Answers are grounded in this paper and cite their sources.
Alzheimer's Disease (AD) is a progressive neurodegenerative disorder marked by cognitive decline, memory loss, and impaired reasoning, caused by the accumulation of amyloid-beta plaques and neurofibrillary tangles of tau protein. This study utilizes neuroimaging Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) and structural MRI and clinical data from 1069 subjects acquired through the Alzheimer's Disease Neuroimaging Initiative (ADNI). Early-stage AD presents challenges for classification between Mild Cognitive Impairment (MCI) and Cognitively Normal (CN) individuals, as structural abnormalities are less pronounced, making sMRI-derived features less discriminative. However, functional connectivity disruptions in brain regions are notable in the early AD stages. This study investigates whether the metabolic brain assessment biomarker, Support Uptake Value Ratio (SUVR) from FDG-PET, can enhance classification performance between MCI and CN classes effectively or not. We introduced a novel weighted random SVM cluster, an ensemble method that outperforms single SVM classifiers by addressing issues of limited data and noise. Using the AAL3 Atlas, which provides finer subdivisions of brain regions (168 regions compared to AAL2's 90), we achieved a more detailed connectivity analysis. Employing rigorous preprocessing and a Stratified K-Fold Nested Cross Validation with k=5 and an 8:2 train-validation to test split ensured robust hyperparameter tuning and model selection to prevent overfitting and selection bias. The proposed method demonstrates good classification accuracy for MCI vs. CN and exhibits good ROC characteristics, indicating sensitivity to detect disruptive metabolic changes, highlighting its potential in early AD detection.
No takes yet. Share an insight, caveat, or question.
Varnagar et al. (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: