AD-Statistical-And-Predictive-Modelling
Alzheimer's disease (AD), the most prevalent form of dementia, affects millions worldwide and demands more effective diagnostic techniques. This study integrates neuroimaging and multi-omic data to develop advanced statistical and predictive models aimed at early and accurate AD diagnosis. We assessed several models across different data modalities, including genetic, gene expression, and electronic health records, to identify novel biomarkers and understand the complex interplay of biological factors contributing to AD. Our comprehensive analysis involves diverse data modalities and classification models, with standout performances from the Ensemble model achieving a remarkable 98% accuracy and Graph Neural Networks demonstrating a high recall rate in genetic data. Crucially, we implemented Non-negative Matrix Factorization (NMTF) for effective dimensionality reduction, which preserved the biological interpretability of the high-dimensional multi-omic data. Extensive statistical tests, including hypothesis testing, ANOVA for variance analysis and non-parametric tests to validate model robustness, were integral in refining our predictive models. These methodologies not only heighten the precision of AD diagnostics but also enrich our understanding of its pathophysiological complexity. By integrating advanced computational techniques and rigorous statistical analysis, our work paves the way for transformative improvements in the diagnosis and management of dementia, promising significant advancements in patient outcomes.