Analysis Of ¹⁸F-FDGPET Brain Images Based On Voxel-based Block-matching Algorithm In The Diagnosis Of Temporal Lobe Epilepsy

Background:Epilepsy,only secondly to stroke,is the second most common disease in neurological diseases.It is a chronic brain disease state with repeated seizures as a common feature.It is not a single disease entity,but has different causes and clinical manifestations.The pathophysiological basis of epilepsy is neuronal cell excitation and inhibition imbalance and lead to synchronous discharge abnormalities of neurons.The quality of life of most people with epilepsy is severely affected.The prevalence of epilepsy is 7‰according to the epidemiological survey in China,which is more than 9 million epilepsy patients in China.Patients with epilepsy generally take anti-epileptic drugs as the main treatment in clinical practice.70 to 80%of patients with epilepsy can be controlled by antiepileptic drugs.Another 20-30%of patients with epilepsy are intractable epilepsy?IE?,which is unable to effectively control seizures despite taking multiple antiepileptic drugs.Temporal lobe epilepsy?TLE?is the most common type of refractory epilepsy.An effective means of treating refractory epilepsy is surgical removal of the epileptogenic focus.Therefore,the precise location of the epileptogenic focus is the key to successful surgery.This has also become the most important research direction in clinical epilepsy.At present,a variety of functional molecular imaging techniques can provide some help information in determining the location and positioning of epileptogenic foci.Among them,Positron Emission Computed Tomography?PET?brain imaging technology is a kind of functional molecular imaging technology that can provide quantitative,localized information and non-invasive.PET brain imaging technology,which is widely used in clinical practice,is the best method currently recognized for non-invasive localization of epilepsy.The principle of PET/CT imaging for the localization of epileptogenic foci is the same as that of PET imaging,which is based on the analysis of the changes in the glucose metabolism of epileptic seizures in epileptic patients and epileptic seizures.Epilepsy epileptogenic zone episodes on PET image has high metabolism,attack epileptogenic zone interphase on the PET images show low metabolism,through computer processing of PET imaging and skull CT image fusion together,cause chronic focal areas can more clearly show it,at the same time analysis of epileptogenic zone with epilepsy brain anatomy and epileptogenic zone and brain tissue,such as language,sports,feel,and other important central area near relation,etc.,can be more accurate positioning,and let the operation more easy and convenient operation.There are two main methods to diagnose temporal lobe epilepsy using PET in clinical practice:visual analysis and Asymmetry index?AI?measurement.Although these two methods have obtained some clinical diagnostic effects,the two methods are susceptible to the level of the diagnostician during the analysis process,subjective factors have a greater impact,and the interference of human factors and directional thinking has a great influence on the analysis results,which affects the results.Credibility,it is highly desirable to find additional analytical methods to provide a useful complement to visual evaluation.In order to overcome the influence of the subjective factors of the above diagnostic methods,the researchers have developed a brain function analysis method based on voxel level,which is a pixel of brain metabolic imaging,which is fully automated and pixel-by-pixel analyzed by computer-aided diagnosis.At the same time,statistical processing was performed.There is no subjective factor in the process of the analysis process,which is completely processed by professional software.The analysis is also fully automatic,with the advantages of high stability,objectivity and reproducibility.At present,one of the hotspots and difficulties in the field of international medical image processing and analysis research is this computer-aided diagnosis,which is also the main trend of medical diagnosis in the future.In the computer-aided diagnosis process,various image processing and analysis techniques such as image segmentation,digital map creation,image registration,etc.must be involved,and image registration technology is the key to image processing and analysis technology.One of the most recognized software in the field of image registration technology is Statistical Parametric Mapping?SPM?,which is widely used in the process of normalization and statistical analysis.Statistical parameter map For the voxel-voxel analysis of ¹⁸F-FDG PET brain metabolic imaging,the image registration step must be performed with the image and the anatomical template.Conventional SPM standardization for image registration,especially in the case of extended anomaly regions,in order to minimize the number of false positive clusters,Gaussian filters are required for image filtering,which causes a certain degree of image distortion.And conventional SPM standardization has limited sensitivity in detecting low metabolizing lesions of medial temporal lobe epilepsy.At present,there is a method of image registration using Block-Matching?BM?to improve the conventional SPM standardization.In this paper,the Block-Matching algorithm is collectively called.The nonlinear transformation in this algorithm is based on:?1?The entire three-dimensional image is divided into small blocks for registration,?2?the most similar small blocks in adjacent regions of the template are identified.The whole process is performed iteratively,which improves the robustness of rigid registration of medical images.Compared to conventional SPM normalization,the Block-Matching algorithm minimizes artifacts and overcomes additional image filtering,minimizing the effects of human factors.Objective:In this study,¹⁸F-FDG PET brain metabolic imaging based on voxel BM algorithm analysis to study the localization diagnosis of patients with temporal lobe epilepsy,and with ¹⁸F-FDG PET brain metabolic imaging based on voxel conventional SPM standardization analysis.By comparing the postoperative pathology and follow-up results of the patients,the results of BM algorithm and conventional SPM standardization were compared and analyzed by statistical methods,and the diagnostic efficacy of each method was evaluated.Selection of BM algorithm methodology,optimization of relevant parameters and establishment of appropriate methods to explore and evaluate the clinical application value of BM algorithm.The significance of this study lies in the establishment and application of an appropriate and accurate BM algorithm based on voxel for the localization diagnosis of temporal lobe epileptic foci,and to obtain various parameters that should be reasonably selected when determining side diagnosis,and to explore and evaluate the clinical application value of BM algorithm in the diagnosis of temporal epileptic foci.Methods:A retrospective analysis of 106 patients from October 2013 to June 2017 at the PET/CT Center of Fuzhou General Hospital of the People's Liberation Army and clinically diagnosed as refractory temporal lobe epilepsy,including 60 males and 46females.Clinically,they have a typical history of seizures and/or EEG.The age is 4 to 54years old,with an average of 27.94±10.01 years.The medical history is 1 to 38 years,with an average of 12.43±9.41 years.The education period is 0 to 16 years.The average is13.48±3.90 years.In this study,70 patients with epilepsy were divided into medial temporal lobe epilepsy group,36 patients with lateral temporal lobe epilepsy and mixed temporal lobe epilepsy.The control group consisted of 38 healthy volunteers,including 24males and 14 females,aged 17-53 years,with an average of 35.68±7.00 years;the education period was 2-16 years,with an average of 14.34±3.79 years.The two groups of subjects matched in terms of gender,age,and education level.The original image of¹⁸F-FDG PET was obtained for all subjects.The obtained ¹⁸F-FDG PET original image was converted into an Analysis format.The analysis includes three parts:?1?Based on the voxel BM algorithm,the side-diagnosis of temporal lobe epilepsy-induced epileptogenic focus.Based on voxel analysis of ¹⁸F-FDG PET brain imaging data of each patient in the patient group compared with the healthy control group,different confidence probability p values were used to calculate and analyze the intracerebral asymmetry of each patient by one-way variance method.The diagnosis of temporal lobe epilepsy was performed,and the diagnostic accuracy of different confidence probabilities was evaluated,and the optimal confidence probability p value was selected.?2?Based on voxel conventional SPM standardization analysis,the side diagnosis of temporal lobe epilepsy-induced epileptogenic focus.Based on voxel normalization,the ¹⁸F-FDG PET brain imaging data of each patient in the patient group was compared with the healthy control group,and the asymmetry of each patient was analyzed by one-way variance method using different confidence probabilities.The diagnosis of temporal lobe epilepsy was performed,and the diagnostic accuracy of different confidence probabilities was evaluated,and the optimal confidence probability p value was selected.?3?Based on the comparison between voxel BM algorithm and conventional SPM standardization group.1.Comparing the optimal results based on the voxel BM algorithm and the optimal results based on the conventional SPM standardization method,observe whether there is a statistically significant difference between the diagnostic accuracy of the two methods.2.Based on voxel BM algorithm and conventional SPM standardization,the comparison between the medial temporal lobe epilepsy and lateral temporal lobe epilepsy and mixed type accuracy was studied.The above-mentioned temporal lobe epilepsy patients were divided into two groups,one group was 70 cases of medial temporal lobe epilepsy group,and the other group was 36 cases of lateral temporal lobe epilepsy and mixed temporal lobe epilepsy.The ¹⁸F-FDG PET images of the two groups of epilepsy patients were analyzed by the BM algorithm and the conventional SPM standardization method and using the respective optimal confidence probability p values.Difference analysis between the two groups of epilepsy patients with BM algorithm and conventional SPM standardized analysis was performed using analysis of variance.Results:?1?Analysis of ¹⁸F-FDG PET brain imaging based on voxel using block matching algorithm.One-way ANOVA was used to calculate the asymmetry of individual patients with confidence probability p<0.05 to accurately diagnose the temporal lobe epilepsy.The rate was 94.34%,which was better than other confidence probabilities,and the difference was statistically significant.?2?Based on voxel normalized SPM normalized analysis ¹⁸F-FDG PET brain imaging was performed by one-way ANOVA using different confidence probabilities p<0.05?FWE correction value?to calculate intracerebral asymmetry in a single patient for temporal lobe epilepsy The accuracy of the fixed side diagnosis was 79.25%,which was superior to other confidence probabilities,and the difference was statistically significant.?3?The comparison between the voxel-based BM algorithm and the conventional SPM standardized analysis group is as follows.1.Comparing the optimal results based on the voxel BM algorithm and the optimal results based on the voxel conventional SPM standardization method,the p value is less than 0.05,and there is a statistically significant difference between the diagnostic accuracy of the two groups.Therefore,based on the voxel BM The algorithmic approach is relatively better for the fixed-sided diagnosis of ¹⁸F-FDG PET brain imaging in patients with temporal lobe epilepsy compared to conventional SPM standardized methods.2.Based on voxel BM algorithm and conventional SPM standardization,the accuracy of the medial temporal lobe epilepsy group and lateral temporal lobe epilepsy and mixed temporal lobe epilepsy group were compared between groups.The accuracy of BM algorithm and conventional SPM standardization in the low temporal metastases of the medial temporal lobe epilepsy group reached 95.71%and 77.14%,respectively,P value was less than 0.05,the two were statistically significant;and the lateral temporal lobe epilepsy and mixed temporal lobe The detection accuracy of the low-metabolic lesions in the epilepsy group reached 91.67%and83.30%,respectively,and the P value was greater than 0.05.The two were not statistically significant.Therefore,the accuracy of the BM algorithm for detecting the low metabolizing lesions of the medial temporal lobe epilepsy is more than that of the conventional SPM standardization method.High,there is no difference between the accuracy of detecting lateral temporal lobe epilepsy and mixed epilepsy and low metabolizing lesions and conventional SPM normalization.Conclusion:Compared with conventional SPM standardization,the voxel-based BM algorithm is better for the diagnosis of ¹⁸F-FDG PET images in patients with temporal lobe epilepsy,especially in patients with medial temporal lobe epilepsy.The main reason is that the BM algorithm detects the medial side.The sensitivity of temporal lobe epilepsy with low metabolizing lesions is higher than that of conventional SPM.Compared with conventional SPM standardization,the BM algorithm performs nonlinear transformations through small blocks,which overcomes the need to process additional filtered images,minimizes artifacts and overcomes additional image filtering,and reduces image distortion to minimize artifacts.The influence of factors can improve the robustness of medical image registration,and the detection efficiency of epileptic foci in temporal lobe epilepsy is high and reproducible,suggesting a higher clinical value in the diagnosis of temporal lobe epilepsy.However,the BM algorithm still needs to conduct further research on different populations to determine its clinical use value.