Abstract:

Cognitive decline presents a significant challenge in the field of neurology, necessitating early and accurate prediction models to enhance patient care and treatment planning. This study leverages the extensive, multi-modal OASIS-3 dataset, encompassing 30 years of longitudinal data, including MR imaging, PET imaging, and comprehensive clinical records, to develop predictive models for cognitive impairment. By integrating these diverse data sources, multi-modal predictive models were shown to outperform single-modality models, achieving higher accuracy (0.85) and AUC (0.90). Key predictors identified include cortical thickness in the medial temporal lobe, amyloid burden, and demographic factors such as age, underscoring their importance in understanding and forecasting cognitive decline. Visualisations, such as 3D brain models and progression charts, provided further insight into data trends and model reliability. These findings highlight the advantages of big data integration in predictive analytics and point towards personalised medicine applications that can significantly impact early diagnosis and targeted treatment. While the study demonstrates robust results, future work will focus on real-time data integration and cross-population model validation to enhance generalisability.

Keywords:

COGNITIVE DECLINE, BIG DATA ANALYTICS, PREDICTIVE MODELLING, MULTI-MODAL IMAGING, <br /> OASIS-3 DATASET, MR IMAGING, PET IMAGING, CORTICAL THICKNESS, AMYLOID BURDEN, PERSONALISED MEDICINE