Two-dimensional (2D) ultrasound echocardiography is the primary technique used to diagnose congenital heart disease before birth. There is, however, a longstanding need for a reliable form of secondary imaging, particularly in cases when more detailed three-dimensional (3D) vascular imaging is required, or when ultrasound windows are of poor diagnostic quality. Fetal MRI, which is well established for other organ systems, is highly susceptible to fetal movement, particularly for 3D imaging. The objective of this study was to investigate the combination of prenatal MRI with novel, motion-corrected 3D image registration software, as an adjunct to fetal echocardiography in the diagnosis of congenital heart disease. Pregnant women carrying a fetus with known or suspected congenital heart disease were recruited via a tertiary fetal cardiology unit. After initial validation experiments to assess the general reliability of the approach, MRI data were acquired in 85 consecutive fetuses, as overlapping stacks of 2D images. These images were then processed with a bespoke open-source reconstruction algorithm to produce a super-resolution 3D volume of the fetal thorax. These datasets were assessed with measurement comparison with paired 2D ultrasound, structured anatomical assessment of the 2D and 3D data, and contemporaneous, archived clinical fetal MRI reports, which were compared with postnatal findings after delivery. Between Oct 8, 2015, and June 30, 2017, 101 patients were referred for MRI, of whom 85 were eligible and had fetal MRI. The mean gestational age at the time of MRI was 32 weeks (range 24-36). High-resolution (0·50-0·75 mm isotropic) 3D datasets of the fetal thorax were generated in all 85 cases. Vascular measurements showed good overall agreement with 2D echocardiography in 51 cases with paired data (intra-class correlation coefficient 0·78, 95% CI 0·68-0·84), with fetal vascular structures more effectively visualised with 3D MRI than with uncorrected 2D MRI (657 [97%] of 680 anatomical areas identified vs 358 [53%] of 680 areas; p<0·0001). When a structure of interest was visualised in both 2D and 3D data (n=358), observers gave a higher diagnostic quality score for 3D data in 321 (90%) of cases, with 37 (10%) scores tied with 2D data, and no lower scores than for 2D data (Wilcoxon signed rank test p<0·0001). Additional anatomical features were described in ten cases, of which all were confirmed postnatally. Standard fetal MRI with open-source image processing software is a reliable method of generating high-resolution 3D imaging of the fetal vasculature. The 3D volumes produced show good spatial agreement with ultrasound, and significantly improved visualisation and diagnostic quality compared with source 2D MRI data. This freely available combination requires minimal infrastructure, and provides safe, powerful, and highly complementary imaging of the fetal cardiovascular system. Wellcome Trust/EPSRC Centre for Medical Engineering, National Institute for Health Research.

Background: Identifying fetuses at risk of severe neonatal coarctation of the aorta (CoA) can be lifesaving but is notoriously challenging in clinical practice with a high rate of false positives. Novel fetal 3-dimensional and phase-contrast magnetic resonance imaging (MRI) offers an unprecedented means of assessing the human fetal cardiovascular system before birth. We performed detailed MRI assessment of fetal vascular morphology and flows in a cohort of fetuses with suspected CoA, correlated with the need for postnatal intervention. Methods: Women carrying a fetus with suspected CoA on echocardiography were referred for MRI assessment between 26 and 36 weeks of gestation, including high-resolution motion-corrected 3-dimensional volumes of the fetal heart and phase-contrast flow sequences gated with metric optimized gating. The relationship between aortic geometry and vascular flows was then analyzed and compared with postnatal outcome. Results: Seventy-two patients (51 with suspected fetal CoA and 21 healthy controls) underwent fetal MRI with motion-corrected 3-dimensional vascular reconstructions. Vascular flow measurements from phase-contrast sequences were available in 53 patients. In the CoA group, 25 of 51 (49%) required surgical repair of coarctation after birth; the remaining 26 of 51 (51%) were discharged without neonatal intervention. Reduced blood flow in the fetal ascending aorta and at the aortic isthmus was associated with increasing angulation ( P =0.005) and proximal displacement ( P =0.006) of the isthmus and was seen in both true positive and false positive cases. A multivariate logistic regression model including aortic flow and isthmal displacement explained 78% of the variation in outcome and correctly predicted the need for intervention in 93% of cases. Conclusions: Reduced blood flow though the left heart is associated with important configurational changes at the aortic isthmus in fetal life, predisposing to CoA when the arterial duct closes after birth. Novel fetal MRI techniques may have a role in both understanding and accurately predicting severe neonatal CoA.

Purpose: Development of a MRI acquisition and reconstruction strategy to depict fetal cardiac anatomy in the presence of maternal and fetal motion.Methods: The proposed strategy involves i) acquisition and reconstruction of highly accelerated dynamic MRI, followed by image‐based ii) cardiac synchronization, iii) motion correction, iv) outlier rejection, and finally v) cardiac cine reconstruction. Postprocessing entirely was automated, aside from a user‐defined region of interest delineating the fetal heart. The method was evaluated in 30 mid‐ to late gestational age singleton pregnancies scanned without maternal breath‐hold.Results: The combination of complementary acquisition/reconstruction and correction/rejection steps in the pipeline served to improve the quality of the reconstructed 2D cine images, resulting in increased visibility of small, dynamic anatomical features. Artifact‐free cine images successfully were produced in 36 of 39 acquired data sets; prolonged general fetal movements precluded processing of the remaining three data sets.Conclusions: The proposed method shows promise as a motion‐tolerant framework to enable further detail in MRI studies of the fetal heart and great vessels. Processing data in image‐space allowed for spatial and temporal operations to be applied to the fetal heart in isolation, separate from extraneous changes elsewhere in the field of view. Magn Reson Med 79:327–338, 2018. © 2017 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Objectives: Fetal cardiovascular magnetic resonance imaging (MRI) offers a potential alternative to echocardiography, although in practice, its use has been limited. We sought to explore the need for additional imaging in a tertiary fetal cardiology unit and the usefulness of standard MRI sequences.Methods: Cases where the diagnosis was not fully resolved using echocardiography were referred for MRI. Following a three‐plane localiser, fetal movement was assessed with a balanced steady‐state free precession (bSSFP) cine. Single‐shot fast spin echo and bSSFP sequences were used for diagnostic imaging.Results: Twenty‐two fetal cardiac MRIs were performed over 12 months, at mean gestation of 32 weeks (26–38 weeks). The majority of referrals were for suspected vascular abnormalities (17/22), particularly involving the aortic arch (n = 10) and pulmonary vessels (n = 4). Single‐shot fast spin echo sequences produced ‘black‐blood’ images, useful for examining the extracardiac vasculature in these cases. BSSFP sequences were more useful for intracardiac structures. Real‐time SSFP allowed for dynamic assessment of structures such as cardiac masses, with enhancement patterns also allowing for tissue characterisation in these cases.Conclusions: Fetal vascular abnormalities such as coarctation can be difficult to diagnose by using ultrasound. Fetal MRI may have an adjunctive role in the evaluation of the extracardiac vascular anatomy and tissue characterisation. © 2016 The Authors. Prenatal Diagnosis published by John Wiley & Sons, Ltd.

In this paper, we present a novel method for the correction of motion artifacts that are present in fetal magnetic resonance imaging (MRI) scans of the whole uterus. Contrary to current slice-to-volume registration (SVR) methods, requiring an inflexible anatomical enclosure of a single investigated organ, the proposed patch-to-volume reconstruction (PVR) approach is able to reconstruct a large field of view of non-rigidly deforming structures. It relaxes rigid motion assumptions by introducing a specific amount of redundant information that is exploited with parallelized patchwise optimization, super-resolution, and automatic outlier rejection. We further describe and provide an efficient parallel implementation of PVR allowing its execution within reasonable time on commercially available graphics processing units, enabling its use in the clinical practice. We evaluate PVR's computational overhead compared with standard methods and observe improved reconstruction accuracy in the presence of affine motion artifacts compared with conventional SVR in synthetic experiments. Furthermore, we have evaluated our method qualitatively and quantitatively on real fetal MRI data subject to maternal breathing and sudden fetal movements. We evaluate peak-signal-to-noise ratio, structural similarity index, and cross correlation with respect to the originally acquired data and provide a method for visual inspection of reconstruction uncertainty. We further evaluate the distance error for selected anatomical landmarks in the fetal head, as well as calculating the mean and maximum displacements resulting from automatic non-rigid registration to a motion-free ground truth image. These experiments demonstrate a successful application of PVR motion compensation to the whole fetal body, uterus, and placenta.

Deep neural networks exhibit limited generalizability across images with different entangled domain features and categorical features. Learning generalizable features that can form universal categorical decision boundaries across domains is an interesting and difficult challenge. This problem occurs frequently in medical imaging applications when attempts are made to deploy and improve deep learning models across different image acquisition devices, across acquisition parameters or if some classes are unavailable in new training databases. To address this problem, we propose Mutual Information-based Disentangled Neural Networks (MIDNet), which extract generalizable categorical features to transfer knowledge to unseen categories in a target domain. The proposed MIDNet adopts a semi-supervised learning paradigm to alleviate the dependency on labeled data. This is important for real-world applications where data annotation is time-consuming, costly and requires training and expertise. We extensively evaluate the proposed method on fetal ultrasound datasets for two different image classification tasks where domain features are respectively defined by shadow artifacts and image acquisition devices. Experimental results show that the proposed method outperforms the state-of-the-art on the classification of unseen categories in a target domain with sparsely labeled training data.