Blood Biomarker-Based Machine Learning Triaging Model for Predicting the Risk of Cognitive Decline Progression from Mild Cognitive Impairment

Early intervention efforts critically depend on pin-pointing individuals with mild cognitive impairment (MCI) who face an elevated risk of progressing to Alzheimer's disease (AD). Therefore, this research evaluated the utility of models leveraging Machine Learning, incorporating blood-based biomarkers and demographic data, to gauge an individual's risk of MCI. We used data from the ADNI dataset with 691 samples, which incorporates five key blood biomarkers and three demographic data. Four Machine Learning models, Support Vector Machines (SVM), Random Forest, Extreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP), were subjected to training and assessment through a 10-fold stratified cross-validation process. The Random Forest model displayed exceptional performance among the set, attaining a mean ROC_AUC score of 0.908 (95% CI: 0.886 - 0.930). These findings suggest that Machine Learning applied to blood biomarkers can effectively stratify MCI patients by their AD progression risk, offering a promising approach for early and precise therapeutic strategies.