Johnsen, SF; Clark, C; Atkinson, D; (2011) Registration of low-SNR high-resolution diffusion-weighted images. In: (Proceedings) Medical Image Analysis and Understanding MIUA 2011.

Text Johnsen_et_al_Registraiton_of_Low-SNR_2011.pdf Download (619kB)

Abstract

This paper introduces a novel, high-speed scheme for intrasubject registration and segmentation of high-resolution multi-shot diffusion-weighted images. Compared to single-shot sequences, multi-shot have advantages in terms of improved spatial resolution and reduced eddy-current and susceptibility artifacts. However, these sequences have prolonged scan times increasing the risk of subject motion, and, a lower signal to noise ratio (SNR) with smaller voxel volumes. The proposed registration algorithm comprises a hybrid thresholding expectation-maximization segmentation method that can cope with the low-SNR, and registers diffusion-weighted to B0 images through fast detection and matching of features found in edge images derived from floating and reference images. We performed validations of the entire pipeline, including assessment of visual appearance by experts, consistency error computations, and analysis of the segmentation, using volunteer images, and found its performance to be comparable with, or exceeding, that of established solutions.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item