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

Abstract

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.

Keywords: Artificial intelligence; Congenital heart disease; Deep learning; Heart transplant; Imaging; Size matching.

Fig. 1

Deep learning architecture components. A DenseNet encoding section precedes the decoding section. B ResNet architecture maximized cross connections between convolutional layers preserving the gradients for small features throughout the pipeline. C Sigmoid activation of the final decoder block’s output is fed into a final convolutional layer for multi label object segmentation. TCV Total Cardiac Volume

Fig. 2

Example Total Cardiac Volume Segmentations (Blue = Ground Truth, Red = Deep Learning). CT Computed Tomography, GT-TCV Ground Truth Total Cardiac Volume, DL-TCV Deep Learning Total Cardiac Volume

Fig. 3

Correlation plot between Ground truth TCV (x axis) and Predicted TCV (y-axis) of the 44 subjects in the validation set. The subgroups are differentiated as follows: black circles (normal cardiac anatomy), red triangles (congenital heart disease), and blue diamonds (cardiomyopathy)

Fig. 4

Violin plot demonstrating the Dice similarity Coefficient (DSC) among subsets of subjects within the validation cohort. The average DSC is highest for the normal cohort, densely concentrated above 0.90. The cardiomyopathy (CM) and congenital heart disease (CHD) cohorts are more uniformly distributed with lower dice score than the normal cohort