Tomographic Reconstruction in the Presence of Internal Radiation Sources on GPU Architecture

This article presents the development of a novel implementation of algorithm for tomographic reconstruction that effectively addresses the challenges posed by internal radiation sources, utilizing GPU architecture for enhanced performance. The proposed algorithm leverages the parallel processing capabilities of GPUs to accelerate the reconstruction process, enabling the efficient handling of complex datasets and improving image quality. The peculiarity of this implementation is in finding independent rows of the projection matrix, which can be processed simultaneously using already known methods. By focusing on the unique characteristics of internal radiation sources, the algorithm aims to provide more accurate and reliable tomographic images, which are crucial for applications in medical imaging, industrial inspection, and geological studies. The paper also presents a computational experiment on the reconstruction of the object of study in the presence of multiple internal sources and receivers at the edges. The results demonstrate significant improvements in reconstruction speed and accuracy compared to traditional methods, highlighting the potential of GPU-based approaches in advancing tomographic techniques.