Identification of cancerous gastric cells based on common features extracted from hyperspectral microscopic images

Siqi Zhu; Kang Su; Yumeng Liu; Hao Yin; Zhen Li; Furong Huang; Zhenqiang Chen; Weidong Chen; Ge Zhang; Yihong Chen

doi:10.1364/BOE.6.001135

Identification of cancerous gastric cells based on common features extracted from hyperspectral microscopic images

Biomed Opt Express. 2015 Mar 4;6(4):1135-45. doi: 10.1364/BOE.6.001135. eCollection 2015 Apr 1.

Authors

Siqi Zhu¹, Kang Su¹, Yumeng Liu¹, Hao Yin¹, Zhen Li¹, Furong Huang¹, Zhenqiang Chen¹, Weidong Chen², Ge Zhang², Yihong Chen³

Affiliations

¹ Key Laboratory of Optoelectronic Information and Sensing Technologies of Guangdong Higher Education Institutes, Jinan University, Guangzhou, Guangdong 510632, China ; Department of Optoelectronic Engineering, Jinan University, Guangzhou, Guangdong 510632, China.
² Key Laboratory of Optoelectronic Materials Chemistry and Physics, Fujian Institute of Research on the Structure of Matter, Chinese Academy of Sciences, Fuzhou, Fujian 350002, China.
³ Key Laboratory of Optoelectronic Information and Sensing Technologies of Guangdong Higher Education Institutes, Jinan University, Guangzhou, Guangdong 510632, China ; Department of Optoelectronic Engineering, Jinan University, Guangzhou, Guangdong 510632, China ; Guangzhou Ante Laser Technology Co. Ltd, Guangzhou, Guangdong 510663, China.

Abstract

We construct a microscopic hyperspectral imaging system to distinguish between normal and cancerous gastric cells. We study common transmission-spectra features that only emerge when the samples are dyed with hematoxylin and eosin (H&E) stain. Subsequently, we classify the obtained visible-range transmission spectra of the samples into three zones. Distinct features are observed in the spectral responses between the normal and cancerous cell nuclei in each zone, which depend on the pH level of the cell nucleus. Cancerous gastric cells are precisely identified according to these features. The average cancer-cell identification accuracy obtained with a backpropagation algorithm program trained with these features is 95%.

Keywords: (110.4234) Multispectral and hyperspectral imaging; (170.0170) Medical optics and biotechnology; (170.6935) Tissue characterization.