A method and system for locating epileptic foci based on magnetic resonance imaging
By preprocessing and three-dimensional reconstruction of MRI images of children with epilepsy, and combining this with neural network identification of epileptic foci, the problem of locating MRI-negative foci has been solved, and the detection rate and localization accuracy of epileptic foci have been improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
- Filing Date
- 2023-11-23
- Publication Date
- 2026-07-14
AI Technical Summary
Current technology makes it difficult to accurately locate MRI-negative lesions in pediatric epilepsy patients using magnetic resonance imaging. Traditional methods are not good at expressing local features, resulting in poor surgical treatment outcomes.
By acquiring and preprocessing MRI images of the patient's brain, labeling focal cortical lesions, extracting morphological features, and using Freesurfer for three-dimensional reconstruction and mapping, a classifier was built using sklearn neural network to identify epileptic foci.
It improved the detection rate of lesions in MRI-negative pediatric epilepsy patients, enhanced the accuracy of epileptic lesion localization, and helped doctors to more accurately assess and surgically treat the lesions.
Smart Images
Figure CN117710635B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of artificial intelligence and medical device technology, and in particular to a method and system for locating epileptic lesions based on magnetic resonance imaging. Background Technology
[0002] Epilepsy is a neurological disorder caused by transient brain dysfunction resulting from excessive synchronous discharge of neurons. It manifests as high-frequency bursts on an electroencephalogram (EEG) and is accompanied by varying degrees of sensory or consciousness loss. Unlike single-source diseases, epilepsy is usually associated with a series of paroxysmal, transient, repetitive, and stereotyped brain dysfunction syndromes. According to WHO estimates, there are approximately 50 million people with epilepsy worldwide, and about 60% of these cases originate in childhood.
[0003] Epilepsy poses a significant threat to patients' health, increasing their psychological and physiological burden and severely reducing their quality of life. Frequent and prolonged seizures can also lead to brain damage, potentially impacting intelligence and even life expectancy.
[0004] Meanwhile, epilepsy is difficult to treat due to the complex structure of the human brain and the diverse and complex causes and pathologies of epilepsy. Approximately 70% of epilepsy patients can effectively control their seizures with regular antiepileptic drug treatment, while the remaining 30% of drug-resistant epilepsy patients require surgical treatment. Especially for children with epilepsy, the likelihood of achieving complete seizure control with antiepileptic drugs alone is very small. Common causes of drug-resistant epilepsy include focal cortical dysplasia and tuberous sclerosis. Focal cortical dysplasia, primarily caused by impaired neuronal migration or cell proliferation in the cerebral cortex, accounts for 40%-50% of children undergoing epilepsy surgery, characterized by altered cortical thickness, blurred gray-white matter boundaries, and abnormal folding.
[0005] Therefore, accurate preoperative assessment and lesion localization are crucial for the effectiveness of surgical treatment. In recent years, magnetic resonance imaging (MRI) has greatly aided preoperative assessment due to its high spatial resolution, multimodal imaging, and rapid, accurate, and non-invasive nature. Clinicians use MRI images of epilepsy patients to manually localize lesions and guide surgery. However, some patients have subtle lesions that are difficult to capture directly on MRI, referred to as MRI-negative lesions. Automated detection of these patients has become one of the biggest challenges in epilepsy clinicians. Traditional methods for localizing epilepsy lesions are ineffective for MRI-negative patients with insufficient local feature expression and are insensitive to subtle imaging abnormalities that are difficult to distinguish visually in MRI-negative epilepsy patients.
[0006] To overcome these shortcomings, this application proposes a method and system for locating epileptic lesions based on magnetic resonance imaging. Summary of the Invention
[0007] The purpose of this application is to provide a method and system for locating epileptic lesions based on magnetic resonance imaging, aiming to solve the problems of limited sample data and unclear features in the research of automated diagnostic technology for children with epilepsy.
[0008] To achieve the above objectives, this application provides the following technical solution:
[0009] This application provides a method for locating epileptic lesions based on magnetic resonance imaging, including:
[0010] MRI images of the brain of the patient to be tested were acquired and preprocessed to obtain statistical data on cortical regions;
[0011] Based on the statistical data of the cortical regions, the focal cortical lesion mask of the patient to be tested is marked;
[0012] Extract the morphological features of the lesion mask;
[0013] The morphological features are input into a classification model for identification and localization of epileptic lesions.
[0014] Furthermore, the steps of acquiring and preprocessing MRI images of the patient's brain to obtain cortical region statistics specifically include the following steps:
[0015] The preprocessing includes: deviation correction stage, subcortical structure segmentation stage, surface reconstruction stage, and calculation stage;
[0016] The MRI images were segmented, registered, and reconstructed in three dimensions using Freesurfer to generate gray-white matter segmentation planes and gray matter-cerebrospinal fluid segmentation planes. Based on the data characteristics calculated from the two segmentation planes, cortical partition statistics were obtained.
[0017] Furthermore, the step of labeling the focal cortical lesion mask of the patient to be tested based on the cortical partition statistics specifically includes the following steps:
[0018] The lesion mask was reconstructed onto the cortical surface of the individual space of the patient to be tested using Freesurfer;
[0019] The lesion mask on the individual space cortical surface is mapped to the standard space using the mni152 template to obtain the unilateral lesion mask in the standard space.
[0020] By using a bilateral symmetrical template, a unilateral lesion mask is mapped onto the contralateral hemisphere of the patient to be tested, thus obtaining a healthy control mask at the same location in the contralateral hemisphere of the patient to be tested.
[0021] The mapping method is as follows: the hemisphere of the brain surface on the side of the lesion is mirrored and flipped to the opposite side, and the flipped brain surface is registered to the opposite side surface.
[0022] Furthermore, the step of extracting the morphological features of the lesion mask specifically includes the following steps:
[0023] The morphological features include: the cortical thickness, curvature, sulcus depth, outer surface area, and gray-white matter signal intensity ratio of the vertices of the triangle covered by the lesion mask;
[0024] Extract the same features from the vertices covered by the mask of healthy controls at the same location in the contralateral hemisphere of the patient to be tested;
[0025] The morphological features are then smoothed using a Gaussian smoothing kernel.
[0026] Furthermore, the step of identifying and locating epileptic lesions after inputting the morphological features into the classification model specifically includes the following steps:
[0027] A nonlinear classifier was built using the sklearn neural network. 80% of the vertices in the lesion mask were randomly selected for training to obtain the classification model. The vertices on the affected side were marked as samples with lesion markings. The same number of vertices were extracted from the non-lesion hemisphere mask on the contralateral side as healthy controls. The remaining 20% of the vertices were used for detection.
[0028] The whole brain vertex features of the patient to be tested are input into the classification model to obtain the predicted value of the vertex; adjacent vertices with predicted values exceeding a threshold are clustered together, where the threshold is determined by calculating the Youden index on the training dataset; if the predicted vertex cluster overlaps with manually delineated lesions, it is identified as a lesion.
[0029] This application provides a magnetic resonance imaging-based epilepsy lesion localization system, comprising:
[0030] Acquisition module: Acquires and preprocesses MRI images of the patient's brain to obtain cortical region statistics;
[0031] Annotation module: Annotates the focal cortical lesion mask of the patient to be tested based on the cortical partition statistics;
[0032] Extraction module: Extracts the morphological features of the lesion mask;
[0033] Identification module: After inputting the morphological features into the classification model, it performs identification and locates the epileptic lesion.
[0034] This application provides an apparatus comprising a processor and a memory coupled to the processor, wherein the memory stores program instructions for implementing a method for locating epileptic lesions based on magnetic resonance imaging; the processor is configured to execute the program instructions stored in the memory to implement the method for locating epileptic lesions based on magnetic resonance imaging.
[0035] This application provides a storage medium storing processor-executable program instructions for executing a method for locating epileptic lesions based on magnetic resonance imaging.
[0036] This application provides a method and system for locating epileptic lesions based on magnetic resonance imaging, which has the following characteristics:
[0037] Beneficial effects:
[0038] By annotating the focal cortical lesion mask of MRI-positive epilepsy patients, and then using Freesurfer to perform three-dimensional reconstruction of the cerebral cortex surface, the lesion mask is first applied to the unilateral surface vertex and then mapped onto a bilateral symmetrical template to obtain the morphological features of the brain surface vertex covered by the lesion mask. At the same time, the same features of the contralateral control region of the lesion mask are obtained. Finally, a classification model based on the original morphological features is established and trained for identification and localization of epileptic foci. This application uses a new brain feature not used for the identification of childhood epilepsy and uses the corresponding brain region of the contralateral hemisphere of the patient's lesion as a control, which can effectively improve the accuracy of the model and improve the detection rate of MRI-negative childhood epilepsy patients. Attached Figure Description
[0039] Figure 1 This is a flowchart illustrating a method for locating epileptic lesions based on magnetic resonance imaging, according to Embodiment 1 of this application.
[0040] Figure 2 This is a flowchart of the preprocessing process in Embodiment 1 of this application;
[0041] Figure 3 This is a lesion mask and a control image of Embodiment 1 of this application;
[0042] Figure 4 This is a schematic diagram showing the gray-white matter signal intensity ratio at different locations in Embodiment 1 of this application;
[0043] Figure 5 This is the ROC curve for vertex classification in Embodiment 1 of this application;
[0044] Figure 6 This is a comparison diagram of the predicted lesion clusters and the pathologically confirmed FCD locations in Example 1 of this application;
[0045] Figure 7This is a schematic diagram of the structure of an epilepsy lesion localization system based on magnetic resonance imaging, according to Embodiment 2 of this application.
[0046] Figure 8 This is a schematic diagram of the device structure in Embodiment 3 of this application;
[0047] Figure 9 This is a schematic diagram of the storage medium structure of Embodiment 4 of this application. Detailed Implementation
[0048] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.
[0049] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0050] Example 1
[0051] Please see Figure 1 This is a flowchart illustrating a method for locating epileptic lesions based on magnetic resonance imaging, according to Embodiment 1 of this application; the steps include:
[0052] S1: Acquire MRI images of the patient's brain and preprocess them to obtain statistical data on cortical regions.
[0053] In this embodiment, the preprocessing includes: a deviation correction stage, a subcortical structure segmentation stage, a surface reconstruction stage, and a calculation stage.
[0054] Please see Figure 2 This is a flowchart of the preprocessing process in Embodiment 1 of this application.
[0055] Deviation correction stage: includes head movement correction, signal deviation correction due to magnetic field inhomogeneity, standardization of signal intensity in white matter region, and skull dissection.
[0056] Subcortical structure segmentation stage: including registration to a standard template, signal correction and intensity normalization, and white matter segmentation.
[0057] Surface reconstruction stage: includes trimming and filling, surface subdivision, smoothing and dilation, surface repair, sphere mapping, creation of gray-white cortex, and construction of pia mater surface.
[0058] The computational phase includes curvature calculation, cortical partition statistics, cortical labeling, white matter partitioning, gray-white matter ratio calculation, and voxel labeling. A triangular mesh is then used for reconstruction.
[0059] This application uses Freesurfer v5.3.0 for preprocessing of the cerebral cortex. FreeSurfer is a magnetic resonance imaging (MRI) data processing software package that can segment, register, and reconstruct three-dimensional images of high resolution MRI; the specific FreeSurfer command is recon-all-subject subjectname-i / path / to / input_volume-all-qcache.
[0060] FreeSurfer can also generate high-precision gray-white matter segmentation surfaces and gray matter-cerebrospinal fluid segmentation surfaces. Based on these two surfaces, cortical thickness and other surface data features at any location can be calculated, such as cortical outer surface area, curvature, gray matter volume, etc. These parameters can also be mapped onto the cerebral cortex surface obtained through the white matter dilation algorithm for intuitive display.
[0061] S2: Mark the focal cortical lesion mask of the patient to be tested based on the cortical partition statistics.
[0062] In this embodiment, under the guidance of a professional clinician, Freesurfer is used to reconstruct the lesion mask onto the cortical surface of the individual space of the patient to be tested;
[0063] The lesion mask on the individual space cortical surface is mapped into the standard space (fsaverage) using the mni152 template to obtain the unilateral lesion mask in the standard space;
[0064] Then, the unilateral lesion mask is mapped onto the contralateral hemisphere of the patient under test using a bilateral symmetric template (fsaverage_sym) to obtain a healthy control mask at the same position in the contralateral hemisphere of the patient under test. The vertices in the affected side and the contralateral side mask appear in pairs.
[0065] The mapping method is as follows: the hemisphere of the brain surface on the side of the lesion is mirrored and flipped to the opposite side, and the flipped brain surface is registered to the opposite side surface.
[0066] Please see Figure 3 The image shows the lesion mask and control diagram of Embodiment 1 of this application. By finding the vertices of the corresponding regions on the opposite hemisphere of the lesion mask as controls, there is less noise and better experimental results compared with other randomly selected vertices.
[0067] S3: Extract the morphological features of the lesion mask.
[0068] In this embodiment, the morphological features include: cortical thickness, area, curvature, sulcus depth, intrinsic curvature, and the ratio of gray matter signal intensity 1 mm below the gray-white matter boundary to white matter signal intensity at 10%, 30%, and 50% of the gray-white matter boundary towards the white matter surface. These features are clinical features used by epilepsy specialists to identify cortical lesions, and they also perform well in the pre-assessment of the ability to differentiate features. Furthermore, the same features are extracted from the apex covered by the mask of a healthy control at the same location in the contralateral hemisphere of the patient being tested.
[0069] The following features were smoothed using a 10mm Gaussian smoothing kernel: cortical thickness, gray-to-white matter signal intensity ratio, and all cortical and subcortical features. Intrinsic curvature and sulcus depth were smoothed using a 20mm Gaussian kernel. All smoothing kernels used were significantly smaller than the lesion mask.
[0070] The above features underwent two normalization processes: first, intra-patient Z-score normalization, which adjusted for differences in mean and standard deviation within individuals; and then inter-patient normalization, which adjusted for regional differences between individuals.
[0071] Please see Figure 4 This is a schematic diagram of the gray and white matter signal intensity ratio at different locations in Embodiment 1 of this application; this application uses multiple linear regression to regress the patient's age and gender, removing their influence on the features, and the p-values of both features are less than the default significance level of 0.05.
[0072] S4: The morphological features are input into the classification model for identification and localization of the epileptic lesion.
[0073] In this embodiment, a nonlinear classifier is built using an sklearn neural network. 80% of the vertices randomly selected from the affected and contralateral masks are used for training to obtain a classification model. The vertices on the affected side are labeled as samples with lesion markings, and the same number of vertices are extracted from the non-lesion hemispherical mask on the contralateral side as healthy controls. The remaining 20% of vertices are used for detection.
[0074] The whole brain vertex features of the patient to be tested are input into the classification model to obtain the predicted value of the vertex; adjacent vertices with predicted values exceeding a threshold are clustered together, where the threshold is determined by calculating the Youden index on the training dataset; if the predicted vertex cluster overlaps with manually delineated lesions, it is identified as a lesion.
[0075] Understandably, to verify the accuracy of the classifier in classifying positive cases, leave-one-out cross-validation was used on 13 patients with positive lesions. Specificity was then calculated in a control group consisting of an equal number of healthy children of similar sex and age. Any cortical abnormalities identified by the classifier in the control group were recorded as false positives.
[0076] Please see Figure 5 The figure shows the ROC curve for vertex classification in Embodiment 1 of this application. A nonlinear classifier was trained using 80% of the vertices randomly selected from the affected and contralateral masks of 13 patients. Vertices on the affected side were labeled as samples with lesion markers, and the same number of vertices were extracted from the contralateral non-lesion hemispherical mask as healthy controls. The remaining 20% of vertices were used for testing. In the vertex classification results, the model achieved a specificity of 0.902, a sensitivity of 0.892, and an accuracy of 0.896.
[0077] Please see Figure 6 This is a comparison image of the predicted lesion clusters and pathologically confirmed FCD locations in Embodiment 1 of this application. After obtaining the predicted values of the vertices, adjacent vertices with predicted values exceeding a specified threshold are clustered together, and clusters with an area less than 50 square millimeters are excluded as noise. In 13 MRI-positive patients, the classifier was able to detect 10 cases (sensitivity = 77%). Multiple lesion clusters were found in the remaining two patients, and no abnormal lesions were detected in the 13 healthy controls (specificity = 100%).
[0078] In summary, Embodiment 1 of this application uses a mask to annotate the focal cortical lesions of patients with positive MRI-patented epilepsy. Then, it uses Freesurfer to perform three-dimensional reconstruction of the cerebral cortex surface. The lesion mask is first applied to the unilateral surface vertices and then mapped onto a bilateral symmetrical template to obtain the morphological features of the vertices of the brain surface covered by the lesion mask. Simultaneously, it obtains the same features of the contralateral control region of the lesion mask. Finally, a classification model based on the original morphological features is established and trained for identification and localization of epileptic lesions. This improves the clinical detection rate of MRI-negative childhood epilepsy and helps doctors accurately assess and locate lesions.
[0079] Example 2
[0080] Please see Figure 7 This is a schematic diagram of the structure of an epilepsy lesion localization system based on magnetic resonance imaging according to Embodiment 2 of this application;
[0081] Acquisition module: Acquires and preprocesses MRI images of the patient's brain to obtain cortical region statistics;
[0082] Annotation module: Annotates the focal cortical lesion mask of the patient to be tested based on the cortical partition statistics;
[0083] Extraction module: Extracts the morphological features of the lesion mask;
[0084] Identification module: After inputting the morphological features into the classification model, it performs identification and locates the epileptic lesion.
[0085] In this embodiment,
[0086] In summary, Embodiment 2 of this application obtains a focal cortical lesion mask through an acquisition module and an annotation module, then uses an extraction module to obtain the morphological features of the vertices of the brain surface covered by the lesion mask, and simultaneously obtains the same features of the control region on the opposite side of the lesion mask; finally, the recognition module inputs the morphological features into the classification model for recognition and localization of the epileptic focus; this application uses a new brain feature not previously used for the identification of childhood epilepsy, and uses the corresponding brain region on the opposite side of the hemisphere where the patient's lesion is located as a control, which can effectively improve the accuracy of the model and increase the detection rate of MRI-negative childhood epilepsy patients.
[0087] Example 3
[0088] Please see Figure 8 This is a schematic diagram of the device structure in Embodiment 3 of this application. The device 50 includes a processor 51 and a memory 52 coupled to the processor 51.
[0089] The memory 52 stores program instructions for implementing the above-described method for locating epileptic lesions based on magnetic resonance imaging.
[0090] The processor 51 is used to execute program instructions stored in the memory 52 to achieve epileptic focus localization based on magnetic resonance imaging.
[0091] The processor 51 can also be referred to as a CPU (Central Processing Unit).
[0092] Processor 51 can be an integrated circuit chip with signal processing capabilities. Processor 51 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor.
[0093] Example 4
[0094] Please see Figure 9This is a schematic diagram of the storage medium in Embodiment 4 of this application. The storage medium in this embodiment stores a program file 61 capable of implementing all the above methods. This program file 61 can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or devices such as computers, servers, mobile phones, and tablets.
[0095] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.
[0096] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.
[0097] Although embodiments of this application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the appended claims and their equivalents.
[0098] Of course, the present invention may have many other embodiments. Based on this embodiment, other embodiments obtained by those skilled in the art without any creative effort are all within the scope of protection of the present invention.
Claims
1. A method for locating epileptic lesions based on magnetic resonance imaging, characterized in that, include: MRI images of the brain of the patient to be tested were acquired and preprocessed to obtain statistical data on cortical regions; Based on the statistical data of the cortical regions, the focal cortical lesion mask of the patient to be tested is marked; Extract the morphological features of the lesion mask; The morphological features are input into a classification model for identification and localization of epileptic lesions; The steps involved in acquiring and preprocessing MRI images of the patient's brain to obtain cortical region statistics include the following: The preprocessing includes: deviation correction stage, subcortical structure segmentation stage, surface reconstruction stage, and calculation stage; The MRI images were segmented, registered, and reconstructed in three dimensions using Freesurfer to generate gray-white matter segmentation planes and gray matter-cerebrospinal fluid segmentation planes. Based on the data features of the two segmentation planes, the cortical partition statistics were obtained. The step of labeling the focal cortical lesion mask of the patient to be tested based on the cortical partition statistics specifically includes the following steps: The lesion mask was reconstructed onto the cortical surface of the individual space of the patient to be tested using Freesurfer; The lesion mask on the individual space cortical surface is mapped to the standard space using the mni152 template to obtain the unilateral lesion mask in the standard space. By using a bilateral symmetrical template, a unilateral lesion mask is mapped onto the contralateral hemisphere of the patient to be tested, thus obtaining a healthy control mask at the same location in the contralateral hemisphere of the patient to be tested. The mapping method is as follows: the brain surface hemisphere on the side of the lesion is mirrored and flipped to the opposite side, and the flipped brain surface is registered to the opposite side surface; The step of extracting the morphological features of the lesion mask specifically includes the following steps: The morphological features include: the cortical thickness, curvature, sulcus depth, outer surface area, and gray-white matter signal intensity ratio of the vertices of the triangle covered by the lesion mask; Extract the same features from the vertices covered by the mask of healthy controls at the same location in the contralateral hemisphere of the patient to be tested; The morphological features are then smoothed using a Gaussian smoothing kernel. The steps of inputting the morphological features into the classification model for identification and locating epileptic lesions specifically include the following steps: A nonlinear classifier was built using sklearn neural network. 80% of the vertices in the lesion mask were randomly selected for training to obtain the classification model. The vertices on the affected side were marked as samples with lesion markings. The same number of vertices were extracted from the non-lesion hemisphere mask on the contralateral side as healthy controls. The remaining 20% of the vertices were used for detection. The whole brain vertex features of the patient to be tested are input into the classification model to obtain the predicted value of the vertex; adjacent vertices with predicted values exceeding a threshold are clustered together, where the threshold is determined by calculating the Youden index on the training dataset; if the predicted vertex cluster overlaps with manually delineated lesions, it is identified as a lesion.
2. A system for locating epileptic lesions based on magnetic resonance imaging according to claim 1, characterized in that, include: Acquisition module: Acquires and preprocesses MRI images of the patient's brain to obtain cortical region statistics; Annotation module: Annotates the focal cortical lesion mask of the patient to be tested based on the cortical partition statistics; Extraction module: Extracts the morphological features of the lesion mask; Identification module: After inputting the morphological features into the classification model, it performs identification and locates the epileptic lesion.
3. A device, characterized in that, The device includes a processor and a memory coupled to the processor, wherein the memory stores program instructions for implementing the method for locating epileptic lesions based on magnetic resonance imaging as described in claim 1; the processor is used to execute the program instructions stored in the memory to implement the location of epileptic lesions based on magnetic resonance imaging.
4. A storage medium, characterized in that, The device stores processor-executable program instructions for performing the method for locating epileptic lesions based on magnetic resonance imaging as described in claim 1.