Glioma eeg electrode positioning system based on multimodal MRI and intraoperative photograph fusion

Abstract

This invention relates to the field of medical control technology, specifically to a glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative photographs. It includes a brain imaging data acquisition module for preoperative acquisition of multimodal brain imaging data and an acquisition module for acquiring intraoperative craniotomy photographs for ECoG electrode placement. The system further aligns the multimodal brain imaging data to the same three-dimensional spatial coordinate system; then segments and corrects the multimodal brain imaging data into different regions; and establishes an individualized brain region atlas based on a healthy brain tissue texture model and segmented brain regions. Vascular structures are extracted, and a three-dimensional spatial coordinate system for vascular feature points and electrode contacts is established and transformed. The minimum Euclidean distance is calculated based on the electrode contacts, and the individualized brain region atlas is labeled. This invention addresses the problems of insufficient spatial positioning accuracy of electrode contacts and difficulty in quantifying tumor-electrode distance during surgery for glioma patients, facilitating intraoperative determination of the anatomical location of the epileptogenic focus by surgeons.

Description

Technical Field

[0001] This invention relates to the field of medical control technology, specifically to a glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative images. Background Technology

[0002] Gliomas are the most common primary intracranial tumors. Currently, most glioma patients experience comorbid epilepsy, also known as glioma-associated epilepsy. Glioma-associated epilepsy is characterized by its widespread incidence; epilepsy is the first symptom in most patients, especially those with low-grade gliomas. Secondly, glioma-associated epilepsy is drug-resistant, difficult to control, and has a high recurrence rate after surgery. Finally, due to the effects of epilepsy itself, the cognitive effects of antiepileptic drugs, and the potential interactions with chemotherapy drugs, glioma-associated epilepsy is also closely related to tumor recurrence, severely impacting patients' mental health and quality of life.

[0003] The pathogenesis of glioma-associated epilepsy is complex according to existing research and records. The epileptogenic focus may not only be directly related to the tumor itself, but may also involve the peritumoral edema area, normal brain tissue invaded by the tumor, and related neural circuits. Therefore, the core challenge in treatment lies in accurately identifying the anatomical location and network connectivity patterns of the epileptogenic focus, maximizing tumor resection while completely eliminating the epileptogenic focus, thereby reducing the postoperative recurrence rate of epilepsy and improving the patient's quality of life. Among these techniques, electrocorticography (ECoG) monitoring is a core technology for identifying the epileptogenic focus during surgery for glioma-associated epilepsy. By placing an electrode array directly on the surface of the cerebral cortex, high spatiotemporal resolution EEG signals can be captured, providing crucial data support for the localization of the epileptogenic focus and the calculation of the epileptogenic network.

[0004] However, existing ECoG electrode localization technology still faces many key challenges in epileptogenic focus localization and network computing applications, severely hindering in-depth research on treatment efficacy and mechanisms:

[0005] 1. Insufficient accuracy in locating epileptogenic foci: Traditional localization methods rely on the surgeon's experience combined with visual assessment of brain anatomical landmarks using two-dimensional intraoperative photographs. This does not fully consider the brain tissue deformation caused by the tumor's space-occupying effect and individual anatomical differences. Furthermore, brain anatomical landmarks are not clearly identifiable in all surgeries, resulting in significant spatial positioning errors of electrode contacts. This makes it impossible to accurately pinpoint the anatomical range of the epileptogenic focus, thereby affecting the accurate identification of epileptogenic network nodes.

[0006] 2. Difficulty in quantifying the spatial correlation between tumor and epileptogenic focus: The boundaries of epileptogenic focus in glioma overlap with the tumor solid and the peritumoral edema area. Traditional methods are difficult to quantify the precise distance between the electrode contact and the tumor edge, making it impossible to clarify the spatial dependence between the epileptogenic focus and the tumor, which hinders the analysis of the origin mechanism of the epileptogenic focus and targeted resection.

[0007] 3. The computation of epileptogenic networks lacks a precise anatomical basis: Existing technologies cannot provide precise coordinates of electrode contacts in individualized three-dimensional cortical models and brain region mapping information, resulting in ambiguous node localization of epileptogenic networks. This makes it difficult to conduct network connectivity analysis based on anatomical structures, and fails to reveal the network propagation path and core nodes of epileptogenic foci, thus restricting innovative mechanism research and the development of individualized treatment plans.

[0008] 4. Reliance on additional equipment and complex operation: Current relatively sophisticated positioning methods require the combination of intraoperative CT scans and expensive neuronavigation systems, which not only increases medical costs and prolongs operation time, but may also bring radiation exposure risks. In addition, the operation process is cumbersome and difficult to promote in special surgical scenarios such as primary hospitals or awake craniotomy, thus limiting its clinical applicability.

[0009] In summary, existing technologies for the treatment of glioma-associated epilepsy have significant shortcomings in terms of precise localization of epileptogenic foci, quantification of the spatial correlation between tumor and epileptogenic foci, and computation of epileptogenic networks. Developing an ECoG electrode-based precise localization system that requires no additional equipment, provides precise localization, is highly efficient, and offers a solid anatomical basis for epileptogenic focus localization and network computation has become an urgent need for clinical treatment and mechanistic research. This system is of great significance for improving the treatment efficacy of glioma-associated epilepsy and promoting innovative research on epileptogenic mechanisms. Summary of the Invention

[0010] To address the aforementioned issues, this invention provides a glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative images. This system solves the problems of insufficient spatial positioning accuracy of electrode contacts and difficulty in quantifying the tumor-electrode distance during surgery for glioma patients, facilitating intraoperative determination of the anatomical location of the epileptogenic focus by surgeons.

[0011] To achieve the above objectives, the technical solution of the present invention is as follows: a glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative photographs, comprising a brain imaging data acquisition module for acquiring preoperative multimodal brain imaging data and an acquisition module for acquiring intraoperative craniotomy photographs corresponding to the placement of ECoG electrodes during craniotomy, wherein the multimodal brain imaging data includes T1w sequence, T2w sequence, FLAIR sequence and T1ce sequence; and further comprising;

[0012] The image preprocessing module is used to convert DICOM format of brain multimodal image data into NIfTI format, and then resample and affine register the brain multimodal image data to align the brain multimodal image data to the same three-dimensional spatial coordinate system.

[0013] The automatic tumor segmentation module is used to segment brain multimodal imaging data into different regions, including the necrotic core region, the enhanced tumor region, and the peritumoral edema region. These different regions are then displayed in the same three-dimensional coordinate system, and corrected according to accurate tumor boundary standards to obtain a tumor mask. ;

[0014] The 3D modeling module for the cerebral cortex is used to input and store texture models of healthy brain tissue corresponding to normal individuals, and to mask tumors based on these healthy brain tissue texture models. Perform virtual infill repair, and mark the repaired image as... Based on By segmenting and reconstructing the cerebral cortex in three dimensions, an individualized brain region map is obtained.

[0015] The brain surface blood vessel modeling module is used to extract vascular structures based on T1ce sequences, obtain vascular response maps, and then apply thresholds. The vascular response map is binarized, and connected components are labeled in the binarized image to obtain the preoperative vascular model;

[0016] The electrode coordinate mapping module is used to perform white balance and contrast adjustment processing based on intraoperative photos. Based on the adjusted intraoperative photos and preoperative vascular models, the corresponding vascular feature points and electrode contacts are marked respectively, and the vascular feature points and electrode contacts in the intraoperative photos and preoperative vascular models are converted into a three-dimensional spatial coordinate system.

[0017] The electrode coordinate association module is used to associate electrode coordinates with electrode contact points. ,calculate The minimum Euclidean distance to the tumor boundary marker in the individualized brain region atlas is used to perform functional region labeling processing based on the minimum Euclidean distance in the set brain region atlas, thus obtaining the individualized indicator model.

[0018] Furthermore, the intraoperative craniotomy images fully cover the ECoG electrode array and the surrounding cortex, and contain at least two clearly identifiable blood vessels on the brain surface, with the images taken perpendicular to the cortical surface.

[0019] Furthermore, the image preprocessing module aligns the brain multimodal image data to the same three-dimensional spatial coordinate system by using the T1ce sequence as the reference space and aligning the T1w sequence, T2w sequence, FLAIR sequence and T1ce sequence to the same three-dimensional spatial coordinate system based on resampling and affine registration methods.

[0020] Furthermore, the necrotic core region is the necrotic area at the center of the tumor, the enhanced tumor region is the enhanced active tumor area in the T1ce sequence, and the peritumoral edema region is the edema region with high signal in the FLAIR sequence.

[0021] Furthermore, the criteria for accurate tumor boundaries include: tumor boundaries consistent with T1ce, T2w, and FLAIR sequences, all infiltrative areas in multimodal brain imaging data, and exclusion of non-tumor edema and blood vessels.

[0022] Furthermore, personalized brain partitioning atlases include gray / white matter segmentation, cortical surface reconstruction, and brain region partitioning. .

[0023] Furthermore, a brain surface blood vessel modeling module is used to extract vascular structures from T1ce sequences based on a Sato filter. The linear response of the vascular structures is... ,in yes An asymmetric exponential function;

[0024] Then normalize using the noise standard deviation and take the maximum response. The vascular response map was obtained.

[0025] Furthermore, the brain surface blood vessel modeling module, based on threshold... The steps for binarizing the vascular response map are as follows: based on the maximum response Establish ;

[0026] Clustering methods based on connected component analysis remove small noise signals, label connected components in the binarized image, and retain those with a volume larger than [missing information]. Connected components Then convert the vascular response map with connected components to NIfTI format.

[0027] Furthermore, the method for marking corresponding vascular feature points in intraoperative photographs is as follows: Mark at least two vascular intersections or bifurcations in the intraoperative photographs, and use these vascular intersections or bifurcations as registration reference points. At the same time, mark the coordinates of the four corner points of the ECoG electrode. ;

[0028] The method for marking vascular feature points corresponding to preoperative vascular models is as follows: The same vascular feature points as those in intraoperative photographs are marked on the visualized preoperative vascular model. .

[0029] Furthermore, the steps for converting vascular feature points and electrode contacts in intraoperative photographs and preoperative vascular models into a three-dimensional spatial coordinate system are as follows:

[0030] Label-based vascular feature points The method employs a similarity transformation to first project the 3D model points in the 3D model space onto the 2D plane of the intraoperative photograph, and then uses the best-fit plane. Determining the plane normal vector through principal component analysis The model points projected into the 3D model space are denoted as... ;

[0031] Compute the minimum registration error of 2D similarity transformation ,in, Scaling factor Given the rotation matrix, Procrustes analysis is used to solve for the optimal transformation parameters;

[0032] Based on minimizing the registration error Map the coordinates of the electrode contacts from intraoperative photographs to the model projection plane. Then, the 2D coordinates are mapped back to a three-dimensional, individualized brain region map through backprojection. .

[0033] Furthermore, the 3D modeling module of the cerebral cortex is also used to calculate the tumor boundary surface based on individualized brain partition maps;

[0034] The tumor boundary surface is calculated as follows, using a tumor mask. Differential diagnosis of brain regions By taking the intersection, the brain regions involved by the tumor can be determined. The labels of the affected brain regions were reassigned. Ultimately, this yields an individualized brain region map containing tumor information.

[0035] Furthermore, the electrode coordinate association module calculates the electrode contact points. The method for finding the minimum Euclidean distance to the tumor boundary in a personalized brain partition map is as follows:

[0036] Based on each electrode contact Calculate electrode contacts Minimum Euclidean distance to the tumor boundary ,in The tumor boundary surface defined in the 3D modeling module of the cerebral cortex;

[0037] Distance calculation employs a KD-Tree accelerated nearest neighbor search algorithm. First, tumor boundaries are extracted from the brain partition map to obtain a set of boundary points. A KD-Tree index structure for boundary points is constructed, and then a nearest neighbor query is performed on each electrode contact to calculate the minimum distance and record the coordinates of the corresponding nearest boundary point. Based on the nearest boundary coordinates and brain region segmentation... Based on the correlation, the functional zones corresponding to the electrode contacts are identified and marked to obtain an individualized indication model.

[0038] Furthermore, the individualized indicator model is displayed using a three-dimensional visualization interface, which displays the individualized brain region map, tumor area, preoperative vascular model, and electrode contact points with different colors and shapes.

[0039] The above approach has the following beneficial effects:

[0040] 1. This solution eliminates the need for additional equipment such as CT scans and expensive navigation systems. It can be performed using only routine intraoperative photographs and preoperative MRI data, significantly reducing the costs associated with equipment procurement, use, and maintenance, and alleviating the financial burden on medical institutions and patients. At the same time, the technical operation process is simple, requiring only the taking of intraoperative photographs and the marking of feature points to proceed with subsequent work. It is easy to promote and apply in neurosurgical clinical departments at all levels and has extremely high clinical application value.

[0041] 2. This approach, based on the stable anatomical landmark of blood vessels on the brain surface, reduces localization errors and provides accurate anatomical references for key surgical procedures such as tumor resection and epileptogenic focus localization, ensuring the safety and effectiveness of the surgery. Furthermore, the image processing and registration process can be completed preoperatively, minimizing the time required for intraoperative image and model registration. This fully meets the needs of real-time intraoperative decision-making, effectively shortens the operation time, improves overall surgical efficiency, and lays the foundation for rapid postoperative recovery for patients.

[0042] 3. This solution integrates multiple core functions such as tumor segmentation, cortical modeling, vascular modeling, electrode localization, and distance calculation. It is applicable to various neurosurgical patients requiring ECoG monitoring and has a wide range of clinical applications. While helping to achieve maximum tumor resection, complete resection of epileptogenic foci, and improve patient prognosis, its precise localization capability also provides important support for in-depth research on the mechanisms of neurological diseases such as epilepsy networks. It can promote basic research and clinical translation in related fields and achieve significant mechanism innovation and technological breakthroughs. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the process of an embodiment of the ECoG electrode localization system for gliomas based on the fusion of multimodal MRI and intraoperative images of the present invention;

[0044] Figure 2 This is a schematic diagram of preoperative multimodal brain images and tumor mask in an embodiment of the glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative photographs of the present invention;

[0045] Figure 3 This is a schematic diagram of a pre-repair image in an embodiment of the ECoG electrode localization system for gliomas based on the fusion of multimodal MRI and intraoperative photographs of the present invention;

[0046] Figure 4This is a schematic diagram of the repaired image in an embodiment of the ECoG electrode localization system for gliomas based on the fusion of multimodal MRI and intraoperative photographs of the present invention;

[0047] Figure 5 This is a schematic diagram of an individualized brain region map in an embodiment of the glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative photographs of the present invention;

[0048] Figure 6 This is a schematic diagram of a preoperative vascular model in an embodiment of the glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative photographs of the present invention;

[0049] Figure 7 This is a schematic diagram of vascular feature points and electrode contact point markings in an intraoperative photograph of the glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to the present invention.

[0050] Figure 8 This is a schematic diagram of the registration of vascular feature points and electrode contact point markers in a three-dimensional spatial coordinate system in an embodiment of the glioma ECoG electrode localization system based on the fusion of multimodal MRI and intraoperative images of the present invention.

[0051] Figure 9 This is a schematic diagram of an individualized indicator model in an embodiment of the ECoG electrode localization system for gliomas based on the fusion of multimodal MRI and intraoperative images of the present invention. Detailed Implementation

[0052] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] The following detailed description illustrates the specific implementation method:

[0054] Example 1:

[0055] As attached Figure 1The system, based on the fusion of multimodal MRI and intraoperative photographs, is a glioma ECoG electrode localization system. It includes a brain imaging data acquisition module for preoperative acquisition of multimodal brain imaging data and an acquisition module for obtaining intraoperative craniotomy photographs corresponding to the placement of the ECoG electrodes. The multimodal brain imaging data includes T1w sequences (T1-weighted images), T2w sequences (T2-weighted images), FLAIR sequences (fluid-attenuated inversion recovery sequences), and T1ce sequences (T1-contrast-enhanced images). The intraoperative craniotomy photographs fully cover the ECoG electrode array and the surrounding cortex, and include at least two clearly identifiable surface blood vessels (arteries or veins). The photographs are taken perpendicular to the cortical surface to avoid excessive tilting for visualization and parameter calculation.

[0056] It also includes an image preprocessing module, an automatic tumor segmentation module, a 3D modeling module for the cerebral cortex, a brain surface blood vessel modeling module, an electrode coordinate mapping module, and an electrode coordinate association module.

[0057] The image preprocessing module converts DICOM format of brain multimodal image data to NIfTI format, then resamples and performs affine registration on the brain multimodal image data to align it to the same three-dimensional spatial coordinate system. The method for aligning the brain multimodal image data to the same three-dimensional spatial coordinate system is as follows: using the T1ce sequence as the reference space, the T1w, T2w, FLAIR, and T1ce sequences are aligned to the same three-dimensional spatial coordinate system based on resampling and affine registration.

[0058] The automatic tumor segmentation module is used to segment different brain multimodal imaging data into different regions, including a necrotic core region, an enhanced tumor region, and a peritumoral edema region. The necrotic core region is the central necrotic area of ​​the tumor, the enhanced tumor region is the enhanced active area of ​​the tumor in the T1ce sequence, and the peritumoral edema region is the high-signal edema region in the FLAIR sequence. These different regions are then displayed in the same three-dimensional coordinate system, and corrected according to accurate tumor boundary standards to obtain a tumor mask. .

[0059] The criteria for accurate tumor boundaries include: tumor boundaries consistent with T1ce, T2w, and FLAIR sequences, all infiltrative areas in multimodal brain imaging data, and exclusion of non-tumor edema and blood vessels.

[0060] For example, such as Figure 2As shown, by acquiring four MRI sequences (T1w, T2w, FLAIR, and T1ce) preoperatively, the morphological characteristics, tissue boundaries, and infiltration range of the tumor can be fully captured from different imaging dimensions, providing comprehensive imaging data support for subsequent segmentation. The tumor was automatically segmented using the 3D U-Net model, and the tumor volume was 106.9 cm³ after manual correction. This not only improved the segmentation efficiency by utilizing intelligent algorithms but also avoided segmentation deviations in complex lesion areas through manual intervention, ensuring the accuracy of tumor volume measurement and boundary localization. This provides a reliable quantitative basis for preoperative tumor grading assessment and surgical planning.

[0061] The FastSurfer-LIT method was used to repair lesions, and FastSurfer was used to reconstruct the cortex, displaying tumor involvement in the frontal lobe in zonal areas. This effectively solved the problems of cortical structural deformation and blurred boundaries caused by tumor space-occupying or infiltration, clearly showing the spatial relationship between the tumor and surrounding normal cortical tissue. This helps surgeons intuitively judge the extent of tumor invasion and the proximity of functional areas, reducing the risk of intraoperative damage to normal brain tissue and providing crucial anatomical references for maximizing safe tumor resection. The Sato filter was used to extract blood vessels on the brain surface, successfully modeling major arteries and veins, clearly restoring the distribution and direction of blood vessels on the brain surface and their spatial proximity to the tumor. On the one hand, the complete vascular model can assist surgeons in planning the optimal surgical approach preoperatively, avoiding important vascular structures, and reducing the risk of intraoperative bleeding; on the other hand, this vascular model can serve as a stable anatomical landmark for subsequent intraoperative image registration, laying the foundation for high-precision localization and ensuring the safety and accuracy of surgical procedures.

[0062] The 3D modeling module for the cerebral cortex is used to input and store texture models of healthy brain tissue corresponding to normal individuals, and to mask tumors based on these healthy brain tissue texture models. Perform virtual infill repair, and mark the repaired image as... Combining such Figure 3 , Figure 4 As shown. Further based on By performing cortical segmentation and 3D reconstruction, an individualized brain region atlas is obtained, combined with, for example... Figure 5 As shown.

[0063] Personalized brain partitioning includes gray / white matter segmentation, cortical surface reconstruction (white matter surface and pia mater surface), and brain region partitioning. Brain regionalization Based on the existing Desikan-Killiany map, there are a total of 68 cortical regions.

[0064] For example, a texture model of healthy brain tissue from a normal individual, encompassing the natural texture features, anatomical morphology, and spatial distribution patterns of brain tissue, provides a highly physiologically accurate reference template for virtual filling and repair of tumor areas. Based on this healthy model, tumor masking can be applied. Image obtained after virtual infill repair It can accurately restore the original normal brain tissue texture and anatomical structure of the tumor-occupying area, effectively solving the problems of brain tissue morphological loss and structural deformation caused by tumor invasion. At the same time, it avoids the defects such as texture distortion and anatomical structure misalignment that are easy to occur in traditional repair methods. The generated repair images not only conform to the physiological characteristics of brain tissue, but also have significant individual attributes, laying a high-quality imaging foundation for subsequent cortical segmentation and reconstruction.

[0065] The generated personalized brain atlas can intuitively reflect the positional relationship between the tumor and cortical functional areas and cerebral blood vessels. Combined with the navigational role of stable cerebral vascular anatomical landmarks, it can assist surgeons in accurately planning the surgical approach preoperatively, maximizing the safe resection of the tumor during surgery, effectively avoiding damage to important functional areas and blood vessels, reducing the risk of postoperative complications such as hemorrhage and neurological deficits, and significantly improving patient prognosis. Furthermore, this technique does not rely on additional expensive equipment; it can be completed through routine image processing, making it simple to operate and cost-effective. The generated personalized atlas can also be used for postoperative efficacy evaluation and long-term follow-up, providing important support for the precision diagnosis and treatment and mechanistic research of neurosurgical diseases.

[0066] The brain surface blood vessel modeling module is used to extract vascular structures from T1ce sequences using a Sato filter based on the Hessian matrix. The linear response of the vascular structures is... ,in yes An asymmetric exponential function; then normalized using the noise standard deviation and taking the maximum response. This yields a vascular response map. Then, based on a threshold... (Usually 0.1-0.2) Binarize the vascular response map, and label the connected components of the binarized image to obtain the preoperative vascular model.

[0067] The binarization steps are as follows, based on the maximum response. Establish ;

[0068] Then, clustering methods based on connected component analysis are used to remove small noise signals, and connected component labeling is performed on the binarized image to retain those with a volume larger than [missing information]. (In this embodiment, the connected component is set to 100 mm³) Then convert the vascular response map with connected components to NIfTI format.

[0069] For example, based on the fixed anatomical location and constant morphological structure of cerebral blood vessels, and relying on the high-resolution imaging advantage of T1ce sequences for vascular tissue, the morphological features of the main trunks and branches of cerebral blood vessels can be accurately captured. By extracting vascular response maps, the grayscale differences between blood vessels and surrounding brain tissue and tumor tissue can be effectively distinguished, solving the problem of insufficient vascular recognition in conventional MRI sequences. Subsequently, by setting a threshold θ to binarize the vascular response maps, the boundary features of vascular structures can be further enhanced, background noise and artifact interference can be filtered out, and the integrity and purity of the vascular region can be ensured. Then, by analyzing the binarized images using a connected component labeling algorithm, a continuous vascular network structure can be automatically identified and constructed. The resulting preoperative vascular model can completely restore the course, branch distribution and spatial morphology of cerebral blood vessels, fully matching the standardized course of the main trunks of cerebral blood vessels and the predictable branch patterns, providing a high-precision vascular anatomy reference for preoperative planning.

[0070] Meanwhile, this modeling process fully leverages the feature preservation advantage of cerebral blood vessels in disease states. Even if blood vessels are slightly displaced due to tumor compression, the preoperative vascular model can still clearly display key structures such as the initiation segment and bifurcation of the blood vessels. Combined with precise processing of threshold binarization and connected component labeling, these identifiable features can be fully preserved in the model, ensuring that the vascular model still has reliable reference value in complex scenarios such as tumor invasion and brain tissue deformation, effectively supporting the accurate judgment of the spatial relationship between tumors and blood vessels during surgery.

[0071] Furthermore, this modeling method does not rely on CT scans or expensive angiography equipment. The vascular model can be constructed using only routine preoperative T1ce sequences, significantly reducing equipment costs and the examination burden on patients. Moreover, the entire processing can be automated through algorithms, eliminating the need for complex manual operations. The generated vascular model can be directly registered with intraoperative images, fully leveraging the navigational role of cerebral blood vessels as stable anatomical landmarks. While improving positioning accuracy, it significantly simplifies clinical procedures and is easier to promote and apply in medical institutions at all levels, providing a practical and feasible technical solution for the precision and safety of neurosurgery.

[0072] The electrode coordinate mapping module is used for white balance and contrast adjustment based on intraoperative images, combined with, for example... Figure 6 As shown, based on the adjusted intraoperative photographs and preoperative vascular model, the corresponding vascular feature points and electrode contacts are marked respectively, combined with, for example... Figure 7 As shown.

[0073] The method for marking corresponding vascular feature points in intraoperative photographs is as follows: Mark at least two (3-4 recommended) significant vascular intersections or bifurcations in the intraoperative photographs, and use these vascular intersections or bifurcations as registration reference points. Simultaneously mark the coordinates of the four corner points or more electrode contacts of the ECoG electrode. .

[0074] The method for marking vascular feature points corresponding to preoperative vascular models is as follows: The same vascular feature points as those in intraoperative photographs are marked on the visualized preoperative vascular model. .

[0075] The specific steps for converting vascular feature points and electrode contacts from intraoperative photographs and preoperative vascular models into a three-dimensional spatial coordinate system are as follows: Based on labeled vascular feature points By employing a similarity transformation (2D to 2D projection), the 3D model points in the 3D model space are first projected onto the 2D plane of the intraoperative photograph, and then the best-fit plane is used. The plane normal vector is determined by principal component analysis (PCA). The model points projected into the 3D model space are denoted as... .

[0076] Calculate the minimum registration error for 2D similarity transformations (including rotation, scaling, and translation). ,in, Scaling factor Given the rotation matrix, Procrustes analysis is used to solve for the optimal transformation parameters;

[0077] Based on minimizing the registration error Map the coordinates of the electrode contacts from intraoperative photographs to the model projection plane. Then, the 2D coordinates are mapped back to a three-dimensional, individualized brain region map through backprojection. Combining such Figure 8 As shown.

[0078] For example, white balance adjustment performed on intraoperative images can effectively eliminate image color cast caused by differences in the color temperature of surgical lighting, restore the true color information of blood vessels, brain tissue and electrode contacts on the brain surface, and avoid feature recognition deviations caused by color distortion; while contrast adjustment can enhance the grayscale difference between blood vessels and surrounding brain tissue, making the course outline, branching morphology of blood vessels and edge features of electrode contacts clearer and more distinguishable, solving the problem of feature blurring caused by factors such as light reflection and tissue occlusion in intraoperative images, and providing a high-quality image foundation for subsequent feature point marking.

[0079] By combining intraoperative photographs with a preoperatively constructed high-precision vascular model for corresponding marking, the advantages of fixed anatomical location and constant morphological structure of cerebral blood vessels can be leveraged to achieve precise matching between intraoperative vascular feature points and preoperative model feature points, significantly reducing the subjective error of manual marking. Simultaneously, the electrode contact marking completed at the same time can directly establish the spatial correspondence between electrode positions and cerebral blood vessels and cortical functional areas, providing a direct reference for subsequent verification of electrode positioning accuracy, ensuring the accuracy of intraoperative image registration, and thus stabilizing the positioning error within 2mm.

[0080] The electrode coordinate association module is used to associate electrode coordinates with electrode contact points. ,calculate The minimum Euclidean distance to the tumor boundary marker in the individualized brain region atlas is used to perform functional region labeling processing based on the minimum Euclidean distance in the set brain region atlas, thus obtaining the individualized indicator model.

[0081] Among them, the 3D modeling module of the cerebral cortex is also used to calculate the tumor boundary surface based on the individualized brain partition map;

[0082] The tumor boundary surface is calculated as follows, using a tumor mask. Differential diagnosis of brain regions By taking the intersection, the brain regions involved by the tumor can be determined. The labels of the affected brain regions were reassigned. Ultimately, this yields an individualized brain region map containing tumor information.

[0083] The method for minimizing Euclidean distance is as follows:

[0084] Based on each electrode contact Calculate electrode contacts Minimum Euclidean distance to the tumor boundary ,in The tumor boundary surface defined in the 3D modeling module of the cerebral cortex;

[0085] The minimum Euclidean distance calculation employs a KD-Tree accelerated nearest neighbor search algorithm. First, tumor boundaries are extracted from the brain partition map to obtain a set of boundary points. A KD-Tree index structure for boundary points is constructed, and then a nearest neighbor query is performed on each electrode contact to calculate the minimum distance and record the coordinates of the corresponding nearest boundary point. Based on the nearest boundary coordinates and brain region segmentation... The association between electrode contacts is determined, and the corresponding functional regions are labeled to obtain an individualized indication model. This maps each electrode contact to a Desikan-Killiany brain atlas, identifying its corresponding functional region. Combining such Figure 9 As shown.

[0086] For example, the minimum Euclidean distance can intuitively reflect the spatial proximity between the electrode contact and the tumor tissue as a quantitative indicator, avoiding the subjective bias of traditional methods that rely on visual observation or experience. This enables precise quantification of the relative relationship between the electrode position and the tumor boundary. This quantitative calculation mode fully integrates the results of preoperative reconstructive imaging and cortical reconstruction, ensuring the authenticity of the spatial reference benchmark for distance calculation. It provides accurate data support for subsequent functional zoning marking and further improves the reliability of surgical positioning.

[0087] By substituting the minimum Euclidean distance parameter into a pre-defined brain region atlas for functional zoning, the quantitative data of electrode localization can be deeply coupled with the distribution of brain functional areas, constructing an individualized indicator model that fits the individual lesion characteristics and brain functional anatomy of the patient. This model can not only clearly distinguish the spatial relationship between the tumor-involved area, the electrode coverage area, and adjacent brain functional areas, but also accurately identify the distance threshold between the functional area and the tumor boundary. This provides a crucial basis for surgeons to formulate differentiated surgical plans, ensuring the goal of maximizing tumor resection while effectively avoiding damage to key functional areas such as motor and language functions, significantly improving the individualization and safety of surgical plans.

[0088] Ultimately, the personalized indicator model can directly serve intraoperative real-time decision-making, providing surgeons with intuitive spatial references, assisting them in dynamically adjusting surgical strategies, reducing the risk of intraoperative neurological damage, and improving postoperative prognosis. Simultaneously, the quantitative distance data and functional zoning labeling information contained in the personalized indicator model can serve as an important data source for basic research such as epilepsy networks and brain functional connectivity, providing detailed personalized data support for revealing the pathogenesis of neurological diseases and developing novel diagnostic and treatment technologies, thus achieving synergistic advancement of clinical application and basic research.

[0089] In the individualized indicator model, an interactive 3D visualization interface was built using the BrainTRACE open-source toolbox based on MATLAB. This interface displays the individualized 3D model of the cerebral cortex (gray semi-transparent), the tumor region (yellow solid), the blood vessels on the brain surface (red tubular structures), the ECoG electrode array (spherical markers), and the electrode numbers. The electrode numbers are sequentially numbered from right to left in the same row, specifically 1-6, 7-12, 13-18, and 19-24.

[0090] The three-dimensional coordinate parameters of the electrode number are shown in the table below:

[0091] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion, comprising a brain imaging data acquisition module for preoperative acquisition of multimodal brain imaging data and an acquisition module for obtaining intraoperative craniotomy photographs corresponding to the placement of ECoG electrodes, wherein, The brain multimodal imaging data includes T1w, T2w, FLAIR, and T1ce sequences; its characteristic is that it also includes; The image preprocessing module is used to convert DICOM format of brain multimodal image data into NIfTI format, and then resample and affine register the brain multimodal image data to align the brain multimodal image data to the same three-dimensional spatial coordinate system. The automatic tumor segmentation module is used to segment brain multimodal imaging data into different regions, including the necrotic core region, the enhanced tumor region, and the peritumoral edema region. These different regions are then displayed in the same three-dimensional coordinate system, and corrected according to accurate tumor boundary standards to obtain a tumor mask. ; The 3D modeling module for the cerebral cortex is used to input and store texture models of healthy brain tissue corresponding to normal individuals, and to mask tumors based on these healthy brain tissue texture models. Perform virtual infill repair, and mark the repaired image as... Based on By segmenting and reconstructing the cerebral cortex in three dimensions, an individualized brain region map is obtained. The brain surface blood vessel modeling module is used to extract vascular structures based on T1ce sequences, obtain vascular response maps, and then apply thresholds. The vascular response map is binarized, and connected components are labeled in the binarized image to obtain the preoperative vascular model; The electrode coordinate mapping module is used to perform white balance and contrast adjustment processing based on intraoperative photos. Based on the adjusted intraoperative photos and preoperative vascular models, the corresponding vascular feature points and electrode contacts are marked respectively, and the vascular feature points and electrode contacts in the intraoperative photos and preoperative vascular models are converted into a three-dimensional spatial coordinate system. The electrode coordinate association module is used to associate electrode coordinates with electrode contact points. ,calculate The minimum Euclidean distance to the tumor boundary marker in the individualized brain region atlas is used to perform functional region labeling processing based on the minimum Euclidean distance in the set brain region atlas, thus obtaining the individualized indicator model.

2. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion as described in claim 1, characterized in that, Intraoperative craniotomy images include complete coverage of the ECoG electrode array and surrounding cortex, and contain at least two clearly identifiable brain surface vessels. The images are taken at an angle perpendicular to the cortical surface.

3. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 2, characterized in that, The image preprocessing module aligns brain multimodal image data to the same three-dimensional spatial coordinate system by using the T1ce sequence as the reference space and aligning the T1w, T2w, FLAIR, and T1ce sequences to the same three-dimensional spatial coordinate system based on resampling and affine registration methods.

4. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 3, characterized in that, The necrotic core region is the necrotic area at the center of the tumor, the enhanced tumor region is the active tumor area that enhances in the T1ce sequence, and the peritumoral edema region is the edema region with high signal in the FLAIR sequence.

5. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 4, characterized in that, Accurate criteria for tumor boundaries include: tumor boundaries consistent with T1ce, T2w, and FLAIR sequences, all infiltrative areas in multimodal brain imaging data, and exclusion of non-tumor edema and blood vessels.

6. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 5, characterized in that, Personalized brain partitioning includes gray / white matter segmentation, cortical surface reconstruction, and brain region partitioning. .

7. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 6, characterized in that, The brain surface blood vessel modeling module is used to extract vascular structures from T1ce sequences based on Sato filters. The linear response of the vascular structures is: ,in yes An asymmetric exponential function; Then normalize using the noise standard deviation and take the maximum response. The vascular response map was obtained.

8. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 7, characterized in that, The brain surface blood vessel modeling module is based on thresholds. The steps for binarizing the vascular response map are as follows: based on the maximum response Establish ; Clustering methods based on connected component analysis remove small noise signals, label connected components in the binarized image, and retain those with a volume larger than [missing information]. Connected components Then convert the vascular response map with connected components to NIfTI format.

9. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 8, characterized in that, The method for marking corresponding vascular feature points in intraoperative photographs is as follows: Mark at least two vascular intersections or bifurcations in the intraoperative photographs, and use the vascular intersections or bifurcations as registration reference points. At the same time, mark the coordinates of the four corner points of the ECoG electrode. ; The method for marking vascular feature points corresponding to preoperative vascular models is as follows: The same vascular feature points as those in intraoperative photographs are marked on the visualized preoperative vascular model. .

10. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 9, characterized in that, The steps for converting vascular feature points and electrode contacts in intraoperative photographs and preoperative vascular models into a three-dimensional spatial coordinate system are as follows: Label-based vascular feature points The method employs a similarity transformation to first project the 3D model points in the 3D model space onto the 2D plane of the intraoperative photograph, and then uses the best-fit plane. Determining the plane normal vector through principal component analysis The model points projected into the 3D model space are denoted as... ; Compute the minimum registration error of 2D similarity transformation ,in, Scaling factor Given the rotation matrix, Procrustes analysis is used to solve for the optimal transformation parameters; Based on minimizing the registration error Map the coordinates of the electrode contacts from intraoperative photographs to the model projection plane. Then, the 2D coordinates are mapped back to a three-dimensional, individualized brain region map through backprojection. .

11. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 10, characterized in that, The 3D modeling module for the cerebral cortex is also used to calculate the tumor boundary surface based on individualized brain partition maps; The tumor boundary surface is calculated as follows, using a tumor mask. Differential diagnosis of brain regions By taking the intersection, the brain regions involved by the tumor can be determined. The labels of the affected brain regions were reassigned. Ultimately, this yields an individualized brain region map containing tumor information.

12. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 11, characterized in that, Electrode coordinate association module calculates electrode contact points The method for finding the minimum Euclidean distance to the tumor boundary in a personalized brain partition map is as follows: Based on each electrode contact Calculate electrode contacts Minimum Euclidean distance to the tumor boundary ,in The tumor boundary surface defined in the 3D modeling module of the cerebral cortex; The minimum Euclidean distance calculation employs a KD-Tree accelerated nearest neighbor search algorithm. First, tumor boundaries are extracted from the brain partition map to obtain a set of boundary points. A KD-Tree index structure for boundary points is constructed, and then a nearest neighbor query is performed on each electrode contact to calculate the minimum distance and record the coordinates of the corresponding nearest boundary point. Based on the nearest boundary coordinates and brain region segmentation... Based on the correlation, the functional zones corresponding to the electrode contacts are identified and marked to obtain an individualized indication model.

13. The glioma ECoG electrode localization system based on multimodal MRI and intraoperative photograph fusion according to claim 12, characterized in that, The individualized indicator model is displayed using a three-dimensional visualization interface, which marks the individualized brain region map, tumor area, preoperative vascular model and electrode contact points with different colors and shapes.

Images