Keywords:
Interventional vascular, Computer applications, Cardiovascular system, Experimental, Computer Applications-3D, Experimental investigations, Angioscopy, Image registration
Authors:
J. Salamon, D. Weller, C. Jung, A. Frölich, M. Möddel, G. Adam, T. Knopp, H. Ittrich, R. Werner; Hamburg/DE
DOI:
10.26044/ecr2019/C-0664
Methods and materials
Selection of case scenarios and development of vascular models
Three typical interventional procedures were selected: Detection of an aneurysm,
thrombectomy and TACE.
Relevant vascular territories (internal carotid artery with aneurysm,
middle cerebral artery and proper hepatic artery and branches) were segmented from flat panel CT or MDCT with resolutions (=/< 1mm) using 3D Slicer.5
Final 3D CAD models were manufactured by a stereo lithography laser printer (Form2,
Formlabs) with a resolution of 50µm.
Image acquisition
Models were connected to a flow system,
using a water pump,
and tank as well as a clinical catheter system.
The flow rates ranged from 160 to 200 ml/min,
which is in the range of physiological blood flow rates in the respective vesselareas.
Vascular models were mounted on the robotic arm system of the MPI and automatically positioned within the scanners FoV.
A 1.5 ml tracer bolus of pure Resovist (108 mg/ml) was manually injected.
Imaging was performed using a preclinical MPI scanner system (Bruker Biospin GmbH,
Ettlingen,
Germany).
A selection gradient field of 1.5 T/m and a drive field amplitude of 14 mT resulted in a measurement field size of 40×40×20 mm3 (length × width × height).
Image reconstruction was performed on the basis of a system function approach.
Image Processing
All data was processed using the online reconstruction framework developed by Knopp et al.6 Visualization of the reconstructed data was performed using the software system VTK,
a well-established C++ toolkit for processing and visualization of medical image data.7 A schematic representation of the visualization pipeline used in this work is displayed in Fig. 1 .
This approach is based on the assumption that Volume rendering is generally superior to surface-based rendering techniques.8 In order to reconstruct comparative two-dimensional images of the investigated vascular anatomies,
conventional maximum-intensity projections (MIP) were derived from the corresponding clinical datasets.
Optimized camera positions were computed based on the generated centerlines and the position of a virtual guide wire tip.
Possible camera positions were sampled along a 360° rotation trajectory around the virtual position of guide wire tip.
The quantitative assessment of a potential view was based on the projection of the vessel section of interest onto a unit sphere centered at the tested camera position.
To determine the optimal viewing angle for the representation of a centerline segment in its maximum extent,
the projected total length of the corresponding centerline segment was maximized.
The optimized view of a vessel bifurcation,
on the other hand,
was derived by maximizing the area of a virtual triangle whose vertices were the branching point itself and one point each on the successor branches.