Utility of deep learning networks for the generation of artificial cardiac magnetic resonance images in congenital heart disease.

BACKGROUND: Deep learning algorithms are increasingly used for automatic medical imaging analysis and cardiac chamber segmentation. Especially in congenital heart disease, obtaining a sufficient number of training images and data anonymity issues remain of concern. METHODS: Progressive generative adversarial networks (PG-GAN) were trained on cardiac magnetic resonance imaging (MRI) frames from a nationwide prospective study to generate synthetic MRI frames. These synthetic frames were subsequently used to train segmentation networks (U-Net) and the quality of the synthetic training images, as well as the performance of the segmentation network was compared to U-Net-based solutions trained entirely on patient data. RESULTS: Cardiac MRI data from 303 patients with Tetralogy of Fallot were used for PG-GAN training. Using this model, we generated 100,000 synthetic images with a resolution of 256 × 256 pixels in 4-chamber and 2-chamber views. All synthetic samples were classified as anatomically plausible by human observers. The segmentation performance of the U-Net trained on data from 42 separate patients was statistically significantly better compared to the PG-GAN based training in an external dataset of 50 patients, however, the actual difference in segmentation quality was negligible (< 1% in absolute terms for all models). CONCLUSION: We demonstrate the utility of PG-GANs for generating large amounts of realistically looking cardiac MRI images even in rare cardiac conditions. The generated images are not subject to data anonymity and privacy concerns and can be shared freely between institutions. Training supervised deep learning segmentation networks on this synthetic data yielded similar results compared to direct training on original patient data.

Prediction of prognosis in patients with tetralogy of Fallot based on deep learning imaging analysis.

OBJECTIVE: To assess the utility of machine learning algorithms for automatically estimating prognosis in patients with repaired tetralogy of Fallot (ToF) using cardiac magnetic resonance (CMR). METHODS: We included 372 patients with ToF who had undergone CMR imaging as part of a nationwide prospective study. Cine loops were retrieved and subjected to automatic deep learning (DL)-based image analysis, trained on independent, local CMR data, to derive measures of cardiac dimensions and function. This information was combined with established clinical parameters and ECG markers of prognosis. RESULTS: Over a median follow-up period of 10 years, 23 patients experienced an endpoint of death/aborted cardiac arrest or documented ventricular tachycardia (defined as >3 documented consecutive ventricular beats). On univariate Cox analysis, various DL parameters, including right atrial median area (HR 1.11/cm², p=0.003) and right ventricular long-axis strain (HR 0.80/%, p=0.009) emerged as significant predictors of outcome. DL parameters were related to adverse outcome independently of left and right ventricular ejection fraction and peak oxygen uptake (p<0.05 for all). A composite score of enlarged right atrial area and depressed right ventricular longitudinal function identified a ToF subgroup at significantly increased risk of adverse outcome (HR 2.1/unit, p=0.007). CONCLUSIONS: We present data on the utility of machine learning algorithms trained on external imaging datasets to automatically estimate prognosis in patients with ToF. Due to the automated analysis process these two-dimensional-based algorithms may serve as surrogates for labour-intensive manually attained imaging parameters in patients with ToF.