Inverse Model Of Cellular Nanostructure Based On The Low-coherence Phase Microscopy

The cellular nanostructure parameters can effectively reflect the subtle changes ofcellular structure so it can be used for early diagnosis of cancer. Studies have shown thatthe system parameters acquired by the low-coherence phase microscopy can effectivelyreflect the cellular nanostructure parameters without trauma. However, there is anon-linear relationship between them, so we need to establish inverse model to get thecellular nanostructure parameters through the cellular system parameters and help torealize early diagnosis of cancer. Therefore, the author studies variety of intelligentinformation processing methods and establish inverse models by the training sample.The main contents of this thesis are as follows:â‘ The author firstly studied the basic principles and design principles of the partialleast squares regression (PLS), BP neural network and least squares support vectormachine (LS-SVM); then proposed the inverse models of the cellular nanostructureparameters that based on these three methods; after that, organized many experiments totest the performance, compared and evaluated these three kinds of inverse models.â‘¡The author used the particle swarm optimization algorithm into the method ofleast squares support vector machine to optimize the parameters of the least squaresSVM and improve the retrieval performance of the SVM. So a new inverse model basedon particle swarm optimization and least squares support vector machine(PSO-LS-SVM) was proposed. In addition, the thesis tested the validity of the inversemodel and analyzed the experimental results concretely.The study of this paper provided a new and effective method for detecting thecellular nanostructure through the low-coherence phase microscope and provided a newtechnical method for early diagnosis of cancer. Therefore, the study has importanttheoretical and practical significance.