Fully‑automated deep‑learning segmentation of pediatric cardiovascular magnetic resonance of patients with complex congenital heart diseases

Abstract

Background For the growing patient population with congenital heart disease (CHD), improving clinical workflow, accuracy of diagnosis, and efficiency of analyses are considered unmet clinical needs. Cardiovascular magnetic resonance (CMR) imaging offers non-invasive and non-ionizing assessment of CHD patients. However, although CMR data facilitates reliable analysis of cardiac function and anatomy, clinical workflow mostly relies on manual analysis of CMR images, which is time consuming. Thus, an automated and accurate segmentation platform exclusively dedicated to pediatric CMR images can significantly improve the clinical workflow, as the present work aims to establish. Methods Training artificial intelligence (AI) algorithms for CMR analysis requires large annotated datasets, which are not readily available for pediatric subjects and particularly in CHD patients. To mitigate this issue, we devised a novel method that uses a generative adversarial network (GAN) to synthetically augment the training dataset via generating synthetic CMR images and their corresponding chamber segmentations. In addition, we trained and validated a deep fully convolutional network (FCN) on a dataset, consisting of $$64$$ 64 pediatric subjects with complex CHD, which we made publicly available. Dice metric, Jaccard index and Hausdorff distance as well as clinically-relevant volumetric indices are reported to assess and compare our platform with other algorithms including U-Net and cvi42, which is used in clinics. Results For congenital CMR dataset, our FCN model yields an average Dice metric of $$91.0\mathrm{\%}$$ 91.0 % and $$86.8\mathrm{\%}$$ 86.8 % for LV at end-diastole and end-systole, respectively, and $$84.7\mathrm{\%}$$ 84.7 % and $$80.6\mathrm{\%}$$ 80.6 % for RV at end-diastole and end-systole, respectively. Using the same dataset, the cvi42, resulted in $$73.2\mathrm{\%}$$ 73.2 % , $$71.0\mathrm{\%}$$ 71.0 % , $$54.3\mathrm{\%}$$ 54.3 % and $$53.7\mathrm{\%}$$ 53.7 % for LV and RV at end-diastole and end-systole, and the U-Net architecture resulted in $$87.4\mathrm{\%}$$ 87.4 % , $$83.9\mathrm{\%}$$ 83.9 % , $$81.8\mathrm{\%}$$ 81.8 % and $$74.8\mathrm{\%}$$ 74.8 % for LV and RV at end-diastole and end-systole, respectively. Conclusions The chambers’ segmentation results from our fully-automated method showed strong agreement with manual segmentation and no significant statistical difference was found by two independent statistical analyses. Whereas cvi42 and U-Net segmentation results failed to pass the t-test. Relying on these outcomes, it can be inferred that by taking advantage of GANs, our method is clinically relevant and can be used for pediatric and congenital CMR segmentation and analysis.

A narrative review of modern approach and outcomes evaluation in congenital heart defects

“…Over the past decade, our field has also seen significant advances in the quality of post-natal diagnosis. Technological advances in post-natal imaging has been dramatically improved by the introduction of threedimensional echocardiography (16,17), by accurate anatomic definition of congenital heart defects lesions through low radiation dose computed tomography (18)(19)(20), and by the functional assessment of flows and myocardial function with magnetic resonance imaging (21,22). Modern imaging techniques have also allowed for more precise clinical decision-making and planning related to the performance of interventional catheter-based procedures as well as for surgical repair through enhanced imaging techniques with segmentation of the heart structures to facilitate three-dimensional rendering or printing of complex heart anatomy (23,24).…”

The role of artificial intelligence in paediatric cardiovascular magnetic resonance imaging

“…The framework achieved good agreement with no clinical oversight, though some RV measures were less well performed by the framework than manual assessment. Importantly, deep learning ventricular segmentation of both the LV and RV has now been shown to be useable in CHD, though accuracy for RV measurements was worse than for LV measurements [21].…”

Artificial Intelligence in Cardiac MRI: Is Clinical Adoption Forthcoming?

“…The neuronal network performed equally or outperformed the human cardiac expert in all parts of left ventricle (LV) and right ventricle (RV) volumetry and mass measurements. Bidhendi et al ( 22 ) expanded the approach and created a fully convolutional network that was applied successfully in paediatric patients with CHD and proved to be superior to the algorithms clinically used in a commercially available platform. An extensive review on these techniques is presented in Chen et al ( 23 ).…”