| Aim to symptom of cognitive function decline of brain tumor patients detected in the current clinical, research about cognitive decline mechanism of brain tumor patients was systematically carried out based on functional magnetic resonance imaging(f MRI) technology. From the local and global function network perspective, brain function network model was constructed based on f MRI, which was used to analyze the impact of brain tumor on functional areas and explore restrictive relationship of network parameters from function network between brain tumor areas and other function areas. Impacting on patient cognitive characteristics and function network evaluation indicators were studied in order to present an evaluation model with certain brain cognitive level in clinical. Based on a large number of clinical tumor data, brain tumor grades and classification were systematicly studied by means of brain functional networks. The main work in the thesis is reproduced below.(1) The cognitive decline mechanism of brain tumor patients was explored from a local functional network perspective. Brain tumor patients and healthy group’s default network and auditory networks were exacted by using adaptive relaxation factor independent component analysis algorithm in the thesis. The results based on the default network showed that, compared with healthy group, functional connectivity strength of brain networks in the brain tumor patient group was significantly reduced in orbital frontal gyrus, left hippocampus and right cingulated, and obvious enhancement occured in the right middle frontal gyrus, left cingulate gyrus and left parietal lobe, which revealed that default mode network connectivity mode was altered in the brain tumor patient. The results further illustrated the default mode network in the brain tumor patient group was in the dysfunction. The results of auditory network showed that compared with the healthy group, functional connectivity strength of brain network for the patient group was significantly reduced in the left medial temporal lobe, the left superior temporal gyrus, the left and right temporal pole. The results indicated that functional connectivity strength in the internal auditory network of the brain tumor patients group was significantly lower than that in the healthy group. The the same method was used for the senior and low-grade brain tumor patients group’s default network and auditory network, the results showed that functional connectivity strength in the auditory networks of senior brain tumor patients was significantly enhanced in the superior temporal gyrus,as well as more obvious enhancement happened in the left parietal gyrus and right posterior cingulate gyrus in the default mode network.(2) The secondary cognitive decline mechanism of brain tumor patients induced by brain tumor was explored by using a functional global network. The small-world network(SWN) analysis method was used to study the resting state brain network topological properties, the results showed that both of patients and healthy groups had small-world properties. It is more important to find an efficiency evaluation parameter such as nodal efficiency with the assessment of cognitive decline for brain tumor patients. The results showed that nodal efficiency of brain functional network of the frontal lobe brain tumor patients was lower than that in the healthy group by two-sample t-test(P<0.05) in the thesis. In order to in-deeply analyze that brain tumor secondarily induced cognitive decline of non-tumor regions, each f MRI data related to tumor was given up and the data from normal brain tissues were retained to reconstruct brain functional network. It was found that nodal efficiency of brain functional network in the patient group in the threshold [0.1, 0.125] was significantly lower than that of brain functional network in healthy volunteer group. In a similar way of processing senior and low-grade brain tumor f MRI data, nodal efficiency of brain functional network in the new network parameters threshold [0.1, 0.125] in senior brain tumor patients was significantly higher than that of brain functional network in the low grade brain tumor patient group. Decreased nodal efficiency corresponding to the brain function areas involved into four functional areas that were the left and right supplementary motor area, the left parietal, and the right posterior cingulate. The conclusion showed that the differences between the brain tumor grades would cause secondary cognitive functions damage. Using both of independent component analysis method and the small-world network method, it was found that function connectivity strength was considerably enhanced in senior brain tumor patients compared with that in the low grade brain tumor patients. The different grade brain tumor would cause different level cognitive functional alterations, especially in the left parietal gyrus, right posterior cingulate and other functional areas.(3) The brain tumor grades classification was built by means of resting state functional brain network character. It’s noted that the average z value from the clusters of the left and right superior temporal gyrus was used as a characteristic factor to classify the different brain tumor grades, classification accuracy was reached by 87.5% in the auditory system; similarly as the z value from the clusters of the right angular gyrus and left parietal gyrus in the default network was used as a characteristic factor were used to classify the different brain tumor grades, accuracy reached by 84.375%. Small-world network parameters such as nodal efficiency and the degree values were utilized to classify and recognize different brain tumor grades, and the classification accuracy reached by 65.625%. The area(Area under the curve, AUC) under the curve corresponding to receiver operation characteristic curve(ROC) was utilized as several quantitative classifier performance indicators.(4) Specific methods for signal processing were improved based on cognitive function research of the brain tumor patients with f MRI technology.Registration methods for brain tumor f MRI data were improved. The combination method of coherence and phase coherence was proposed for brain tumor segmentation. In order to achieve regional brain tumor precise segmentation, mask method based on cost function was implemented to normalize the f MRI data from brain tumor patients. When the brain tumor area is larger, the register precision by mask method based on cost function was lower than that registration method based on the Powell algorithm for f MRI data was presented in this section. Powell method was utilized to register patients’ f MRI data to standard template space through normal brain gray image to the template image. Infomax ICA and Fast ICA algorithm were improved. By comparing with Infomax ICA algorithm, combining method of particle swarm optimization(PSO) with Newton iteration method were proposed based on particle swarm, which provides initial value, which would be provided as the initial value as Newton iteration start. This method is fully exerted PSO population search virtue and advance of partial meticulous of Newton’s method search. This combined method overcomes the inefficiencies of PSO and initial value sensitive defects of Newton iteration method. By comparing Fast ICA algorithm, a method of the adaptive relaxation factor was proposed based on independent component analysis algorithm. Sum of the square of the skewness variable was used as relaxation factor applied to improve the Fast ICA that named adaptive relaxation factor Fast ICA, which had third-order convergence speed and could reduce the computational complexity. It was also quite suitable for handling large amounts of f MRI data.Significant conclusions were obtained as follows.(1) By compared with healthy group, the default mode network occurred dysfunction, and shows that the functional connectivity strength of brain networks in the brain tumor patient group was significantly reduced in orbital frontal gyrus, left hippocampus and right cingulated, as wel as significant enhancement in the right middle frontal gyrus, left cingulate gyrus and left parietal gyrus.(2) Compared with healthy group, functional connectivity strength of auditory network in the brain tumor patient group was significantly reduced in the left medial temporal lobe, the left superior temporal gyrus and the left and right temporal pole.(3) Bothy methods of improved ICA and small world network were adopted in the senior brain tumor patients’ group that functional connectivity strength of brain function networks was significantly enhanced in the left parietal gyrus, and right posterior cingulate gyrus compared with that in low grade brain tumor patients’ group.(4) Character evaluation parameters were found and could be used to evaluate the effect of tumor on cognitive function such as functional connectivity strength(z value) and nodal efficiency. Characteristic factor(z value) was used to classify brain tumor grades and noticed that when the average z values from the clusters of the left and right superior temporal gyrus was used as a characteristic factor, which classification accuracy reached by 87.5% in the auditory system. By ROC evaluation, it’s found that the method based on the auditory network(AUC=0.87) was the best one with respect to the performance of the other methods.Preliminary assessment about brain function cognitive decline for brain tumor patients were made and meanwhile essential reference indicators were provided for the clinical surgery program, which directly to avoid producing cognitive impairment and improving neurological surgery effect. Our research will be helpful to further make f MRI instrument functional extension and expansion, as well as to provide a reference for related software development and lay the foundation for the development of new medical devices. |
PDF Full Text Request