The prediction of cardiovascular disease based on trace element contents in hair and a classifier of boosting decision stumps.

The early discovery of cardiovascular disease (CVD) is crucial for performing successful treatments. This study aims at exploring the feasibility of Adaboost (ensemble from machining learning) using decision stumps as weak classifier, combined with trace element analysis of hair, for accurately predicting early CVD. A total of 124 hair samples composed of two groups of samples (one is healthy group from 100 healthy persons aged 24-72 while the other is patient group from 24 cardiovascular disease patients aged 36-81) were used. Nine kinds of trace elements, i.e., chromium (Cr), manganese (Mn), cadmium (Cd), copper (Cu), zinc (Zn), selenium (Se), iron (Fe), aluminum (Al), and nickel (Ni), were selected. In a preliminary analysis, no obvious linear correlations between elements can be observed and the concentration of Cr, Fe, Al, Cd, Ni, or Se for healthy group is higher than that for patient group while the opposite is true for Mn, Cu, or Zn, indicating that both low Se/Fe and high Mn/Cu can be identified as major risk factors. Based on the proposed approach, the final ensemble classifier, constructed on the training set and contained only four decision stumps, achieved an overall identification accuracy of 95.2%, a sensitivity of 100% and a specificity of 94% on the independent test set. The results suggested that integrating Adaboost and trace element analysis of hair sample can serve as a useful tool of diagnosing CVD in clinical practice.