Research On The Processing And Identification Techniques Of Spontaneous Raman Fingerprint Spectra In Acute Leukemia Cells

Raman spectroscopy is a fingerprint spectrum of molecular vibration,which can indicate the biological molecular changes of cancerous cells,and its label-free and non-destructive properties make it have a broad application prospect in the clinical identification of single-cell canceration.The spontaneous Raman spectrum signal is weak,and it is easily interfered by a variety of spectral lines in the detection of clinical cancer cells,which seriously affects the identification and analysis of the fingerprint spectrum of cells.This paper studies the methods of removing noise,substrate spectrum and fluorescence background in Raman spectra clinical cancer cells,as well as the classification and identification of fingerprint spectra,and realizes the identification of clinical acute leukemia samples through Raman spectroscopy.The details are as follows.(1)The physical mechanism and characteristics of several main interference lines in Raman spectra of clinical samples were studied.Aiming at the spikes in the spectrum,a spike removal method based on morphological operations is proposed.This method is simple to operate and can remove spikes in any spectrum.In order to eliminate the substrate spectrum that is common in the Raman spectrum of clinical sample,a specific-scale analysis algorithm based on wavelet transform is proposed.It can realize the linear separation of the substrate spectrum through multiresolution analysis.Fluorescence spectra are commonly found in the Raman spectra of biological samples,and are characterized by a smooth curve with slow changes.We propose a fluorescence estimation method based on cubic spline interpolation fitting,and then develop an automatic fluorescence estimation method combining local minimum points with zero-order Savitzky-Golay filter.Both methods can achieve fluorescence removal.(2)The classification and recognition methods of Raman spectroscopy are studied.The spectra of gastric cancer cell lines and leukemia cell lines were classified by principal component analysis(PCA)combined with linear discriminant analysis(LDA).Through the analysis of the classification process,the reliability and importance of the pre-processing algorithm in spectral recognition are proved.Aiming at the deficiency of principal component analysis(PCA)and linear discriminant analysis(LDA)in complex sample recognition,a spectral feature amplification method based on random forest algorithm was proposed.In this method,spectral parameter features were extracted from multiple random forests,and the original spectrum was amplified according to the extracted features.The spectral data after feature amplification can significantly improve the classification accuracy of PCA-LDA algorithm under the nonlinear boundary of a variety of samples,and greatly expand the application of PCA-LDA algorithm.(3)The label-free and rapid detection of spontaneous Raman spectroscopy in clinical acute leukemia was studied.Through the processing and identification of Raman spectra in clinical white blood cell samples,the accurate distinction between normal white blood cells and two acute leukemia white blood cells was realized,and the accuracy of identification was 98.62%.By comparing the characteristic peaks of Raman spectrum of different samples,it was found that the peaks of 673 cm^-1,728 cm^-1,782 cm^-1,1483 cm^-1 and 1577 cm^-1 characterized by nucleic acid in acute leukemia were all stronger than those of normal cells,while the peaks of lipids were lower than those of normal leukocytes.The identification results of Raman spectroscopy on a variety of leukemia indicate that Raman fingerprint identification technology of cancer cells has important reference value for accurate and rapid diagnosis and treatment of clinical leukemia.