| Surface-enhanced Raman Spectroscopy(SERS)is capable of providing various information on molecular changes,offering advantages such as high speed and accuracy,and has been extensively applied in the field of biomedicine.From this,we proceeded to collect serum RNA surface-enhanced Raman Spectroscopy samples from patients with hysteromyoma and patients with cervical cancer,and carried out feature selection on the two types of spectral data through the feature selection algorithm,combined with the machine learning classification model to classify the two groups of spectra analysis and classification predictions were performed for screening cervical cancer.This paper mainly conducts research in the following three aspects:1.Collection and analysis of surface-enhanced Raman spectroscopy signals.The average spectroscopy after preprocessing shows that the bands of uterine fibroids and cervical cancer show obvious features at wave numbers of 448cm-1,518cm-1,698cm-1,750cm-1,1003cm-1 and 1076cm-1 peak.The change of the vibration mode of the biochemical substance corresponding to the characteristic peak is consistent with the change of the vibration mode of the biochemical substance in the patient’s body during the cervical cancer process.For example,the 1076cm-1 is related to the symmetrical phosphoric acid stretching mode,which is usually derived from the phosphodiester group.And associated with increased nucleic acid in malignant tissue.2.A new method for screening cervical cancer patients based on serum RNA is proposed.First,feature selection was performed on the surface-enhanced Raman signal of cervical serum RNA using principal component analysis and independent sample t-test,and classification and diagnosis were performed by combining LDA,SVM and AdaBoost machine learning models.The results showed that the diagnostic results with accuracy,specificity and sensitivity higher than 94%were obtained by using the AdaBoost model,which was better than that without feature selection.3.The surface-enhanced Raman spectroscopy signal of cervical serum RNA samples combined with the ant colony algorithm was studied.The feature selection of feature grouping and feature non-grouping is compared,and the cumulative feature selection frequency of repeated 100 runs is used as the basis to enhance the robustness of the overall model.The results show that feature selection based on feature grouping combined with the ant colony algorithm is more helpful to the establishment of subsequent classification models.In summary,the feature selection algorithm studied in this paper plays an important role in the establishment of subsequent classification and diagnosis models.It not only identifies biochemical substances with significant differences in cervical cancer lesions but also further improves the accuracy of cervical cancer screening.Using surface-enhanced Raman technology with feature selection and machine learning classification models for cervical cancer screening has broad application research prospects. |
