Retrospective study investigates serum elemental profiling and machine learning for Alzheimer's identification and cognitive prediction, suggesting a less invasive approach.
Key Points
The study aims to explore whether serum elemental profiling can aid in Alzheimer's disease identification and cognitive score prediction using machine learning methods.
Retrospective cross-sectional study of 874 participants from 2017 to 2023
Utilized inductively coupled plasma mass spectrometry to quantify serum element concentrations
Developed machine learning models for AD identification and regression models for MMSE and MoCA score prediction.
Odds of AD increased with higher serum lead (OR=4.95, 95% CI: 3.42–7.37) and tin (OR=1.45, 95% CI: 1.20–1.78) levels.
Random forest model achieved 88% accuracy and an AUC of 0.94 in distinguishing AD from NC.
Random forest regression showed strong correlations for predicting cognitive scores: MMSE (r=0.48, p<0.001) and MoCA (r=0.62, p<0.001).