Felix Sebastian Leo Thomsen1, Claudio Augusto Delrieux2, Rodrigo de Luis-García3. 1. CONICET and Departamento de Ingeniería Eléctrica y Computadoras, Universidad Nacional del Sur, Avenida Alem 1253, 8000, Bahía Blanca, Buenos Aires, Argentina. Felix.Thomsen@uns.edu.ar. 2. CONICET and Departamento de Ingeniería Eléctrica y Computadoras, Universidad Nacional del Sur, Avenida Alem 1253, 8000, Bahía Blanca, Buenos Aires, Argentina. 3. ETSI Telecomunicación, Universidad de Valladolid, Campus Miguel Delibes, s/n, 47011, Valladolid, Spain.
Abstract
PURPOSE: Descriptors extracted from magnetic resonance imaging (MRI) of the brain can be employed to locate and characterize a wide range of pathologies. Scalar measures are typically derived within a single-voxel unit, but neighborhood-based texture measures can also be applied. In this work, we propose a new set of descriptors to compute local texture characteristics from scalar measures of diffusion tensor imaging (DTI), such as mean and radial diffusivity, and fractional anisotropy. METHODS: We employ weighted rotational invariant local operators, namely standard deviation, inter-quartile range, coefficient of variation, quartile coefficient of variation and skewness. Sensitivity and specificity of those texture descriptors were analyzed with tract-based spatial statistics of the white matter on a diffusion MRI group study of elderly healthy controls, patients with mild cognitive impairment (MCI), and mild or moderate Alzheimer's disease (AD). In addition, robustness against noise has been assessed with a realistic diffusion-weighted imaging phantom and the contamination of the local neighborhood with gray matter has been measured. RESULTS: The new texture operators showed an increased ability for finding formerly undetected differences between groups compared to conventional DTI methods. In particular, the coefficient of variation, quartile coefficient of variation, standard deviation and inter-quartile range of the mean and radial diffusivity detected significant differences even between previously not significantly discernible groups, such as MCI versus moderate AD and mild versus moderate AD. The analysis provided evidence of low contamination of the local neighborhood with gray matter and high robustness against noise. CONCLUSIONS: The local operators applied here enhance the identification and localization of areas of the brain where cognitive impairment takes place and thus indicate them as promising extensions in diffusion MRI group studies.
PURPOSE: Descriptors extracted from magnetic resonance imaging (MRI) of the brain can be employed to locate and characterize a wide range of pathologies. Scalar measures are typically derived within a single-voxel unit, but neighborhood-based texture measures can also be applied. In this work, we propose a new set of descriptors to compute local texture characteristics from scalar measures of diffusion tensor imaging (DTI), such as mean and radial diffusivity, and fractional anisotropy. METHODS: We employ weighted rotational invariant local operators, namely standard deviation, inter-quartile range, coefficient of variation, quartile coefficient of variation and skewness. Sensitivity and specificity of those texture descriptors were analyzed with tract-based spatial statistics of the white matter on a diffusion MRI group study of elderly healthy controls, patients with mild cognitive impairment (MCI), and mild or moderate Alzheimer's disease (AD). In addition, robustness against noise has been assessed with a realistic diffusion-weighted imaging phantom and the contamination of the local neighborhood with gray matter has been measured. RESULTS: The new texture operators showed an increased ability for finding formerly undetected differences between groups compared to conventional DTI methods. In particular, the coefficient of variation, quartile coefficient of variation, standard deviation and inter-quartile range of the mean and radial diffusivity detected significant differences even between previously not significantly discernible groups, such as MCI versus moderate AD and mild versus moderate AD. The analysis provided evidence of low contamination of the local neighborhood with gray matter and high robustness against noise. CONCLUSIONS: The local operators applied here enhance the identification and localization of areas of the brain where cognitive impairment takes place and thus indicate them as promising extensions in diffusion MRI group studies.
Entities:
Keywords:
Alzheimer’s disease; Diffusion tensor imaging; Local texture; White matter
Authors: Mark Jenkinson; Christian F Beckmann; Timothy E J Behrens; Mark W Woolrich; Stephen M Smith Journal: Neuroimage Date: 2011-09-16 Impact factor: 6.556