西安电子科技大学学报（自然科学版）2025，Vol.52Issue(6)：45-57,13.DOI:10.19665/j.issn1001-2400.20251007
乳腺癌细胞拉曼光谱的数据增强与识别方法
Method of data augmentation and recognition for the Raman spectroscopy of breast cancer cells
摘要
Abstract
Raman spectroscopy data are characterized by high noise,high dimensionality,and costly acquisition,which complicates data mining and utilization,consequently limiting the performance of deep learning models.To enhance the analytical performance of models for Raman spectra,this paper proposes a breast cancer cell identification method that incorporates Raman spectral data augmentation.First,the data is augmented using four methods,namely Gaussian filtering,random erasing,multi-sample fusion,and signal weakening.Subsequently,a feature repair technique is employed to optimize the augmented data.Then,a neural network model is designed to extract Raman spectral features of breast cancer cells,and primary classification models are trained separately on the four augmented datasets.Finally,these primary models are integrated using a Stacking ensemble framework,with a meta-learner trained to perform the final classification of breast cancer cells.Experimental results show that the proposed data augmentation improvements can increase the model accuracy by 0.21％ to 16.16％.Furthermore,our method outperforms existing similar methods in terms of accuracy,sensitivity,and F1-score for breast cancer cell identification.This study enhances the effectiveness of deep learning methods for identifying breast cancer cells and contributes to promoting the application of Raman spectroscopy-based cell identification in breast cancer diagnosis.关键词
拉曼光谱/深度学习/乳腺癌/数据处理/人工智能Key words
Raman spectroscopy/deep learning/breast cancer/data handling/artificial intelligence分类
信息技术与安全科学引用本文复制引用
郭阳,陈生聘,兰静芬,于强..乳腺癌细胞拉曼光谱的数据增强与识别方法[J].西安电子科技大学学报（自然科学版）,2025,52(6):45-57,13.基金项目
陕西省自然科学基础研究计划(2024JC-YBMS-044) （2024JC-YBMS-044）
