Thesis Details
Acceleration of Photoacoustic Imaging
The goal of this thesis is to provide a new method of image reconstruction out of data generated using Photo-Acoustic imaging. Photo-Acoustic imaging is a very popular biomedical in-vivo imaging modality based on the non-invasive laser-induced generation of ultrasound waves recorded by the acoustic sensors, during which very large amounts of data are generated. The amount of data makes the image reconstruction process very time-consuming. This thesis demonstrates image reconstruction using Back-Projection, an algorithm that is simple enough to be optimized for execution on modern accelerated processor architectures. Two versions of this algorithm are designed: from the perspective of the pixel and from the perspective of the sensor. Both versions are implemented using 3 different execution acceleration methods: vector-level parallelism, thread-level parallelism, and parallelism on the Graphical Processing Unit (GPU). All 3 implementations of both algorithm versions are tested and their results are compared to the much slower but more accurate Time-Reversal reconstruction method. The results have shown that the GPU parallelism implementation offers the fastest execution, which is faster more than 200 times on average compared to the Time-Reversal method. This possibly makes it suitable even for real-time applications.
Photo-Acoustic Imaging, Back-Projection, GPU, CUDA, OpenMP, Vectorization, Multi-threading
Čadík Martin, doc. Ing., Ph.D. (DCGM FIT BUT), člen
Lengál Ondřej, Ing., Ph.D. (DITS FIT BUT), člen
Malinka Kamil, Mgr., Ph.D. (DITS FIT BUT), člen
Polčák Libor, Ing., Ph.D. (DIFS FIT BUT), člen
@mastersthesis{FITMT21922, author = "Sava Nedeljkovi\'{c}", type = "Master's thesis", title = "Acceleration of Photoacoustic Imaging", school = "Brno University of Technology, Faculty of Information Technology", year = 2020, location = "Brno, CZ", language = "english", url = "https://www.fit.vut.cz/study/thesis/21922/" }