The next seminar “Modelling of materials – theory, model reduction and efficient numerical methods” will take place this Wednesday (Nov 10, 2021) from 9:00 till 10:00 in K3. The talk will be given by Vojtěch Kulvait through Zoom (using projector in K3). Please see the details below.
Speaker: Vojtěch Kulvait
Title: Developing fast tools for CT reconstruction, time resolved tomography and processing C-arm CT perfusion data
Abstract: Perfusion imaging is a diagnostic method for assessing cerebral tissue blood flow in patients with ischemic stroke with large vessel occlusion. It is important modality for the decision on mechanic thrombectomy in a patients after 6h from symptoms onset. It is typically performed using CT or MRI, where the patient is injected contrast material and the contrast agent dynamics is evaluated over a period of circa one minute. The idea of using C-arm CT for perfusion imaging instead of conventional CT is based on the fact that this device is available in the operating room where thrombectomy is performed. However, when using C-arm CT, we encounter problems due to the higher noise level, lower frame rate and the slow rotation of the device, so that during reconstruction we cannot consider the reconstructed object being static due to the dynamics of the contrast agent. In this talk I will present methods and software solutions for processing perfusion maps from C-arm CT perfusion so that the results are comparable with CT perfusion results. These include dimension reduction of the projection data by means of the Time separation technique, algebraic reconstruction of the fitted data and deconvolution processing to obtain resulting perfusion maps. Further I will talk about the optimization of the algebraic reconstruction and the CT projection operator by using parallel GPU programming in OpenCL to evaluate CT projection and backprojection (adjoint to projection) operators. Using these optimizations, it is possible to reduce the processing of typical C-arm perfusion scan so that it can be performed in a diagnostically acceptable times of circa 5min and provide accurate information of the perfusion status of the brain.