Associations of Amyloid Deposition and FDG Uptake in Aging and Cognitively Impaired Elders With and Without Moderate to Severe Periventricular White Matter Hyperintensities Using Simple Machine Learning Techniques
Year: 2020
Type: dissertation
Study design: Observational Study
Abstract: Background: Cerebral small vessel disease (SVD) often coexists with Alzheimer’s Disease (AD), but subjects with significant SVD were usually considered mixed disease and excluded from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and other international cohorts. We use 2 machine learning algorithms (random forests (RFs) and K-means clustering) and 2 radiopharmaceuticals (18F-Florbetabir and 18F-FDG) to study associations of periventricular white matter hyperintensity (pvWMH), amyloid deposition and 18F-FDG uptake in aging and cognitively impaired elders. Methods: A multi-center, observational, prospective cohort study was designed to acquire data in subjects with significant pvWMH. Baseline data from the first 57 subjects recruited through tertiary memory clinics or stroke prevention clinics formed the basis for analysis. All subjects had a 3T MRI, 18F-Florbetabir PET and neurological evaluation; 18F-FDG PET was available in 55. A matched cohort of 57 subjects with minimal pvWMH was derived from ADNI. PETs were interpreted by 2 dual certified radiologists/ nuclear medicine physicians with MIM software (MIM Software Inc, Cleveland, Ohio) and processed to obtain quantitative data in regions of interest (ROIs). Algorithms programmed in MATLAB were used to classify 18F-Florbetabir PET as positive or negative for amyloid deposition, suggest ROIs for classification, and study associations with pvWMH. Discussion: This work complements current data on associations of amyloid, 18F-FDG and pvWMH. RFs (supervised) and K-means clustering (unsupervised) had similar classification accuracy for 18F-Florbetabir PET with clinical interpretation as the gold standard. ROIs for classification were, most commonly: the left posterior cingulate, left precuneus and left middle fontal gyrus; key nodes of the default node network, where amyloid deposition preferentially begins in the brain of patients with AD. Subjects with significant pvWMH had slightly more amyloid deposition across a spectrum of ROIs compared with their matched ADNI cohort. 18F-FDG uptake was slightly higher in several cortical ROIs with lower uptake in the basal ganglia. Possibly these findings can be explained by poor amyloid clearance related to venous collagenosis in subjects with significant pvWMH coupled with metabolic compensation. Future research will include evaluation of 2-year follow-up, correlation of PET with MRI, neurological assessment and ApoE4 status in the full cohort of subjects.
Source: TSpace (University of Toronto)
Author: Katherine Zukotynski
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W3093664434