What are the main predictors of age-related cognitive decline?
Objective 1: To analyze all available patient data, stratified by extent of age-related cognitive decline, using standard statistical software
Objective 2: To build a machine-learning (ML) based predictive model that incorporates all collected data and identifies factors associated with cognitive decline.
Age-related cognitive decline is an extremely common phenomenon. Outside of cases with a specific medical diagnosis, the cause of cognitive deterioration is usually unknown and therefore difficult to treat. Likely, the etiology is multifactorial, involving genetics, lifestyle factors, and comorbid medical conditions. By identifying the most prominent risk factors for cognitive decline, we can better advise patients on prevention and time interventions at the most critical timepoints. We may be able to develop new therapies which can be individualized by patient etiology.
We seek to include all available patient data, including demographics, clinical information, laboratory results, and genomics, in order to have the broadest possible scope of possible risk factors and to evaluate confounding variables. This data will be analyzed using descriptive statistics with standardized statistical software such as MatLab and Stata, and the results will be compared to that of our ML model.
ML offers a powerful toolset for developing predictive models capable of detecting complex, non-linear relationships within large datasets. Unlike conventional statistical methods, ML algorithms can continually learn from data, improving predictive accuracy over time. Several ML algorithms have shown promise for clinical use, both alone or in combination, including Random Forest, Convolutional Neural Networks, and Support Vector Machine. Previous studies have shown the efficacy of ML-based models in early detection of breast cancer, colon cancer, and melanoma, among others.