Publicação

An efficient software tool to segment slice and view electron tomograms

Ver documento

Detalhes bibliográficos
Resumo:Segmentation is a key method to extract useful information in Electron Tomography. Manual segmentation is the most commonly used method, but it is subject to user bias and the process is slow. The lack of adequate automated processes, due to the high complexity and to the low signal-to-noise ratio of these tomograms, provided the main challenges for this dissertation: to develop a software tool to efficiently handle electron tomograms, including a novel 3D segmentation algorithm. Tomograms can be seen as a stack of 2D images; operations on tomograms usually lead to computationally intense tasks. This is due to the large amount of involved data and to the strided and random memory access patterns. These characteristics represent serious problems on novel computing systems, which rely on complex memory hierarchy architectures to hide memory access latency time. A software tool with a user-friendly interface — TomSeg — was designed, implemented and tested with experimental datasets, built with sequences of Scanning Electron Microscopy images obtained using a Slice and View technique. This tool lets users align, crop, segment and export electron tomograms, using computationally efficient processes. TomSeg takes advantage of the most usual architectures of modern compute servers, namely based on multicore and many-core CPU devices, exploring vector and parallel programming techniques; it also explores the available GPU-devices to speedup critical code functions. Validation and performance results on a compute server are presented together with the performance improvements obtained during the implementation and test phases. TomSeg is an open-source tool for Unix and Windows that can be easily extended with new algorithms to efficiently handle generic tomograms.
Autores principais:Sousa, Paulo Rafael da Costa e
Ano:2017
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade do Minho
Idioma:inglês
Origem:RepositóriUM - Universidade do Minho

Registos relacionados

Registos relacionados