The intricate nature of congenital heart disease requires understanding of the complex, patient-specific three-dimensional dynamic anatomy of the heart, from imaging data such as three-dimensional echocardiography for successful outcomes from surgical and interventional procedures. Conventional clinical systems use flat screens, and therefore, display remains two-dimensional, which undermines the full understanding of the three-dimensional dynamic data. Additionally, the control of three-dimensional visualisation with two-dimensional tools is often difficult, so used only by imaging specialists. In this paper, we describe a virtual reality system for immersive surgery planning using dynamic three-dimensional echocardiography, which enables fast prototyping for visualisation such as volume rendering, multiplanar reformatting, flow visualisation and advanced interaction such as three-dimensional cropping, windowing, measurement, haptic feedback, automatic image orientation and multiuser interactions. The available features were evaluated by imaging and nonimaging clinicians, showing that the virtual reality system can help improve the understanding and communication of three-dimensional echocardiography imaging and potentially benefit congenital heart disease treatment.

This study aimed to evaluate the accuracy and reliability of a virtual reality (VR) system line measurement tool using phantom data across three cardiac imaging modalities: three-dimensional echocardiography (3DE), computed tomography (CT) and magnetic resonance imaging (MRI). The same phantoms were also measured using industry-standard image visualisation software packages. Two participants performed blinded measurements on volume-rendered images of standard phantoms both in VR and on an industry-standard image visualisation platform. The intra- and interrater reliability of the VR measurement method was evaluated by intraclass correlation coefficient (ICC) and coefficient of variance (CV). Measurement accuracy was analysed using Bland–Altman and mean absolute percentage error (MAPE). VR measurements showed good intra- and interobserver reliability (ICC ≥ 0.99, p < 0.05; CV < 10%) across all imaging modalities. MAPE for VR measurements compared to ground truth were 1.6%, 1.6% and 7.7% in MRI, CT and 3DE datasets, respectively. Bland–Altman analysis demonstrated no systematic measurement bias in CT or MRI data in VR compared to ground truth. A small bias toward smaller measurements in 3DE data was seen in both VR (mean −0.52 mm [−0.16 to −0.88]) and the standard platform (mean −0.22 mm [−0.03 to −0.40]) when compared to ground truth. Limits of agreement for measurements across all modalities were similar in VR and standard software. This study has shown good measurement accuracy and reliability of VR in CT and MRI data with a higher MAPE for 3DE data. This may relate to the overall smaller measurement dimensions within the 3DE phantom. Further evaluation is required of all modalities for assessment of measurements <10 mm.

Work supported by the NIHR i4i funded 3D Heart project [II-LA-0716-20001] Background/Introduction Virtual Reality (VR) has recently gained great interest for examining 3D images from congenital heart disease (CHD) patients. Currently, 3D printed models of the heart may be used for particularly complex cases. These have been found to be intuitive and to positively impact clinical decision-making. Although positively received, such printed models must be segmented from the image data, generally only CT/MR may be used, the prints are static, and models do not allow for cropping / slicing or easy manipulation. Our VR system is designed to address these issues, as well as providing a simple interface compared to standard software. Building such a VR system, one with intuitive interaction which is clinically useful, requires studying user acceptance and requirements. can a prototype VR system be easily learned and used by clinicians unfamiliar with VR. We tested a VR system which can display 3D echo images and enables the user to interact with them, for instance by translating, rotating and cropping. Our system is tested on a transoesophageal echocardiogram from a patient with aortic valve disease. 13 clinicians evaluated the system including 5 imaging cardiologists, 5 physiologists, 2 surgeons and an interventionist, with their clinical experience ranging from trainee to more than 5 years’ of experience. None had used VR regularly in the past. After a brief training session, they were asked to place three anatomical landmarks and identify a particular cardiac view. They then completed a questionnaire on system ease of learning and image manipulation. Results: Results are shown in the figure below. Learning to use the system was perceived as easy for all but one participant, who rated it as ‘Somewhat difficult’. However, once trained, all users found the system easy to use. Participants found the interaction, where objects in the scene are picked up using the controller and then track the controller’s motion in a 1:1 way, to be particularly easy to learn and use. Our VR system was accepted by the vast majority of clinicians, both for ease of learning and use. Intuitiveness and the ability to interact with images in a natural way were highlighted as most useful - suggesting that such a system could become accepted for routine clinical use in the future. Abstract P1417 Figure. VR system evaluation participant feedbac Abstract P1417 Figure. VR system evaluation participant feedbac

Automatic segmentation of the placenta in fetal ultrasound (US) is challenging due to the (i) high diversity of placenta appearance, (ii) the restricted quality in US resulting in highly variable reference annotations, and (iii) the limited field-of-view of US prohibiting whole placenta assessment at late gestation. In this work, we address these three challenges with a multi-task learning approach that combines the classification of placental location (e.g., anterior, posterior) and semantic placenta segmentation in a single convolutional neural network. Through the classification task the model can learn from larger and more diverse datasets while improving the accuracy of the segmentation task in particular in limited training set conditions. With this approach we investigate the variability in annotations from multiple raters and show that our automatic segmentations (Dice of 0.86 for anterior and 0.83 for posterior placentas) achieve human-level performance as compared to intra- and inter-observer variability. Lastly, our approach can deliver whole placenta segmentation using a multi-view US acquisition pipeline consisting of three stages: multi-probe image acquisition, image fusion and image segmentation. This results in high quality segmentation of larger structures such as the placenta in US with reduced image artifacts which are beyond the field-of-view of single probes.