Premium
IC‐P‐083: MRI Patch‐Based Imaging Biomarker for Automatic Detection of Parkinson Disease
Author(s) -
Fonov Vladimir S.,
Dadar Mahsa,
Yiming Xiao,
Collins D. Louis
Publication year - 2016
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2016.06.112
Subject(s) - subthalamic nucleus , parkinson's disease , substantia nigra , receiver operating characteristic , dopaminergic , basal ganglia , medicine , magnetic resonance imaging , biomarker , disease , nuclear medicine , neuroscience , pathology , psychology , dopamine , deep brain stimulation , radiology , biology , central nervous system , biochemistry
Background:Parkinson’s disease (PD) is a common neurodegenerative disorder affecting the motor system due to loss of dopaminergic neurons in the substantia nigra in addition to nondopaminergic system degeneration mainly in the basal ganglia region. Finding sensitive biomarkers of PD would facilitate the clinical management, particularly in the early stages of disease. Sensitive biomarkers could also effectively help advance drug development for the condition. Recently, we have developed a method for early detection of Alzheimer’s disease based on the automatic analysis of T1w MRI scans. Here we adapt this method for detection of PD. Methods: The method is based on nonlocal patch-based algorithm, estimating anatomical patterns from a given subject in a specific brain region and comparing them with a database composed of healthy subjects and patients. The output of the method is a grading value showing similarity towards NC (-1) or PD (+1).We tested the method using the baseline 3T high resolution T1-weighted MRI scans obtained from the Parkinson’s Progression Markers Initiative (PPMI) database (www.ppmi-info.org/data), corresponding to newly diagnosed PD patients (n1⁄4198) and an age-matched control group (n1⁄493). Results: We performed 10fold cross-validation experiments and measured grading values for the substantia nigra (SN) and subthalamic nucleus (STN). The classifier performance was tested by estimating the Area Under the receiver operating characteristic Curve (AUC). Our results showed statistically significant differences in grading values for left SN (p1⁄40.013) and left STN (p1⁄40.0002) between patients and controls. When used to classify individual subjects in the crossvalidation experiment the performance of the classifier yields an AUC1⁄40.662 for the left SN, and AUC1⁄40.733 for the left STN. Conclusions:We have shown that our method could be applied for detecting PD with AUC similar to Alzheimer Disease vs NC detection shown previously (AUC1⁄40.73 for hippocampal grading in AD, Coupe et al, 2015).