| In recent years, the brain functional network which can be constructed by functional connectivity has been studied in all aspects because it can directly show the functional mechanism of the brain activity. These studies are based on the rapid development of brain imaging techniques and different methods to calculate functional connectivity in the brain. The study results have been applied in many fields, especially in the pathologic diagnosis of various neurologic diseases. In this paper, we used partial coherency to get functional connectivity in MEG. The analysis results were used in the diagnosis of epilepsy.The MEG recordings contain crosstalk among different channels and are easily affected by brain volume conductivity. Partial coherency analysis can lower the interference of the crosstalk. Meanwhile, some researchers proved that the imaginary part of coherency is not affected by volume conductivity. Therefore, the imaginary part of partial coherency may represent a more accurate functional connectivity in MEG and in this paper, we used it to analyze the simulated brain signals to prove its effectiveness.There are three ways to get the partial coherency:direct method, iterative method and inversion method. The direct method isn’t commonly used in practical calculation due to its calculation error. The iterative method is usually applied to analyze low-order partial coherency. The inversion method is quick and convenient but it relies on the condition of the spectrum matrix. In this paper, we found that sometimes, when the spectrum matrix was in ill condition, the analysis results of the inversion method were wrong while the analysis of the iterative method could still get the correct results. However, on the other hand, the iterative method is not fit to calculate the high order coherency because of the limitations of the memory and speed in a regular computer. Due to this, in this paper, we optimized the computing procedure of the iterative method. The new procedure can greatly reduce the time and memory cost of the computer so that we can use the optimized iterative method to calculate high-order partial coherency.At the end of the paper, we analyzed the functional connectivity of275-ch MEG signals based upon the optimized iterative method. Functional network was built using the analysis results. Comparing the functional network of the healthy people and people with epilepsy, we found significant functional connection in the patients’ network at some frequencies and we thought that this abnormal connectivity was related to the epilepsy focus. |
