AI-based quantification of planned radiotherapy dose to cardiac structures and coronary arteries in breast cancer patients.

PURPOSE To develop and evaluate an automatic deep learning method for segmentation of cardiac chambers and large arteries, and localization of the three main coronary arteries in radiotherapy planning CT. To determine the planned radiotherapy dose to cardiac structures for breast cancer therapy. MATERIALS AND METHODS Eighteen contrast-enhanced cardiac scans acquired with a dual-layer-detector CT scanner were included for method development. Manual reference annotations of cardiac chambers, large arteries and coronary artery locations were made in the contrast scans and transferred to virtual non-contrast (VNC) images, mimicking non-contrast-enhanced CT. Additionally, 31 non-contrast-enhanced radiotherapy treatment planning CTs with corresponding dose-distribution maps of breast cancer patients were included for evaluation. For reference, cardiac chambers and large vessels were manually annotated in two 2D-slices per scan (26 scans, totaling 52 slices) and in 3D scan volumes in five scans. Coronary artery locations were annotated in 3D. The method employs an ensemble of convolutional neural networks with two output branches that perform two distinct tasks: segmentation of the cardiac chambers and large arteries, and localization of coronary arteries. Training was performed using reference annotations and VNC cardiac scans. Automatic segmentation of the cardiac chambers and large vessels, and coronary artery locations was evaluated in radiotherapy planning CT with Dice score (DSC) and average symmetric surface distance (ASSD). The correlation between dosimetric parameters derived from the automatic and reference segmentations was evaluated with R2. RESULTS For cardiac chambers and large arteries, median DSC were 0.76-0.88 and the median ASSD were 0.17-0.27 cm in 2D slice evaluation. 3D evaluation found a DSC of 0.87-0.93 and an ASSD of 0.07-0.10 cm. Median DSC of the coronary artery locations ranged from 0.80 to 0.91. R2 values of dosimetric parameters were 0.77-1.00 for the cardiac chambers and large vessels, and 0.76-0.95 for the coronary arteries. CONCLUSIONS The developed and evaluated method can automatically obtain accurate estimates of planned radiation dose and dosimetric parameters for the cardiac chambers, large arteries, and coronary arteries.