Background 4D flow MRI allows the analysis of hemodynamic changes in the aorta caused by pathologies such as thoracic aortic aneurysms (TAA). For personalized management of TAA, new biomarkers are required to analyze the effect of fluid structure iteration which can be obtained from 4D flow MRI. However, the generation of these biomarkers requires prior 4D segmentation of the aorta.
Objective To develop an automatic deep learning model to segment the aorta in 4D from 4D flow MRI.
Methods Segmentation is addressed with a U-Net based segmentation model that treats each 4D flow MRI frame as an independent sample. Performance is measured with respect to Dice score (DS) and Hausdorff distance (HD). In addition, the maximum and minimum surface areas at the level of the ascending aorta are measured and compared with those obtained from cine-MRI.
Results The segmentation performance was 0.90 ± 0.02 for the DS and the mean HD was 9.58 ± 4.36 mm. A correlation coefficient of r = 0.85 was obtained for the maximum surface and r = 0.86 for the minimum surface between the 4D flow MRI and cine-MRI.
Conclusion The proposed automatic approach of 4D aortic segmentation from 4D flow MRI seems to be accurate enough to contribute to the wider use of this imaging technique in the analysis of pathologies such as TAA.
An aortic aneurysm is an increase in diameter greater than or equal to 50% of its expected size . As a consequence of this pathology, the aortic wall can rupture or dissect, causing lethal consequences. The overall incidence per year of thoracic aortic aneurysms (TAAs) is 5 to 10 per 100,000 people, becoming the 19th leading cause of death overall . In clinical practice, the decision to intervene surgically on an aneurysm is mainly taken considering its diameter and growth rate. However, it has been observed that rupture and dissection can occur in aneurysms smaller than the sizes indicated in the guidelines [3,4]. Therefore, the generation of new biomarkers (including aorta shape, movement and constrain flow) that allow a personalized treatment of TAA is essential. In this context, 4D flow MRI is a cornerstone for assessing these parameters and opens the door to a new way of analyzing by considering the local and global hemodynamic characteristics and particularly the changes produced by cardiovascular pathologies such as TAA.
Despite the potential of 4D flow MRI for a widespread use in clinical practice for flow analysis and biomarker computation, it is necessary to first overcome some challenges such as automatic segmentation of the aortic wall. In the literature, in most studies, only 3D segmentation is proposed on 3D images generated from 4D flow MRI. The main objective of generating 3D images is to enhance the contrast between the aorta and the background. Kohler et al.  generated a maximum intensity projection image (tMIP) using the time steps of the magnitude image. Then, they ran a graph cut-based algorithm manually initialized by a user. Similarly, other studies have applied pre-processing techniques to create a 3D phase contrast magnetic resonance angiography (PCMRA) image from 4D flow MRI [6–8]. The drawback with PCMRA generation is that the temporal information is lost, which leads to a bias in the position of the aorta. Using PCMRA, active surface , or 3D neural networks [7,8] based segmentation algorithms have been implemented. Although these methods have shown high performance on PCMRA images, generating the segmentation for the whole cardiac cycle with comparable results is still an open problem.
The method evaluated in this paper for 4D segmentation of dilated aorta from 4D flow MRI magnitude images showed results comparable to those obtained in recent works for 4D segmentation of subjects without this pathology. Results shows that the 3D model adapts to the various shapes of the aorta during the cardiac cycle. Thus, the proposed approach of 4D aortic segmentation from 4D flow MRI could contribute to the expanded use of 4D flow MRI in the analysis of pathologies such as TAA. The results encourage further exploration of biomarkers correlated with aortic dilatation, such as wall share stress that, due to the challenge in 4D segmentation, have currently been evaluated assuming a static aorta.