, risk factors, and clinical outcomes of pericardial effusions within 6 months after pediatric heart transplantation (HT).

METHODS

A single-center retrospective cohort study was performed on all pediatric HT recipients from 2004 to 2018. Logistic regression was used to identify factors associated with pericardial effusions post-HT, and survival was compared using log-rank test.

RESULTS

During the study period, 97 HTs were performed in 93 patients. Fifty patients (52%) had a ≥small pericardial effusion within 6 months, 16 of which were, or became, ≥moderate in size. Pericardial drain was placed in 8 patients. In univariate analysis, larger recipient body surface area (p = .01) and non-congenital heart disease (p = .002) were associated with pericardial effusion development. Donor/recipient size ratios, post-HT hemodynamics, and rejection did not correlate with pericardial effusion development. In multivariable analysis, non-congenital heart disease (adjusted odds ratio 3.3, p = .01) remained independently associated with development of pericardial effusion. There were no significant differences in post-HT survival between patients with and without ≥small (p = .68) or ≥moderate pericardial effusions (p = .40).

CONCLUSIONS

Pericardial effusions are common after pediatric HT. Patients with cardiomyopathy, or non-congenital heart disease, were at higher risk for post-HT pericardial effusions. Pericardial effusions increased morbidity but had no effect on mortality in our cohort. The risk factors identified may be used for anticipatory guidance in pediatric HT.

Evidence supporting total cardiac volumes instead of weight for transplant size-matching

Total cardiac volume (TCV)-based size matching for heart transplantation offers individualization in size matching that increases the number of suitable donors. Here we describe our clinical protocol for using TCV to determine an acceptable donor weight range for heart transplant candidates. We compare candidate imaging-derived TCV to a nomogram of subjects with normal TCV to determine a precise maximum donor weight at the time of listing. For nearly half of our transplant patients, we have increased weight range by an average of 70% with no oversizing related adverse events, such as delayed chest closure to avoid tamponade or bronchial compression. Widespread adoption of TCV-based size matching can lead to a more efficient heart allocation system by the data-driven bypass of poor size matches.

Reducing the wait: TCV can expand the donor pool for heart transplant candidates

A 16-year-old with new-onset dilated cardiomyopathy underwent VAD placement, later complicated by low flow from outflow graft kinking. To expedite heart transplantation, TCV was calculated and compared with 141 normal patients pinpointing the upper weight threshold. He was transplanted 2 days later within the expanded weight range with no post-transplant complications.

A novel method of donor‒recipient size matching in pediatric heart transplantation: A total cardiac volume‒predictive model

BACKGROUND

The pediatric heart transplant community uses weight-based donor-to-recipient size matching almost exclusively, despite no evidence to validate weight as a reliable surrogate of cardiac size. Donor size mismatch is the second most common reason for the refusal of donor hearts in current practice (∼30% of all refusals). Whereas case-by-case segmentation of total cardiac volume (TCV) by computed tomography (CT) for direct virtual transplantation is an attractive option, it remains limited by the unavailability of donor chest CT. We sought to establish a predictive model for donor TCV on the basis of anthropomorphic and chest X-ray (CXR) cardiac measures.

METHODS

Banked imaging studies from 141 subjects with normal CT chest angiograms were obtained and segmented using 3-dimensional modeling to derive TCV. CXR data were available for 62 of those subjects. A total of 3 predictive models of TCV were fit through multiple linear regression using the following variables: Model A (weight only); Model B (weight, height, sex, and age); Model C (weight, height, sex, age, and 1-view anteroposterior CXR maximal horizontal cardiac width).

RESULTS

Model C provided the most accurate prediction of TCV (optimism corrected R2 = 0.99, testing set R2 = 0.98, mean absolute percentage error [MAPE] = 8.6%) and outperformed Model A (optimism corrected R2 = 0.94, testing set R2 = 0.94, MAPE = 16.1%) and Model B (optimism corrected R2 = 0.97, testing set R2 = 0.97, MAPE = 11.1%).

CONCLUSIONS

TCV can be predicted accurately using readily available anthropometrics and a 1-view CXR from donor candidates. This simple and scalable method of TCV estimation may provide a reliable and consistent method to improve donor size matching.

A coordinated approach to improving pediatric heart transplant waitlist outcomes: A summary of the ACTION November 2019 waitlist outcomes committee meeting

The number of children needing heart transplantation continues to rise. Although improvements in heart failure therapy, particularly durable mechanical support, have reduced waitlist mortality, the number of children who die while waiting for a suitable donor organ remains unacceptably high. Roughly, 13% of children and 25% of infants on the heart transplant waitlist will not survive to transplantation. With this in mind, the Advanced Cardiac Therapies Improving Outcomes Collaborative Learning Network (ACTION), through its Waitlist Outcomes Committee, convened a 2-day symposium in Ann Arbor, Michigan, from 2-3 November 2019, to better understand the factors that contribute to pediatric heart transplant waitlist mortality and to focus future efforts on improving the organ allocation rates for children needing heart transplantation. Using improvement science methodology, the heart failure-transplant trajectory was broken down into six key steps, after which modes of failure and opportunities for improvement at each step were discussed. As a result, several projects aimed at reducing waitlist mortality were initiated.