Label-Free Cervical Cancer Detection in Tissue Samples Using Raman Spectroscopy Combined with Machine Learning technology

Abstract

Cervical cancer is one of the most common malignant tumors of the female reproductive tract, ranking fourth among all female malignancies. Currently, the primary method for diagnosing cervical cancer is histopathological examination, which mainly relies on hematoxylin and eosin (H&E) staining of tumor tissues to evaluate cellular and structural morphology. However, conventional H&E staining depends heavily on the pathologist's experience and lacks the ability to explore molecular changes in tissue samples. In this study, Raman spectroscopy, an optical analysis technology, was used to characterize molecular feature of tissue samples from cervical cancer patients. Furthermore, a classification model was constructed using the principal component analysis (PCA) and linear discriminant analysis (LDA) algorithm for spectral data analysis. The results revealed alterations in biomolecular components such as lipids, proteins, and glycogen, enabling discrimination between malignant and normal tissues with an accuracy of 100%. Moreover, staging analysis was performed, achieving an accuracy of 70.8% in distinguishing early-stage (I-II) from advancedstage (III-IV) cervical cancer tissues. Finally, we generated heat maps based on Raman signals to explore a novel visualization-based approach for differentiating cervical cancer tissues. This work integrating Raman spectroscopy with machine learning would provide a rapid and accurate method based on molecular profiling for cervical cancer detection.