T-ray relevant frequencies for osteosarcoma classification W. Withayachumnankul a,d , B. Ferguson b,d , T. Rainsford d , D. Findlay c , S. P. Mickan d , and D. Abbott d a Department of Information Engineering, Faculty of Engineering, King Mongkut’s Institute of Technology Ladkrabang, Bangkok 10520, Thailand b Tenix Systems Pty Ltd, 2nd Avenue, Mawson Lakes, SA 5095, Australia c Centre for Biomedical Engineering (CBME) and Department of Orthopaedics & Trauma, The University of Adelaide and Hanson Institute, SA 5005, Australia d Centre for Biomedical Engineering (CBME) and Department of Electrical & Electronic Engineering, The University of Adelaide, SA 5005, Australia ABSTRACT We investigate the classification of the T-ray response of normal human bone cells and human osteosarcoma cells, grown in culture. Given the magnitude and phase responses within a reliable spectral range as features for input vectors, a trained support vector machine can correctly classify the two cell types to some extent. Performance of the support vector machine is deteriorated by the curse of dimensionality, resulting from the comparatively large number of features in the input vectors. Feature subset selection methods are used to select only an optimal number of relevant features for inputs. As a result, an improvement in generalization performance is attainable, and the selected frequencies can be used for further describing different mechanisms of the cells, responding to T-rays. We demonstrate a consistent classification accuracy of 89.6%, while the only one fifth of the original features are retained in the data set. Keywords: Terahertz time-domain spectroscopy, T-rays, support vector machines, feature selection, osteosar- coma, signal classification, cancer detection, curse of dimensionality 1. INTRODUCTION T-rays, spanning the range from 0.1 to 10 THz in the electromagnetic spectrum, have a great potential in biomedical applications. 1, 2 This results from distinctive properties of biomolecules in this frequency range. DNA and specific molecules, such as amino acids, peptides, and proteins, have resonances at T-ray frequencies. 3 T-rays are non-ionizing radiation, and represent a totally non-invasive diagnostic technique. 4 Due to strong absorption by water, T-rays produce skin-depth level contrast, in which X-rays fail. Optical and infrared frequencies suffer from Rayleigh scattering, not present with T-rays due to the longer wavelength. 5 Biomaterial classification is one promising application of T-rays. It employs time-gated detection techniques 6 using terahertz time-domain spectroscopy (THz-TDS) to produce high SNRs and coherent signals. The signals, when passing through materials with different quantities of interstitial water, are subject to different amounts of attenuation and dispersion. This information leads to rich features useful for classification. Woodward et al. 7, 8 investigated and classified basal cell carcinoma, one form of skin cancer, in vitro and in vivo with a T-ray reflection geometry. Ferguson et al. 9 distinguished two types of meats using chirped probe T-ray imaging system. L¨offleretal. 10 classified tumors in sliced tissues with T-ray pulsed imaging. W. Withayachumnankul, Email: kwwithaw@kmitl.ac.th; B. Ferguson, Email: brad.ferguson@tenix.com; T. Rainsford, Email: tamath@eleceng.adelaide.edu.au; D. Findlay, Email: david.findlay@adelaide.edu.au; S. P. Mickan, Email: spmickan@eleceng.adelaide.edu.au; D. Abbott, Email: dabbott@eleceng.adelaide.edu.au Photonics: Design, Technology, and Packaging II, edited by Derek Abbott, Yuri S. Kivshar, Halina H. Rubinsztein-Dunlop, Shanhui Fan, Proc. of SPIE Vol. 6038, 60381H, (2006) · 0277-786X/06/$15 · doi: 10.1117/12.637964 Proc. of SPIE Vol. 6038 60381H-1 Downloaded From: http://proceedings.spiedigitallibrary.org/ on 11/12/2012 Terms of Use: http://spiedl.org/terms