CMR-Analysis (including machine learning)
Andrea Guala, PhD
Senior Researcher
Hospital Universitari Vall d'Hebron, CIBER-CV
barcelona, Catalonia, Spain
Lydia Dux-Santoy, PhD
Researcher
Hospital Universitari Vall d'Hebron, Catalonia, Spain
Juan Garrido-Oliver, MSc
Researcher, engineer
Hospital Universitari Vall d'Hebron, Spain
Jose F Rodriguez Palomares, MD, PhD, FSCMR
Cardiologist
Hospital Universitari Vall d'Hebron
Barcelona, Spain
Gisela Teixidó-Turà, MD, PhD
Cardiologist
Hospital Universitari Vall d'Hebron, Spain
Aroa Ruiz-Muñoz, MSc
Researcher
Hospital Universitari Vall d'Hebron, Spain
Guillem Casas, MD
MD
Hospital Universitari Vall d'Hebron, Spain
Filipa Valente, MD
Cardiologist
Hospital Universitari Vall d'Hebron
Barcelona, Spain
Laura Galian-Gay, MD, PhD
Cardiologist
Hospital Universitari Vall d'Hebron
Barcelona, Catalonia, Spain
Alejandro Carrasco-Poves, MSc
Predoctoral researcher
Hospital Universitari Vall d'Hebron, Spain
Ruben Fernandez-Galera, MD
Cardiologist
Hospital Universitari Vall d'Hebron, Spain
Hug Cuéllar, MD, PhD
Radiologist
Hospital Universitari Vall d'Hebron, Spain
Gemma Burcet, MD
Radiologist
Hospital Universitari Vall d'Hebron, Spain
Anton Aubanell, MD
Radiologist
Hospital Universitari Vall d'Hebron, Spain
Artur Evangelista, MD, PhD
Cardiologist
Hospital Universitari Vall d'Hebron, Spain
Ignacio Ferreira-Gonzáiez, MD, PhD
Cardiologist
Hospital Universitari Vall d'Hebron, Spain
Current assessment of aortic size and growth is based on manual measurements performed at specific locations identified by anatomical landmarks, often precluding the capacity to identify the location of fastest dilation rate (1). Three-dimensional aortic growth mapping (AGM) can be obtained via deformable registration of serial contrast-enhanced CT angiographies (1,2) and can give new insights into aortic dilation physiopathology (3). However, to limit radiation exposure, MR angiography (MRA) is preferred for patients’ follow-up, especially in young subjects. This work aimed to validated AGM applied to contrast-enhanced (CE-) and non-contrast, self-navigated 3D whole heart (SN-) MRA (4).
Methods:
Two independent, experienced observers measured thoracic aorta diameters (ascending and descending aorta, and diaphragm level) and their growth rates (GR) in 30 patients followed with CE-MRA and 15 followed with SN-MRA acquired at least 1 year apart, both manually and with the AGM technique. In patients followed with SN-MRA, aortic root (maximum, cusp-to-cusp, cusp-to-commissure) diameters and GRs were also obtained. Agreement between manual and AGM measurements and interobserver reproducibility were assessed. 3D maps of thoracic aortic diameters and GR by AGM were computed (FIGURE 1) and its reproducibility was assessed.
Results:
Follow-up duration was 3.8±2.3 years for patients followed with CE-MRA (46% male, age 38.7±18.8 years) and 2.7±1.6 years for patients with SN-MRA (40% male, age 27.2±14.5 years). Registration failed in two patients followed with CE-MRA. In the assessment of aortic diameters, the AGM technique presented low bias and excellent agreement with manual measurements both for CE- (bias=0.301 mm; ICC=0.984) and SN-MRA (bias=0.854 mm; ICC=0.968) and similar inter-observer reproducibility (ICC >0.9) (TABLE).
In aortic GR the AGM technique showed limited agreement with manual measurements using CE-MRA (bias=0.217 mm/y; ICC= 0.359) and moderate agreement using SN-MRA (bias=0.138 mm/y; ICC=0.538) (TABLE). Aortic GR by AGM presented much higher inter-observer reproducibility than manual measurements, both for CE- (ICC=0.837 vs 0.345) and SN-MRA (ICC=0.799 vs 0.595) (FIGURE 2). Manual GR showed moderate reproducibility using SN-MRA (ICC=0.595) but limited with CE-MRA (0.345), while AGM GR reproducibility was good both with CE and SN-MRA. Using SN-MRA, aortic root GR by AGM showed higher reproducibility than manual assessment (ICC=0.879 vs 0.519, respectively).
Regional thoracic aortic maps were highly reproducible, with overall excellent reproducibility for aortic diameters (mean ICC >0.9) and good reproducibility for GRs (mean ICC >0.85) both with CE-MRA and SN-MRA.
Conclusion:
Accurate and reproducible assessment of aortic size and growth rate via deformable registration of serial MRA is feasible both with CE-MRA and SN-MRA images, and allows to obtain robust 3D aortic growth mapping without the drawbacks of CT.