Preliminary Study On Internal Information Of The Skin Tissue Based On Multi-position Scattering Spectroscopy

ObjectiveIn order to eliminate the superficial interference of biological tissue and achieve the high precision measurement and analysis method of spectral analysis on skin tissue’s internal information, this study built a multi-position scattering spectrum collection device and designed a new data analysis method which called BP neural network combine with M+N theory. A mimicking phantom experiment and an excised animal tissue experiment are designed to proof the feasibility of multi-position scattering spectral method to analyze the internal information of skin tissue.Methods1. According to the transmission path of light in tissue of the "banana model" theory, this research designs a multi-position scattering spectrum acquisition system which includes micro displacement platform, spectrometer, computer, and based on MCU and Lab VIEW software to achieve automatically control of the acquisition system.2. A mimicking phantom experiment is designed. This phantom model includes two layers:the upper layer is pellicle and the bottom layer is milk solution. The milk solution concentration will be calculated after multi-position scattering spectral acquisition and data analysis of BP neural network combine with the "M+N" theory.3. An excised animal tissue experiment is designed. In the experiment, five pieces of pork are injected of different volume of water in subcutaneous fat layers. The volume of water will be calculated after multi-position scattering spectral acquisition and data analysis of BP neural network combine with the "M+N" theory.Results1. Automatic control of multi-position scattering spectrum system is achieved, and the stability of system is good. The modulated light source system decreases the influence of ambient light and dark current noise in measuring process, and improves the accuracy of measurement. 2. In milk-membrane phantom experiment, we make a comparison of the single position spectral data and multi-position spectral data, the common BP neural network modeling and BP neural network modeling combine with "M+N" theory. The results show that the predict accuracy of multi-position scattering spectral data with BP neural network modeling combine with "M+N" theory is highest, reaching more than95%.3. In excised animal experiment, the predict accuracy of multi-position data modeling is higher than the single point position data modeling. The result shows that multi-position scattering spectral method can effectively predict the volume of water in subcutaneous fat layers of the pork.Conclusions1. Through the phantom experiment and excised animal experiment, the results conclude that multi-position scattering spectroscopy can cover more internal information of tissue, and improve the measurement accuracy.2.The BP neural network combine with "M+N" theory modeling method has much higher accuracy than the general BP neural network.