Research On MRI-guided Intracranial Hematoma Monitoring Algorithm

In view of the fact that the existing clinical imaging methods and intracranial monitoring equipment are not suitable for rapid,continuous,real-time,non-invasive and accurate monitoring of intracranial hematomas,this topic has carried out research on the real-time monitoring of intracranial hematomas.This study combines near-infrared light technology,imaging methods and deep neural network algorithms,and uses magnetic resonance imaging(MRI)of the patient’s head to provide prior information on the location and size of intracranial hematomas,and continuous detection using near-infrared light technology The method achieves real-time monitoring of intracranial hematoma,and deep neural network realizes the reconstruction of hematoma image.The main research contents of the paper include:(1)The establishment of an optical model guided by MRI brain images.The introduction of MRI brain images can overcome individual differences.Based on the image,the brain tissue can be segmented and the patient’s brain structure information can be obtained to establish an optical model of individual tissue.The optical model can provide a priori information of the location and shape of the hematoma,and has the characteristic that the tissue structure and the shape of the hematoma fit the real brain.(2)Research on the amplification method of artificial data set of intracranial hematoma.The training process of neural networks requires a large amount of data,and it is difficult to obtain a large amount of learning and training data for individuals in clinical practice.This topic proposes to use simulation methods to construct large data sets,that is,to achieve reasonable amplification of sample data sets by extracting and scaling the hematoma boundary.(3)Research on the applicability of regular and irregular hematoma imaging based on intracranial hematoma image reconstruction method.In this subject,simulation experiments are carried out on the reconstruction effect of regular and irregular hematoma images and the network anti-noise.The experiment shows that the image reconstruction method of this subject has high accuracy and adaptability in the reconstruction of different hematoma morphology.(4)3D image reconstruction algorithm of hematoma based on deep neural network.In this study,stacked auto-encoder(SAE)deep neural network is used for image reconstruction and the research of network anti-noise.The experiment shows that the curve fitting R~2 value of the regular hematoma reconstruction volume and the real hematoma volume under different noise conditions is over 0.95.Aiming at the problem of poor noise immunity of irregular hematoma networks,this study proposes a multi-level noise network establishment method,which reduces the average reconstruction volume error by 82%and effectively improves the noise immunity of the network.(5)The simulation experiment verifies the feasibility of the intracranial hematoma monitoring algorithm.Using polyoxymethylene and fat emulsion solutions to prepare optical phantoms,the transmitted light intensity information data of corresponding points are collected as the prediction set of SAE deep neural network.The experiment concluded that the deep neural network based on SAE can realize the accurate prediction of hematoma change trend.Through simulation experiment and simulation experiment research and analysis,this subject concludes that the intracranial hematoma monitoring method in this study can realize fast,continuous,real-time,non-invasive and accurate monitoring of intracranial hematoma,and provide an effective auxiliary diagnosis basis for the treatment of intracranial hematoma.