Biomedical Engineering Reference
In-Depth Information
of the patient-specific aorta anatomy (Sect.
2.2
), (3) estimation of the patient-specific
aorta flow (Sect.
2.3
), (4) CFD simulation.
2.1 Measurements of Anatomy and Flow
The CFD simulations within our framework are driven by patient-specific
measurements coming from MR images. We would like to emphasize that we
directly process raw DICOM images produced by MR scanner of all major vendors.
From a typical study that contains on average 1,500 2D images we utilize two
sequences: (1) a stack of contrast enhanced MR angiography (CE-MRA) slices to
reconstruct the 3D anatomy of the aorta (2) a 2D PC -MRI Cine sequence that
contains through-plane flow velocity information distal to the AV and the region of
the descending aorta, distal to the supra-aortic arteries. The reconstruction and
image selection is accomplished within approximately 3 min (Fig.
1
a).
The 3D CE-MRA images are coronal stacks of dimension 256
256 to 512
640 with 56-140 slices, in-plane resolution isotropic 0.605-1.562 mm, slice
thickness of 0.889-1.8 mm. The coronal images usually contain the left ventricular
outflow tract (LVOT), the AV, the aortic arch, including the 3 superior arteries, and
the descending thoractic aorta. Due to the length of the acquisition, the angiograms
correspond to the aortic lumen motion averaged over the duration of the scan.
Flow-sensitive ECG-gated Cine PC MR images are typically oblique axial time-
series
encoding through-plane velocities
in the
isotropic
resolution of
0.742-2.083 mm, dimension 126
512, VENC found in the range
of 140-300 cm/s. The slices are routinely positioned to provide two different cross-
sections, one somewhere between the LVOT and AAo, the other in the DAo
(if there is a stent implant, in the direction of the blood flow below the stent
location). Each time series corresponds to one heart cycle, and has 20-40 frames
per cycle. The heart rates of the examined patients are ranging from 60 to 75 bpm.
All patient data was acquired employing 1.5 Tesla scanners as part of existing
clinical practice, and not specifically for this simulation study.
144 to 384
2.2 Estimation of the Patient-Specific Aorta Anatomy
2.2.1 Aorta Centerline Extraction
Within this step, the centerline of the aorta is automatically extracted from the 3D
anatomy image. Two seed points in the lumen of the aorta are required to initialize the
algorithm, with the source seed point placed in the left ventricle outflow tract
(LVOT) region and the sink seed point in the descending thoracic aorta. The
automatic extraction is performed by means of a robust vessel tree modeling algo-
rithm similarly as in [
12
]. The graph-based optimization algorithm estimates vessel
