Open-source, fully-automated hybrid cardiac substructure segmentation: development and optimisation

Abstract

Radiotherapy for thoracic and breast tumours is associated with a range of cardiotoxicities. Emerging evidence suggests cardiac substructure doses may be more predictive of specific outcomes, however, quantitative data necessary to develop clinical planning constraints is lacking. Retrospective analysis of patient data is required, which relies on accurate segmentation of cardiac substructures. In this study, a novel model was designed to deliver reliable, accurate, and anatomically consistent segmentation of 18 cardiac substructures on computed tomography (CT) scans. Thirty manually contoured CT scans were included. The proposed multi-stage method leverages deep learning (DL), multi-atlas mapping, and geometric modelling to automatically segment the whole heart, cardiac chambers, great vessels, heart valves, coronary arteries, and conduction nodes. Segmentation performance was evaluated using the Dice similarity coefficient (DSC), mean distance to agreement (MDA), Hausdorff distance (HD), and volume ratio. Performance was reliable, with no errors observed and acceptable variation in accuracy between cases, including in challenging cases with imaging artefacts and atypical patient anatomy. The median DSC range was 0.81–0.93 for whole heart and cardiac chambers, 0.43–0.76 for great vessels and conduction nodes, and 0.22–0.53 for heart valves. For all structures the median MDA was below 6 mm, median HD ranged 7.7–19.7 mm, and median volume ratio was close to one (0.95–1.49) for all structures except the left main coronary artery (2.07). The fully automatic algorithm takes between 9 and 23 min per case. The proposed fully-automatic method accurately delineates cardiac substructures on radiotherapy planning CT scans. Robust and anatomically consistent segmentations, particularly for smaller structures, represents a major advantage of the proposed segmentation approach. The open-source software will facilitate more precise evaluation of cardiac doses and risks from available clinical datasets. Graphical abstract

Predicting Cardiac Issues From Echocardiograms: A Literature Review Using Deep Learning and Machine Learning Techniques

“…This section examines the literature on efficient procedures and compares their performance to find out how they stack up against one another. For the cardiac image segmentation, BiSeNet [26], U-Net [30], FASTR-SCANN [32], U-Net-Transformer [33], OFHCSS [35] and U-Net-YOLOv7 [36] are considered for the evaluation. Then, CNN-ResNet [37], CNN-LSTM [41], Xception [45], 1D-CNN [46], CNN [47] and ShuffleNet [48] are taken for the cardiac view classification.…”

“…An open-source, fully-automated hybrid cardiac substructure segmentation system (OFHCSS) was developed by Finnegan et al [35]. An automated segmentation of the whole heart, its chambers, major vessels, valves, coronary arteries, and conduction nodes was achieved by the use of a multi-stage approach that combined DL with multi-atlas mapping and geometric modeling.…”

Heart and Lung Dose as Predictors of Overall Survival in Patients With Locally Advanced Lung Cancer. A National Multicenter Study

“…The heart and substructure were automatically delineated using a hybrid segmentation method for automatic segmentation, previously described by Finnegan et al. 14 , 23 The hybrid methods’ ability for dose prediction in a Danish setting has been evaluated previously. 23 The automatically segmented structures consisted of the heart, the four chambers, and the coronary arteries (left main coronary artery, LAD, circumflex artery [CX], and right coronary artery [RCA]) as seen in Figure 1 .…”

“…This problem can be addressed using automatic heart and substructures segmentation based on standardized guidelines. 14 , 15 …”

PlatiPy: Processing Library and Analysis Toolkit for Medical Imaging in Python

“…To date, two auto-segmentation models developed using PlatiPy have been deployed and made available through the library for direct use by researchers. The first is a cardiac sub-structure auto-segmentation model, which utilises a deep learning component to segment the whole heart, followed by an atlas-based segmentation and geometric definitions to segment 17 cardiac sub-structures on radiotherapy CT images (Finnegan et al, 2023). The second is a bronchial tree segmentation algorithm that employs threshold techniques to segment the lungs, followed by the airways in radiotherapy lung CT images (Ghandourh et al, 2021).…”