Research On Detection And Localization Of Intractable Epilepsy Based On Radiomics Analysis

Epilepsy is a common chronic brain disease.Currently,nearly 30%of patients with epilepsy have no reliable cure.This type of epilepsy is called intractable epilepsy.For this type of intractable epilepsy,the main effective treatment method is surgical epilepsy resection.Therefore,during the treatment of clinical intractable epilepsy,an accurate diagnosis of preoperative epilepsy lesions is crucial for the formulation of surgical plans and subsequent diagnosis and treatment.The traditional detection methods of intractable epilepsy lesions usually perform morphological analysis on brain images based on the clinical diagnosis knowledge from professional doctors.In recent years,there is another radiomics analysis method gradually applying in the field of medical imaging research.By extracting a series of semi-quantitative or quantitative features on medical images,radiomics quantifies the morphology,texture and other features on a large number of images,and based on these features,establishes a series of clinical diagnosis models for different diseases,efficacy evaluation models,etc.Various studies have shown that radiomics features can be used as MRI imaging biomarkers in tumor classification tasks and applied in tasks such as brain glioma detection and lung cancer detection,but have not been deeply explored in the field of epilepsy lesions.Based on radiomics analysis,this study proposes detection and localization methods for several refractory epilepsy lesions,and analyzes the correlation between radiomics features and several types of epilepsy lesions by processing and analyzing MRI images.The result of detection model can assist doctors in formulating treatment plans.This thesis firstly focuses on learning the related study and research on radiomics analysis methods and machine learning,and the detection of hippocampal sclerosis lesions was set to be the initial research in the whole experiment.In the hippocampal sclerosis detection experiment,a combination of wavelet filtering and Gaussian filtering algorithms were proposed to extract radiomic features.By comparing the evaluation indicators of machine learning classification models constructed with different features,the radiomic features and classification models were evaluated,and the ability of different radiomic features to characterize epilepsy lesions were analyzed.Then,based on four machine learning models,a multi-model fusion classification algorithm is proposed,and the algorithm is evaluated by comparing the evaluation indicators of different classification models.Then,based on the above experimental results,this thesis transferred the experimental method to other kinds of intractable epilepsy lesion localization tasks,and designed the temporal lobe epilepsy lesion localization experiment and the frontal cortical dysplasia lesion localization experiment.In the task of localizing the epileptogenic side of temporal lobe epilepsy,a method for locating the epileptogenic side of temporal lobe epilepsy based on the extraction of radiomic features from the temporal lobe cortex was proposed to verify that the radiomic features based on cortical extraction had the ability to represent temporal lobe epilepsy lesions,and analytical assessments of these radiomic features were performed.Focusing on the results of feature analysis,the thesis continued to design research on the localization of dysplasia in the frontal cortex,and proposed a method for localizing dysplasia in the frontal cortex based on the radiomic features extracted from the sub-regions of the frontal cortex,so as to achieve a more specific localization of epilepsy lesions.This thesis confirms that the radiomics analysis method can be applied to the detection task of hippocampal sclerosis,and finds several types of radiomics features that are sensitive to epilepsy lesions.Through the detection results of hippocampal sclerosis,the experiment further verified the feasibility of applying radiomics method to temporal lobe epilepsy and frontal cortical dysplasia,more specific location of the lesion can be detected using radiomics.The results can assist doctors in the effective diagnosis and treatment of some of intractable epilepsy.