Photoacoustic Signals Analysis Methods For In Vivo Photoacoustic Flow Cytometry

In this master’s subject,student optimized the photoacoustic signals analysis methods for in vivo photoacoustic flow cytometry.The theoretical principle of in vivo photoacoustic flow cytometry is the photoacoustic effect of biological tissues.When living tissues are irradiated with specific laser,target cells would absorb the energy of electromagnetic waves,causing local temperature rise and volume expansion,and finally the ultrasonic wave are emitted.The PAFC(In Vivo Photoacoustic Flow Cytometry,PAFC)we built realize the routine in vitro detection and the in vivo detection of target cells,meanwhile ensures the noninvasive and realtime detection.Melanoma is a malignant tumor,which is common in the circulation system in the period of tumor metastatic formation.The PAFC we built can detect the circulating melanoma cells in the circulation system.Based on the result of detection,analysis for the photoacoustic signals about the features can be made,thus realizing the early diagnosis.Currently the most common analysis methods for photoacoustic signals includes threshold analysis in time domain,full-width-half-maximum method,fast Fourier transform(FFT),power spectrum analysis method and so on.However in the in vivo experiments,the photoacoustic signals acquired from target cells are hard to be sifted out because of the noise pollution.In this master’s project,optimized methods are introduced based on Pearson correlation coefficient,signal features like FWHM supplemented with machine learning analysis to solve the problem the true signals were buried in the noises.The wave form of photoacoustic signals in time domain have similar form of vibration,based on which Pearson correlation coefficient method can be applied to sift away the noises having high amplitudes acquired by threshold analysis.The method based on signal features like FWHM can be used to filter the true photoacoustic signals in time domain furthermore.Finally combined with machine learning methods,we can get the best result relatively.The methods included in the thesis pioneer the idea of using the features of target cells’ PA signals to filter and classify the unknown single period signals,which greatly improve the precision and efficiency of threshold method in time domain.Meanwhile the excavation of features of PA signals is meaningful to the study and exploration of features of target cells and medium.By applying the data analysis methods in the thesis to the in vivo PA signals,we reached more than 85% accuracy of detection of target cells’ PA signals,while ensure the processing speed about 1000 period per second on our lab’s work station.The analysis methods mentioned above are based on the existence of complete true photoacoustic signals’ database of target cells,on the foundation of which we can apply these methods for in vivo detection of circulating melanoma cells,thus realizing the early detection and diagnosis of melanoma.