Author: Ciprian N. Ionita, Parisa Naghdi, Ahmad Rahmatpour 👨🔬
Affiliation: University at Buffalo, SUNY: University at Buffalo 🌍
Purpose: Quantitative angiography (QA) evaluates neurovascular hemodynamics, but is affected by X-ray parameters, vessel foreshortening, and imaging chain factors like detector efficiency, pixel size, etc. To study these effects and the link between QA and hemodynamics, we propose a method to create digital twin angiograms with known hemodynamics.
Methods: Using clinical 3D DSA images, we extracted high-resolution 3D geometry of aneurysms. These geometries were used to create 4D angiography simulations within a computation fluid dynamic (CFD) solver (ANSYS). To consider radiation interactions with these 4D angiograms, we simulated the details of the imaging system using GATE Monte Carlo (MC) simulation toolkit. Using the CFD output as a voxelized phantom, radiation interactions which may happen in DSA were simulated to create new DSA projections with realistic physical considerations. We implemented MC simulations for CFD volumes at several time points, to see the contrast movement through the vessel. Furthermore, we performed the simulation for different parameters such as energy ranges, photons per second, detector voxel dimension, etc. The quality of simulated projections was assessed by using signal-to-noise ratio (SNR) metric.
Results: Using more simulated photons, we identified a meaningful improvement in SNR, from 3.2 to 6.85, 9.85, and 11, which are proportional to the square root of the number of primary particles. In contrast, in constant radiation parameters, SNR decreased to half when we reduced the image acquisition time four times. We applied both methods to ensure the simulation mirrors well understood physical principles. Different kilovoltage was applied and we observed improvement in SNR by factor 2 when we change the tube voltage from 40 KVp to 80 kVp.
Conclusion: We present a workflow for creating a digital DSA database, using CFD and MC simulations. This approach creates in-silico DSA images with radiation interaction consideration, supporting QA research optimization and clinical applicability.