Texture Features of Magnetic Resonance Images: A Marker of Slight Cognitive Deficits in Parkinson's Disease

Abstract

Background Cognitive impairment is a frequent nonmotor symptom of Parkinson's disease. Depending on severity, patients are considered to have mild cognitive impairment or dementia. However, among the cognitively intact patients, some may have deficits in a less severe range. The early detection of such subtle symptoms may be important for the initiation of care strategies. Objective To identify imaging markers of early cognitive symptoms, potentially before usual signs, such as atrophy, become manifest. Methods A total of 102 patients with Parkinson's disease and 17 age‐matched cognitively intact healthy controls underwent extensive neuropsychological assessment and T1‐weighted magnetic resonance imaging. Parkinson's disease patients were separated into 3 groups according to their cognitive status: intact, with slight slowing, and with mild deficits in executive functions. Texture features as measured by first‐order and second‐order statistics were computed in the following 6 brain regions: the hippocampus, thalamus, amygdala, putamen, caudate nucleus, and pallidum. They were tested between the groups, and their correlation with cognition was examined. Volumetric measurements were made for comparison. Results Texture analysis showed significant between‐group differences for 2 features—skewness and entropy in the hippocampus, the thalamus, and the amygdala—and the volume analysis revealed no between‐group difference. These features were significantly correlated with cognitive performance. Conclusion These results support the assumption that signal alterations associated with Parkinson's disease–related cognitive decline can be captured very early by texture analysis. As these changes appear to reflect clinical phenomena, texture analysis may be a promising marker for helping cognitive phenotyping in Parkinson's disease. © 2019 International Parkinson and Movement Disorder Society

Posterior Cortical Cognitive Deficits Are Associated With Structural Brain Alterations in Mild Cognitive Impairment in Parkinson’s Disease

“…To study structural parameters less sensitive to segmentation accuracy, texture analyses were performed on the left and right caudate nuclei, thalami, and hippocampi. Detailed presentation of the methodology can be found in Betrouni et al ( 2020 ). Briefly, from the binary masks of each structure obtained from the volBrain segmentation, we have computed four first-order features to study the gray-level distribution in all voxels composing a given structure without considering spatial relationships, namely, the mean, the standard deviation, the kurtosis, and the skewness.…”

“…In the shape analysis, each point in a patient’s mesh was scaled with the total intracranial volume of this patient. In the texture analysis, given the large number of variables (six brain structures, 11 features by structure), we used a feature selection strategy similar to that of Betrouni et al ( 2020 ) to reduce the dimensionality. Spearman’s correlation coefficients were computed between the texture features and the PD-MCI subtypes for each structure.…”

Detection of Parkinson’s Disease from 3T T1 Weighted MRI Scans Using 3D Convolutional Neural Network

“…It was observed from the study that the textural features can be considered for discriminating the structural changes. Moreover, similar studies were also performed by [4,23] where the textural and morphological features were only considered from specific subcortical structures of the brain for the prediction of Parkinson's Disease. But concerning the implementation of CNN and deep learning on MRI scans for the detection of Parkinson's Disease, there are quite a few notable works that have been performed.…”

Anxiety in Parkinson's disease is associated with changes in the brain fear circuit

“…Texture analysis is an image processing method for the quantification of grey levels inside an image [16]. The procedure is detailed in Betrouni et al [17]. Here, we compared four first-order and seven second-order texture parameters detailed in Supplementary Table 1.…”