Alcoholism via wavelet energy entropy and support vector machine

Abstract

Alcohol has become a common drink in social etiquette that people inattention to their alcohol intake, resulting in alcoholism. Clinically, it is difficult for the physician to quickly determine whether a patient is at risk for alcoholism unless the physician has sufficient experience to achieve a rapid diagnosis. However, the non-obvious manifestations and potential harms have not been solved in time, so that patients often miss the optimal adjustment period. Based on the common electroencephalogram (EEG), some studies have attempted to combine imaging with computer-aided diagnosis to assist physicians to complete a more sophisticated diagnosis. Various computer-aided methods emerge endlessly and bring good potential application prospects. In this paper, we propose a new method to extract the energy entropy of brain image after wavelet transformation, which combined with a support vector machine classifier for alcohol intoxication detection. In the experiment, our method obtained 92.34±1.86% sensitivity, 92.72±1.00% specificity and 92.53±0.80% accuracy, showing better application ability compared with the new technique.