Study On Time-Domain Diffuse Optical Tomography For Early Breast Tumor Diagnosis

Early detection is a key method for improving the curative ratio of breast cancer and reducing its mortality. Traditional X-ray mammography and ultrasonic inspection have shown some deficiencies in sensitivity, specificity, security and comfortableness. As a new noninvasive optical measurement technique, optical mammography of diffuse optical tomography (DOT) has attracted a lot of attentions for providing a richer functional image.This thesis gives an overview of existing early breast cancer inspection techniques, whereafter, proposes a method of time-domain diffuse optical tomography for early breast tumor diagnosis. Based on the sensitive connection between light and the physiological index, time-domain DOT can efficiently reconstruct the tissue optical parameter in centimeter scale by measuring the boundary flux with time resolved system.The solution procedure of finite element method (FEM) in the forward problem is simplified since the forward model is built by time-domain DOT based on generalized pulse spectrum technique (GPST). According to the Green's function and the modified GPST, the solution procedure of Jacobian matrix is accelerated in the inverse problem of DOT.To avoid the absolute system scaling, a relative data type is introduced in this paper. Two scaling methods of Jacobian matrix are proposed with a detailed theoretical derivation. The interrelated simulated and experimental validations are also revealed for two scaling methods of Jacobian matrix.The influence of mesh generation on the reconstructed image performances is studied. The mesh density and the mesh generation manner are not only associated with the iteration time but also relevant to the quantitativeness and spatial resolution of the reconstructed image. This investigation was validated by simulatied experiments.The optimization of fiber position strategy and the image segmentation strategy are proposed to improve the image quality. The effects of aboved strategies are validated by simulatied and experimental data. The optimization of fiber position strategy can be widely applied in known tumor distribution. With the introduction of this strategy, the reconstructed image perfomances can be improved by decreasing the ill-posed problem with a full use of information carried by the finite optodes. The image segmentation strategy can be widely applied in general investigation of breast. With a reduced reconstructed region and a reduced number of reconstructed variables, the underdetermined problem and ill-posed problem are well improved by the reduced Jacobian matrix.A set of experimental system of time-domain DOT is set up based on the time resolved cone-shaped compressed scanning manner. The experimental data is obtained by the strategy of optimizing fiber position, and therefore the image of solid phantom is reconstructed with two Jacobian matrix scaling methods. The reconstructed images according as the experimental data validate the algorithm of time-domain DOT that based on GPST and the image segmentation strategy.