Classification Of Terahertz Pulsed Signals From Breast Tissues Using Wavelet Packet Energy Feature Exaction And Machine Learning Classifiers

Here, we propose an effective classification strategy for THz pulsed signals of breast tissues based on wavelet packet energy (WPE) feature exaction and machine learning classifiers. The paraffin-embedded breast tissue samples were adopted in this study and identified as tumor (226 samples), healthy fibrous tissue (233 samples) or adipose tissue (178 samples) based on the histological results. Firstly, the THz pulsed signals of tissue samples were acquired using a standard transmission THz time-domain spectrometer. Then, the signals were decomposed by the wavelet packet transform (WPT) and the features of the WPE were extracted. To reduce the dimensionality of extracted features, the principal components analysis (PCA) method was employed. Six different machine learning classifiers were then performed and compared for automatic classification of different tissue samples. The highest classification accuracy is up to 97% using the fine Gaussian support vector machine (SVM) approach. The results indicate that the WPE feature exaction combined with machine learning classifier can be used for automatic evaluation of biological tissue THz signals with good accuracy.