A Novel Method of Copula Connectivity-based and Collatz Patterns Schemes for Detection of Mild Cognitive Impairment and Alzheimer Patients using Magnetoencephalogram

Biglari, Hadis; Yazdi, Mehran; Taghavi, Mahsa; Sanei, Saeid; Boostani, Reza

doi:10.24200/sci.2026.66179.10058

A Novel Method of Copula Connectivity-based and Collatz Patterns Schemes for Detection of Mild Cognitive Impairment and Alzheimer Patients using Magnetoencephalogram

Articles in Press

Document Type : Research Article

Authors

Hadis Biglari ¹

Mehran Yazdi ¹

Mahsa Taghavi ²

Saeid Sanei ³

Reza Boostani ¹

¹ School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran.

² Psychiatric Department, Medical School, Azad University, Kazerun Branch, Kazerun, Fars, Iran.

³ Department of Electronic Engineering, Imperial College London, London, London, UK.

10.24200/sci.2026.66179.10058

Abstract

Differentiating between Alzheimer’s Disease (AD), Mild Cognitive Impairment (MCI), and healthy controls (HCs) remains a significant challenge in neuroscience. The existing methods based on electroencephalogram (EEG) or magnetoencephalogram (MEG) connectivity analysis, often focus on two-class classification and struggle to effectively capture both functional and causal brain connectivity for a three-class differentiation. To address this gap, we propose a novel multilevel connectivity scheme that integrates Copula theory and the Collatz pattern to estimate both linear and nonlinear connectivity levels between MEG channels. MEG signals from 62 participants (24 AD, 14 MCI, and 24 HC) were analyzed, with features combined using canonical correlation analysis (CCA) and refined through principal component analysis (PCA) and minimum redundancy maximum relevance (mRMR). This hybrid method, utilizing Frank Copula and Collatz pattern-based features, achieved a classification accuracy of 98.94% for three-class differentiation. The innovation stems from the ability of Copula features to capture both functional and causal relationships. When these features are combined with Collatz pattern features, the fusion approach outperforms the existing connectivity methods, including Granger causality. This approach provides a more accurate and comprehensive method for distinguishing between AD, MCI, and HC, filling a crucial gap in current neuroimaging research.

Keywords

Alzheimer

Collatz Patterns

Copula Function

Brain Connectivity

MEG

Feature Fusion

Subjects