Laser & Optoelectronics Progress, Volume. 60, Issue 2, 0230001(2023)
Raman Spectral Segmentation Method for Tongue Squamous Cell Carcinoma Using Deep Learning
Raman spectrum can indicate the changes in the molecular structure of living tissues and be used for the detection of tongue squamous cell carcinoma tissues. While the existed technologies can only identify the characteristics of tongue squamous cell carcinoma tissue and establish whether the tissue is cancerous, they cannot locate crucial band sections of the Raman spectrum of tongue squamous cell carcinoma tissue. Therefore, based on a deep learning algorithm, this study aims to present a spectral region segmentation technique for identifying significant bands of Raman spectra of the tongue squamous cell carcinoma. First, the Raman spectrum data of 44 tumor tissues from 22 patients were obtained using fiber-optic Raman spectroscopy acquisition equipment. The data were preprocessed, annotated, and split into the training set and testing set. Next, a band region deep convolutional neural network model was created. This model is composed of three fundamental modules, namely, Raman spectral feature extraction network, crucial spectral band recommendation network, and critical spectral band regression network. Among these, the Raman spectral feature extraction network is used to extract the spectral characteristics of tongue squamous cell carcinoma tissues and crucial bands. The crucial spectral band recommendation network and the crucial spectral band regression network are used to segment the essential band regions of the tongue squamous cell carcinoma tissue spectrum. Experimental findings show that the average precision of the proposed method for significant bands in tongue squamous cell carcinoma tissue is 99% under the criterion of interest of union value of 0.7.
Get Citation
Copy Citation Text
Jinyang Liu, Mingxin Yu, Shengnan Ji, Lianqing Zhu, Tao Zhang, Jingya Ding, Jiabin Xia. Raman Spectral Segmentation Method for Tongue Squamous Cell Carcinoma Using Deep Learning[J]. Laser & Optoelectronics Progress, 2023, 60(2): 0230001
Category: Spectroscopy
Received: Oct. 11, 2021
Accepted: Nov. 29, 2021
Published Online: Jan. 6, 2023
The Author Email: Yu Mingxin (yumingxin@bistu.edu.cn)