Automatic determination of aortic compliance based on MRI and adapted curvilinear detector. (February 2018)

• Record Type: Journal Article
• Authors: Mitéran, J.; Bouchot, O.; Cochet, A.; Lalande, A.
• Is Part Of: Biomedical signal processing and control. Volume 40(2018)
• Page Start: 295
• Page End: 311
• Publication Date: 2018-02
• Subjects: Aortic compliance -- MRI -- Curvilinear detector; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2017.09.002
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Automatic aorta parameters computation from cine-MRI (compliance, distensibility). Adaptation of a curvilinear detector to near circular regions. Study of robustness against noise using synthetic data is performed. Comparisons between manual segmentation and previously published method are provided, for 40 real cases. Proposition of a fully automatic tool which can be applied in clinical practice. Abstract: Parameters of aortic elasticity, such as aortic compliance or aortic distensibility, can be estimated from cine-MRI through the knowledge of the aortic contour on each image. In this context, a completely automatic method for the measurement of aortic elasticity is proposed in this study, and compared with previously published methods which are not fully automatic. An adaptation of a curvilinear region detector was used for the aortic wall detection over the entire cardiac cycle, to completely automatically evaluate the aortic stiffness in a pilot study including 40 volunteers. Near circular regions were detected (ascending and descending aorta cross-sections) in each image of the sequence using robust scale-space based method with removing of false positives using probabilistic approach. Robustness against noise was studied and an evaluation of area estimation was performed. A comparison between manual segmentation by two experts is provided on whole images from the patient dataset. The global mean relative errors for the area are 2.83 ± 1.88% and 1.44 ± 1.52% for the ascending and descending aorta, respectively. The global means of the Dice's coefficient are 0.97 ± 0.01 for the ascending aorta and 0.97 ± 0.01 for the descending aorta. These values are high and very stable. Finally, the Bland–Altman plots for compliance and distensibility values show good agreements between our method and experts, with a mean of difference always close to zero, and a low standard deviation. Then the proposed tool allows a precise and accurate automatic measurement of aortic stiffness from cine-MRI and can be applied in clinical practice.