Abstract-Diagnosis of hepatocellular carcinoma based on a terahertz signal and VMD-CWSE

Haishun Liu, Ke Zhao, Xiangyi Liu, Zhenwei Zhang, Jingyu Qian, Cunlin Zhang, and Meiyan Liang

Schematic sketch of the equipment

https://www.osapublishing.org/boe/abstract.cfm?uri=boe-11-9-5045

A novel strategy on combining variational mode decomposition (VMD) and composite weighted-scale sample entropy (CWSE) modified from composite multiscale entropy (CMSE) is proposed to screen hepatocellular carcinoma (HCC) by measuring the terahertz (THz) pulse signals of ten normal and ten HCC serums. Eight measured HCC specimens are negative in serum biomarker alpha fetoprotein (AFP) determination. In CWSE, the time series with weighted-scales are generated from the weighted average processing in the coarse-grained time series corresponding to each scale of the CMSE algorithm. VMD served as a preprocessing method was introduced into decomposing THz signal to obtain the mode functions of specific bandwidth for identification. Final results reveal that more obtainable entropy values of CWSE for recognition in comparison to those of CMSE on the basis of the rule of statistically significant difference and effect size and also manifest the stronger discriminability than the traditional THz parameters. This study provides a new potential auxiliary tool for diagnosis HCC and develops the methodology on the discrimination for similar THz signals.

© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

Abstract-Dimensionality Reduction for Identification of Hepatic Tumor Samples Based on Terahertz Time-Domain Spectroscopy

 Haishun Liu,  Zhenwei Zhang, Xin Zhang,  Yuping Yang, Zhuoyong Zhang, Xiangyi Liu, Fan Wang,  Yiding Han, Cunlin Zhang

https://ieeexplore.ieee.org/document/8333778/

Terahertz time-domain spectroscopy (THz-TDS) combining with chemometrics methods was proposed for the identification of hepatic tumors. Two linear compression methods, principle component analysis and locality preserving projections (LPPs), and a nonlinear method, Isomap, were used to reduce the dimensionality of the measured dataset. Comparing two-dimensional (2-D) data reduced by these three dimensionality reduction techniques, only 2-D Isomap plot could separate the distances between two classes for the THz time-domain data and LPP had capacity of distinguishing two types of samples building on frequency-domain data. The best classification accuracies from 2-D time-domain data were 99.81±0.30% and 99.69±0.61% given by Isomap probabilistic neural network (PNN) and Isomap support vector machine (SVM), respectively, while the best classification results of 2-D frequency-domain data were 100.00±0.00%, 99.75±0.32% provided by LPP-PNN, LPP-SVM. The results showed that Isomap and LPP are appropriate techniques to reflect the nonlinear manifold of the THz data. The THz technology either in time-domain or frequency-domain coupled with Isomap-PNN or LPP-PNN could offer a potential procedure to identify hepatic tumors.