A hybrid system based on information gain and principal component analysis for the classification of transcranial Doppler signals

作者：Uguz Harun^*

来源：Computer Methods and Programs in Biomedicine, 2012, 107(3): 598-609.

DOI：10.1016/j.cmpb.2011.03.013

• Feature selection
• Principal component analysis
• Information gain
• Discrete wavelet transform
• Support vector machine

A transcranial Doppler (TCD) is a non-invasive, easy to apply and reliable technique which is used in the diagnosis of various brain diseases by measuring the blood flow velocities in brain arteries. This study aimed to classify the TCD signals, and feature ranking (information gain - IG) and dimension reduction methods (principal component analysis - PCA) were used as a hybrid to improve the classification efficiency and accuracy. In this context, each feature within the feature space was ranked depending on its importance for the classification using the IG method. Thus, the less important features were ignored and the highly important features were selected. Then, the PCA method was applied to the highly important features for dimension reduction. As a result, a hybrid feature reduction between the selection of the highly important features and the application of the PCA method on the reduced features were achieved. To evaluate the effectiveness of the proposed method, experiments were conducted using a support vector machine (SVM) classifier on the TCD signals recorded from the temporal region of the brain of 82 patients, as well as 24 healthy people. The experimental results showed that using the IG and PCA methods as a hybrid improves the classification efficiency and accuracy compared with individual usage.

• 出版日期2012-9