Breast Cancer classification using extracted parameters from a terahertz dielectric model of human breast tissue.

Our previous study proposed a dielectric model for human breast tissue and provided initial analysis of classification potential of the eight model parameters and their multiparameter combinations with the support vector machine (SVM). A combination of three model parameters could achieve a leave-one-out cross validation accuracy of 93.2%. However, the SVM approach fails to exploit the combinations of more than three model parameters for classification improvement. Thus, the Bayesian neural network (BNN) method is employed to overcome this problem based on its advantages of handling our small data and high complexity of the multiparamter combinations. The BNN successfully classifies the data using the combinations of four model parameters with an accuracy, estimated by leave-one-out cross validation, of 97.3%. Overall performance assessed by leaveone-out and repeated random-subsampling cross validations for all examined combinations is also remarkably improved by BNN. The results indicate the advance of BNN as compared to SVM in utilising the model parameters for detecting tumour from normal breast tissue.

High correlation of double Debye model parameters in skin cancer detection.

The double Debye model can be used to capture the dielectric response of human skin in terahertz regime due to high water content in the tissue. The increased water proportion is widely considered as a biomarker of carcinogenesis, which gives rise of using this model in skin cancer detection. Therefore, the goal of this paper is to provide a specific analysis of the double Debye parameters in terms of non-melanoma skin cancer classification. Pearson correlation is applied to investigate the sensitivity of these parameters and their combinations to the variation in tumor percentage of skin samples. The most sensitive parameters are then assessed by using the receiver operating characteristic (ROC) plot to confirm their potential of classifying tumor from normal skin. Our positive outcomes support further steps to clinical application of terahertz imaging in skin cancer delineation.

Near-infrared parameters extraction: A potential method to detect skin cancer.

The wavelength-dependent absorption coefficients can be used to analyse optical properties of human skin. Existing absorption models for narrow ranges in the visible and near infrared are insufficient to simultaneously incorporate the spectral contrast produced by differences in chromophores, water and lipid content of skin tissue into skin cancer detection. In the broad range up to 1600 nm, recent analysis approaches for absorption spectra do not consistently provide significant differences between healthy and cancerous skins. We propose an absorption model to fit the absorption coefficient spectra of skin samples over the range from 400 nm to 1600 nm and an advanced algorithm to find the optimal estimation. The extracted parameters of this model are analysed by a statistical t-test. The test results demonstrate the significant differences between all pairs of tumour-normal skin. Therefore, our approach has strong potential for early skin cancer detection using near infrared spectroscopy (NIRS).

Global optimization for human skin investigation in terahertz.

In this paper, the electromagnetic interaction between human skin and terahertz radiation is investigated through the double Debye parameters' extraction algorithm. The changes of skin content are contrasted at the frequencies below one terahertz(THz) but the recent approaches could provide only a rough estimation. We propose an global optimization based identification, which results in globally accurate estimators in the frequency range up to two THz, and thus supports the validity of Debye model for Terahertz wave's propagation and reflection in skin. Simulation results confirm our prominent methodology.

Optimal design of FIR triplet halfband filter bank and application in image coding.

This correspondence proposes an efficient semidefinite programming (SDP) method for the design of a class of linear phase finite impulse response triplet halfband filter banks whose filters have optimal frequency selectivity for a prescribed regularity order. The design problem is formulated as the minimization of the least square error subject to peak error constraints and regularity constraints. By using the linear matrix inequality characterization of the trigonometric semi-infinite constraints, it can then be exactly cast as a SDP problem with a small number of variables and, hence, can be solved efficiently. Several design examples of the triplet halfband filter bank are provided for illustration and comparison with previous works. Finally, the image coding performance of the filter bank is presented.