Replacement MRI construction method based on three-dimensional head scanning and related device

Abstract

The invention discloses an alternative MRI (Magnetic Resonance Imaging) construction method based on three-dimensional head scanning and a related device, and the method comprises the steps: carrying out the preprocessing of an image of an MRI database, obtaining head MRI data, and constructing an MRI head three-dimensional model; acquiring three-dimensional head scanning data, and performing point cloud registration and shape similarity matching by using the three-dimensional head scanning data and the MRI head three-dimensional model; and selecting the MRI head three-dimensional model with the highest matching degree with the three-dimensional head scanning data as a substitution model according to the point cloud registration and shape similarity matching result. The method has the advantages of being convenient to operate, low in cost, low in requirement for subjects and the like, a set of automatic retrieval and substitution generation process from three-dimensional scanning to MRI is constructed, the method can be directly used for TI electrode layout, simulation and stimulation navigation, and therefore the application range of the TI technology in clinical and scientific research environments is remarkably widened.

Description

Technical Field

[0001] This invention belongs to the field of neuromodulation and medical image processing technology, and relates to an alternative MRI construction method and related devices based on three-dimensional head scanning. Background Technology

[0002] Temporal Interference (TI) is an emerging non-invasive neuromodulation technique that selectively modulates deep brain structures by applying two high-frequency alternating currents to the scalp surface and creating a difference-frequency envelope electric field deep within the brain. Because it can generate deep neural stimulation without punctures or tissue damage, TI shows great potential for applications in brain disease treatment, functional area localization, and neuroscience research. However, the electric field distribution of TI is highly dependent on a precise model of the individual's head structure; therefore, individual MRI scans are typically required to establish conductivity models and calculate the electric field distribution before performing TI.

[0003] In practical applications, obtaining MRI scans is limited by high costs, long processing times, and the fact that some individuals cannot access MRI equipment (such as those with implanted metal parts, claustrophobic patients, and pregnant women). Furthermore, MRI imaging may be completely unavailable in primary healthcare facilities or research environments. Therefore, the inability to obtain MRI scans severely impacts the accessibility of TI technology and limits its widespread adoption in clinical and research settings.

[0004] Current research on MRI-free conditions mainly focuses on using universal head models or standard templates (such as the MNI template) to replace individual MRI. However, universal models cannot reflect the significant differences in head shape between individuals, leading to significant errors in TI electric field calculations, which in turn shifts the peak position of the electric field, affecting the effectiveness and safety of stimulation. In addition, there are research directions that use statistical shape models or deep learning methods to estimate individual head structure, but these methods rely on large amounts of labeled data or fine meshes, limiting their practical generalizability.

[0005] With the development of 3D scanning technology, it has become possible to quickly acquire the 3D shape of the head using structured light or depth cameras, obtaining high-precision head point cloud data within seconds. However, a mature technical solution is still lacking for how to efficiently and accurately match the scan results to an MRI database and find the closest MRI head structure for a given individual, thereby constructing an alternative MRI model in the absence of an MRI. Existing 3D point cloud registration methods (such as ICP) suffer from problems such as unstable matching, low computational efficiency, and difficulty in guaranteeing registration accuracy when the head shape is complex, the pose is inconsistent, or there is a lot of data noise. In addition, many methods used for biomedical image registration cannot be directly used for matching sparse point clouds with volume data, and therefore cannot be directly applied to the pre-model construction for TI electric field calculations.

[0006] Therefore, there is an urgent need for a method that can generate individual alternative MRIs by efficiently matching three-dimensional head scans with MRI databases in the absence of MRI, in order to support personalized modeling and navigation of temporal interference stimuli. Summary of the Invention

[0007] The purpose of this invention is to provide an alternative MRI construction method and related device based on three-dimensional head scanning, which solves the problem that existing time-interference stimulus individualized modeling is highly dependent on MRI and cannot perform deep electric field calculations without MRI.

[0008] To achieve the above objectives, the present invention employs the following technical solution: An alternative MRI construction method based on 3D head scanning includes: Images from the MRI database were preprocessed to obtain head MRI data, and a three-dimensional MRI head model was constructed. Acquire 3D head scan data, and use the 3D head scan data to perform point cloud registration and shape similarity matching with the MRI head 3D model; Based on the point cloud registration and shape similarity matching results, the MRI head 3D model with the highest matching degree to the 3D head scan data was selected as the alternative model.

[0009] Furthermore, image preprocessing of the MRI database includes binarization of MRI images, hole filling of MRI images, boundary extraction of MRI images, and segmentation of MRI images.

[0010] Further, the binarization process of MRI images includes: Find a threshold based on the grayscale of MRI images to maximize the inter-class variance between the foreground and background; After performing automatic thresholding on the MRI image, regions with voxel values ​​higher than the threshold are set to 1, and regions with voxel values ​​lower than the threshold are set to 0, thereby obtaining the binarized result of the image; Hole filling in MRI images involves filling the background area completely surrounded by the foreground along the X, Y, and Z axes of each two-dimensional slice after binarization, thereby eliminating internal holes. Boundary extraction in MRI images includes: Perform dilation and erosion operations on the binary image of the head, respectively; After completing the expansion and corrosion processes, perform an XOR operation on the results of the two processes, extract the difference between the expanded and corroded regions, and obtain the corresponding head boundary region. MRI image segmentation includes: Browse MRI slices in three directions—axial, sagittal, and coronal—using a visual image user interface; Based on the selected direction, specific slice locations are selected for cropping, retaining the MRI data above the target area and removing the rest.

[0011] Furthermore, the inter-class variance between the foreground and background is:

[0012]

[0013]

[0014] In the formula, For example, the inter-class variance between foreground and background. For the threshold, and These represent the pixel ratios of the background and foreground, respectively. grayscale value Histogram probability, The number of gray levels. and These are the average gray values ​​for the background and foreground, respectively.

[0015] Furthermore, methods for constructing 3D models of the MRI head include: Isosurfaces are extracted from the preprocessed MRI data. By traversing the voxel units in the volume data, the position of the isosurface in the voxel is determined by linear interpolation based on the relationship between the voxel vertices and a set threshold, and the three-dimensional voxel data is converted into a continuous triangular mesh model.

[0016] Further methods for point cloud registration and shape similarity matching include: Uniform sampling is performed on the surface of the MRI head 3D model, and the data is converted into point cloud data. The iterative nearest point algorithm is used to rigidly register the point cloud. The rotation and translation parameters are iteratively optimized to make the source point cloud and the target point cloud coincide in space. Calculate the Chamfer distance between the head point cloud from the 3D head scan data and the head point cloud generated from each MRI image, and perform a traversal comparison of all 3D head models from MRI.

[0017] Furthermore, the MRI head 3D model with the highest matching degree to the 3D head scan data is the MRI head 3D model with the smallest Chamfer distance, which is:

[0018] In the formula, and This represents two sets of 3D point clouds that need to be compared. Represents a set Any point in, Represents a set Any point in, and Representing sets respectively and set The total number of midpoints and Point and points The square of the Euclidean distance between them.

[0019] An alternative MRI reconstruction system based on three-dimensional head scanning, comprising: The preprocessing module is used to preprocess images from the MRI database to obtain head MRI data and construct a three-dimensional MRI head model. The matching module is used to acquire 3D head scan data and perform point cloud registration and shape similarity matching with the 3D head MRI model. The selection module is used to select the MRI head 3D model with the highest matching degree with the 3D head scan data as the alternative model based on the point cloud registration and shape similarity matching results.

[0020] A terminal device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method.

[0021] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method.

[0022] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method for constructing alternative MRI based on 3D head scanning. First, images from an MRI database are preprocessed to obtain standardized 3D MRI head models. Then, individual head shape information is acquired using 3D scanning. The subject's scan model is matched with point clouds of multimodal head models in the database to retrieve the MRI most closely matching the individual's head morphology, thus constructing an individual alternative MRI model. This allows for personalized navigation and electric field simulation for temporal interference stimuli (TI) even in the absence of an MRI, achieving approximately 90% matching accuracy and ensuring high consistency of the alternative MRI at the head shape level. This invention proposes an alternative MRI generation method for TI navigation in the absence of individual MRI scans. By acquiring a 3D scan model of the subject's head and comparing its shape similarity with 3D head models modeled in an MRI database, the method automatically selects the most closely matching alternative MRI, thereby obtaining a head structure model suitable for TI electric field simulation and navigation without the need for an individual MRI scan. Compared to traditional methods relying on individual MRI scans, this invention, for the first time, combines 3D scanning technology with MRI database retrieval. Only a 3D head scan is needed to select an alternative MRI, avoiding the problems of long MRI scan times, high equipment requirements, and subject limitations. This enables the construction of an alternative MRI model for time-interference stimulation electric field calculation even without an MRI, significantly improving the accessibility and clinical application efficiency of TI navigation. Compared to using universal templates or standard head models, this invention can select the MRI with the most similar morphology for each individual, significantly improving the rationality of TI electric field simulation. Furthermore, this invention has been validated on a large head database, showing that the shape matching degree between the alternative MRI and the individual's head scan is significantly higher than that of the universal template model, reaching approximately 88% structural similarity, thus ensuring that the electric field simulation error is within an acceptable range. Compared to the traditional "one model fits everyone" approach, this invention minimizes electric field deviation caused by structural differences. In addition, this invention has advantages such as ease of operation, low cost, and fewer requirements for subjects. It constructs an automated retrieval and alternative generation process from 3D scan to MRI, which can be directly used for TI electrode layout, simulation, and stimulation navigation. By reducing the high cost and barriers to individual MRI acquisition, TI technology can be extended to resource-constrained medical environments, populations that cannot receive MRI, or research scenarios that require rapid modeling. It is applicable to neuromodulation scenarios such as TI surgical planning and TI treatment, thereby significantly improving the applicability of TI technology in clinical and research environments.

[0023] Furthermore, this invention employs point cloud registration and Chamfer distance to achieve quantitative matching between head 3D scans and MRI models, thereby improving the accuracy of alternative MRI selection.

[0024] Furthermore, this invention enables rapid comparison between head scans and MRI databases through three-dimensional point cloud preprocessing, standardized registration, and shape matching and optimization algorithms based on distance metrics; and by constructing alternative MRI models, TI applications no longer rely on expensive, time-consuming, or difficult-to-obtain individual MRI scans, significantly improving the accessibility and scalability of temporal interference stimuli.

[0025] Furthermore, this invention performs spatial alignment on the scanned head point cloud model and achieves rigid registration between the scanned point cloud and the outer surface of each MRI head model through an iterative nearest-point algorithm. This allows point clouds from different sources to be evaluated for similarity in the same coordinate system, thereby improving the accuracy of Chamfer distance calculation.

[0026] Furthermore, the present invention performs region cropping on the scanned point cloud, including removing the portion below the earlobe and the chin region, retaining only the top of the head to the face region required for matching the MRI model, so that the comparison area remains consistent and the matching stability is improved.

[0027] Furthermore, this invention imposes a point count limit on the outer surface point cloud extracted during MRI modeling. By setting a downsampling strategy of num_points=5000, a balance is achieved between computational cost and shape fidelity, thereby improving matching efficiency.

[0028] Furthermore, this invention uses Chamfer distance as a point cloud similarity evaluation index. By calculating the bidirectional nearest neighbor distance between the scanned point cloud and the MRI model point cloud one by one, the matching results have advantages such as robustness to noise, simple calculation, and applicability to sparse scanning. Attached Figure Description

[0029] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart of the alternative MRI construction method based on three-dimensional head scanning according to the present invention.

[0031] Figure 2 This document provides a detailed dataset, research method, and research flowchart for this invention.

[0032] Figure 3 This is a flowchart of the most similar sample finding method of the present invention.

[0033] Figure 4 This is a flowchart illustrating the verification process of the most similar sample finding method of the present invention.

[0034] Figure 5 This is a demonstration diagram of the image segmentation GUI interface of the present invention.

[0035] Figure 6 This is a schematic diagram of a preferred embodiment of the present invention, which is an alternative MRI reconstruction system based on three-dimensional head scanning.

[0036] Figure 7 This is a schematic diagram of the electronic device structure according to a preferred embodiment of the present invention. Detailed Implementation

[0037] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0038] Obviously, the described embodiments are only some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0039] It should be noted that the terminals involved in the embodiments of this application may include, but are not limited to, mobile phones, personal digital assistants (PDAs), wireless handheld devices, tablet computers, personal computers (PCs), MP3 players, MP4 players, wearable devices (e.g., smart glasses, smartwatches, smart bracelets), smart home devices, and other smart devices.

[0040] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0041] The present invention will now be described in further detail with reference to the accompanying drawings: See Figure 1This invention provides a method for constructing an alternative MRI based on 3D head scanning. It utilizes 3D head scan data and an MRI database for point cloud registration and shape similarity matching. Even without an MRI, it automatically selects the MRI image from the database that most closely resembles the subject's head shape, generating an alternative MRI image suitable for TI simulation and treatment. This invention completes the selection of the alternative model through image preprocessing, 3D modeling, database retrieval of the most similar sample, and verification. This eliminates the reliance on individual MRI images for personalized navigation, improving the accessibility and scalability of TI technology.

[0042] The complete process of this invention from 3D scan input to alternative MRI output mainly includes: MRI image binarization, MRI image filling, MRI image boundary determination, MRI image segmentation, 3D modeling, distance metric calculation, and database traversal matching. By calculating the scan point cloud and the MRI outer surface point cloud one by one, this invention can quantify the shape differences of different MRI models and finally select the MRI with the highest matching degree with the scan point cloud as the alternative model. Based on this alternative MRI, subsequent time-interference stimulation electric field modeling and stimulation parameter design can be carried out. Figure 2 As shown, the specific steps include: Step 1, MRI image preprocessing. This invention uses Python or MATLAB to preprocess the images, and MRIcroGL software is used to view the images when viewing them.

[0043] Step 1.1, Binarization of MRI Images. MRI generates images based on the magnetic resonance signals of hydrogen nuclei (mainly protons in water molecules) in human tissues. Different tissues (such as gray matter, white matter, cerebrospinal fluid, fat, and bone) will have different signal intensities due to differences in proton density, T1 relaxation time, T2 relaxation time, and magnetic susceptibility, which will manifest as differences in grayscale values. The differences in proton density and magnetic susceptibility are determined by the characteristics of the tissue itself (for example, the content of water molecules and fat in the tissue varies; cerebrospinal fluid (CSF) has a high water content, a high proton density, and a stronger magnetic resonance signal; bone has a low water content, a low proton density, and a weaker magnetic resonance signal). Different tissues respond differently to magnetic fields (e.g., iron-containing blood or calcified tissue can affect the homogeneity of the local magnetic field, thus affecting signal intensity), while T1 relaxation time and T2 relaxation time are determined by the different imaging sequences during MRI scans (e.g., T1-weighted imaging emphasizes T1 relaxation differences, with high (bright) fat signal and low (dark) cerebrospinal fluid signal, often used to observe anatomical structures; T2-weighted imaging emphasizes T2 relaxation differences, with high (bright) cerebrospinal fluid signal and moderate fat signal, often used to detect edema or lesions).

[0044] For the reasons mentioned above, head MRI images typically exhibit different grayscale values. In this invention, the focus is on the outline shape of the head (skull, skin). The brain tissue inside the head will affect the results in subsequent modeling and registration steps. Therefore, it is necessary to use a certain method to remove the internal brain tissue and retain only the outer outline of the head. First, the MRI image is binarized: The data portion of the .nii format MRI file used in this invention is a three-dimensional array storing voxels, so Python's NumPy arrays can be used for processing during MRI image preprocessing.

[0045] The Otsu method is used to find a threshold based on the grayscale of the MRI image. This maximizes the inter-class variance between the foreground (voxels above a threshold) and the background (voxels below a threshold). The formula for inter-class variance is: (1) In the formula, and These represent the pixel proportions (weights) of the background and foreground, respectively, where: (2) In the formula, grayscale value Histogram probability, The number of gray levels.

[0046] and These are the average grayscale values ​​for the background and foreground, respectively, where: (3) Iterate through all possible thresholds to find the one that makes Maximum threshold .

[0047] After using the Otsu algorithm to perform automatic thresholding segmentation on MRI images, voxel values ​​higher than the threshold are... The region is set to 1, below The region is set to 0 to obtain the binarized image result. However, in actual observation, it was found that directly using the threshold automatically calculated by Otsu causes the original MRI image boundary to shrink, i.e., the edges shift inward. This problem mainly stems from the assumption of Otsu's method, that the image's gray-level histogram exhibits a bimodal distribution, which is only applicable to images containing two types of tissue. However, the brain's tissue structure is complex, and its gray-level distribution is not a strictly bimodal pattern. Therefore, the global threshold calculated by Otsu... It may be too high, which could lead to the shrinkage of the boundary.

[0048] To alleviate the above problems, the initial threshold was adjusted a second time. Through comparison of multiple sets of experiments, it was found that when the threshold calculated by Otsu was adjusted... Scale to approximately At the optimal threshold, the boundary effect of the binarization result is most ideal. If the threshold is set too high, the boundary region is easily affected by external stray points in the MRI image, resulting in boundary expansion; while if the threshold is set too low, it may further exacerbate the inward shrinkage of the boundary. The binarized image obtained after adjustment can better preserve the edge information of the original tissue structure.

[0049] Step 1.2, Hole Filling in MRI Images: To accurately extract the external contour structure of the head in MRI images, it is necessary to remove as much internal brain tissue as possible, retaining only the external envelope structure of the head. In early processing, each MRI slice was treated as an independent two-dimensional image, identifying closed regions representing the head, and assuming that the largest connected region within these regions corresponds to the brain tissue to be removed. Based on this assumption, the largest connected region was found in each slice, and all voxel values ​​within that region were set to zero, thereby achieving the effect of removing the internal brain structure while preserving the external head structure.

[0050] However, in practice, this method has been found to have significant limitations and cannot achieve ideal processing results. Specifically, the processed image still retains a large amount of internal structure. The problem mainly stems from the following aspects: First, MRI images themselves have a three-dimensional spatial structure, while this method relies solely on the analysis of individual two-dimensional slices. This 2D-based processing ignores the spatial continuity and structural consistency between adjacent slices, causing regions that are connected in three-dimensional space to be misidentified as disconnected or independent in two-dimensional slices.

[0051] Secondly, due to variations in tissue signal intensity, imaging artifacts, and the complexity of anatomical structures in actual MRI images, the head region in a single slice is often not a completely closed connected region. For example, brain tissue (such as gray and white matter) may appear as multiple isolated parts due to the presence of low-signal areas (regions that have been segmented into 0 or 1 in binarization) such as ventricles, cerebrospinal fluid, or blood vessels. Furthermore, gaps or signal interruptions between bones and skin may also cause the skull contour to appear as multiple separate regions in a two-dimensional image. These situations undermine the assumption that "the largest connected region is the target region," making it impossible to accurately remove brain tissue through simple two-dimensional connected component analysis.

[0052] To address the aforementioned issues, a novel strategy was employed: the entire MRI volume was filled with holes to merge all connected parts into a single, larger structure, and this was combined with other methods to preserve the head's contour.

[0053] Hole filling refers to the process of filling in background areas (usually with a value of 0, called "holes") completely surrounded by foreground (usually a value of 1) in a binary image, thereby eliminating internal voids. In the preceding processing steps, the binarized MRI image has been obtained. Then, hole filling is performed on each two-dimensional slice along the X, Y, and Z axes. This multi-axial filling strategy maximally repairs voids caused by discontinuities in tissue structure or signal loss, allowing previously scattered areas to form a more complete and unified structure in three-dimensional space.

[0054] Step 1.3: After binarizing and filling the three-dimensional holes in the MRI head image, a head region with complete connectivity and eliminated internal holes is obtained. Based on this binarized three-dimensional volume data, this invention uses a morphological boundary extraction method to determine the external contour of the head.

[0055] A small, symmetrical 3D structuring element is selected, and dilation and erosion operations are performed on the binary image of the head. The dilation operation expands the foreground region, causing the head boundary to expand outward, thereby compensating for the discontinuity of the local boundary; the erosion operation shrinks the foreground region, causing the head boundary to shrink inward, and suppressing any isolated noise regions that may exist.

[0056] After completing the expansion and corrosion processes, an XOR operation is performed on the results to extract the difference between the expanded and corroded regions, yielding the corresponding head boundary region. This boundary accurately reflects the external envelope morphology of the head.

[0057] Since the preceding processing has ensured the connectivity and integrity of the head region in three-dimensional space, the morphological boundary extraction method used can stably obtain a continuous and clear outer contour of the head, providing a reliable boundary basis for subsequent head shape reconstruction, contour matching and related applications.

[0058] Step 1.4: After boundary extraction of the MRI image, a complete external head contour has been obtained. The ultimate goal of this invention is to perform three-dimensional modeling based on the processed MRI image to obtain a three-dimensional head model, and to calculate the similarity between this model and the head model obtained through three-dimensional scanning. This provides an alternative to MRI data for preoperative target localization in temporal interferometry (TI) therapy in the absence of MRI scanning conditions.

[0059] In actual scanning, MRI images often contain non-target areas such as the neck and shoulders, which can interfere with subsequent 3D head modeling and similarity calculations. Therefore, before performing 3D modeling and similarity comparison, it is necessary to segment the MRI images, retaining only the head portions related to the brain regions.

[0060] To address the issue of inconsistent head positions in MRI images due to differences in height, body position, and scanning conditions among different subjects, this invention employs a segmentation method based on human interaction. A visual image user interface (GUI) was developed in a Python environment, such as... Figure 5 As shown, this interface is used for browsing and segmenting 3D MRI data. It supports browsing MRI slices in three directions: axial, sagittal, and coronal. Specific slice locations can be selected based on the chosen direction for cropping, and the segmented results are finally saved as new NIFTI format files.

[0061] In this invention, segmentation is preferably performed from the axial plane. The user determines the cropping position by dragging the slice selection control. Once the selection is complete, the cropping operation is performed, retaining the MRI data above the target area and removing the rest.

[0062] Specifically, the segmentation location was selected as the MRI slice corresponding to the first appearance of the earlobe structure in the axial section. This location has relatively stable anatomical features across different subjects, providing a clear basis and good repeatability for the segmentation process. This effectively removes the neck and lower regions, retaining only the head MRI image from the top of the head to the earlobe, providing a more accurate and consistent data foundation for subsequent 3D modeling and similarity calculation.

[0063] Step 2: After segmenting the MRI images, this invention obtains MRI data containing only the area from the top of the subject's head to the earlobe. This data effectively focuses on head structures related to brain regions, reducing interference from non-target areas below the neck on subsequent modeling and similarity calculations. Based on the segmentation results, a 3D model of the MRI data is constructed to create a corresponding 3D head model, which is then compared in shape similarity with the head model obtained from the 3D scan.

[0064] This invention utilizes a Python environment to perform 3D modeling of MRI volumetric data. The Marching Cubes algorithm is employed to extract isosurfaces from the preprocessed MRI data, converting the 3D voxel data into a continuous triangular mesh model. This algorithm traverses the voxel cells in the volumetric data and, based on the relationship between voxel vertices and a set threshold, uses linear interpolation to determine the position of the isosurfaces within the voxels, thereby generating a triangular mesh structure that approximates the outer surface of the head.

[0065] During the modeling process, the MRI data used underwent preprocessing such as binarization, hole filling, boundary extraction, and earlobe segmentation to ensure the connectivity and integrity of the head region in three-dimensional space. Therefore, the three-dimensional model generated based on the MarchingCubes algorithm can realistically reflect the external envelope morphology of the head and avoid interference from non-target tissues on the model's geometry.

[0066] Through the above Python modeling process, a stable 3D head model can be obtained for subsequent 3D shape similarity calculation, providing a reliable data foundation for preoperative target localization of temporal interferometry (TI) therapy in the absence of MRI scans.

[0067] Step 3: After completing the preprocessing and 3D modeling of the MRI images, this invention performs a comprehensive comparison of the 3D head model generated from the MRI data to find the sample most similar to the head model obtained from the subject's 3D scan. To ensure consistency in the comparison, the 3D head model obtained from the scan is first processed in the same way as the MRI data, retaining only the head model from the top of the subject's head to the earlobe, in order to eliminate the influence of non-target areas below the neck on the similarity calculation.

[0068] like Figure 3 and Figure 4 As shown, in the similarity evaluation process, this invention selects Chamfer distance as the metric for 3D model similarity. Chamfer distance, calculated based on the nearest neighbor distance between points, reflects the geometric differences between the overall outer contours of two models. It has advantages such as high computational efficiency, insensitivity to changes in point cloud density, and robustness to a small number of outliers, making it suitable for rapid traversal and matching of large-scale MRI model databases. The application goal of this invention is to find the sample with the closest overall head contour; therefore, using Chamfer distance satisfies the balance between accuracy and efficiency.

[0069] Since 3D head models are typically stored as triangular meshes, directly calculating similarity based on these meshes is computationally intensive. To improve computational efficiency and standardize the representation of different models, this invention uniformly samples the surface of the 3D model, converting it into point cloud data. The sampled point cloud can better reflect the overall shape features of the model while significantly reducing computational complexity, making it suitable for subsequent registration and similarity calculations.

[0070] Before calculating point cloud similarity, considering the potential differences in position and pose between different 3D models in space, this invention employs the Iterative Closest Point (ICP) algorithm for rigid registration of the point clouds. ICP iteratively optimizes rotation and translation parameters to ensure that the source and target point clouds overlap as much as possible in space, thereby eliminating the influence of initial pose differences and improving the accuracy of subsequent Chamfer distance calculations.

[0071] After completing point cloud sampling and ICP alignment, this invention calculates the Chamfer distance between the scanned head point cloud and each head point cloud generated from an MRI image, and compares all MRI models. The MRI head model with the smallest Chamfer distance is identified as the model most similar to the target scan model, and its corresponding MRI scan sample is used as a substitute sample for the subject's head MRI scan for preoperative target localization in subsequent temporal interferometry (TI) therapy.

[0072] Charmfer distance: (4) In the formula, and This represents two sets of 3D point clouds that need to be compared. Represents a set Any point in (three-dimensional coordinates) ), Represents a set Any point in, and Representing sets respectively and set The total number of midpoints (base number). and Point and points The square of the Euclidean distance between them.

[0073] This invention also provides an alternative MRI reconstruction system based on three-dimensional head scanning, such as... Figure 6 As shown, the system includes: a preprocessing module, a matching module, and a selection module.

[0074] The preprocessing module is used to preprocess images from the MRI database to obtain head MRI data and construct a three-dimensional MRI head model. The matching module is used to acquire 3D head scan data and perform point cloud registration and shape similarity matching with the 3D head MRI model. The selection module is used to select the MRI head 3D model with the highest matching degree with the 3D head scan data as the alternative model based on the point cloud registration and shape similarity matching results.

[0075] It is understood that the alternative MRI construction system based on three-dimensional head scanning provided by the present invention corresponds to the alternative MRI construction method based on three-dimensional head scanning provided in the foregoing embodiments. The relevant technical features of the alternative MRI construction system based on three-dimensional head scanning can be referred to the relevant technical features of the alternative MRI construction method based on three-dimensional head scanning, and will not be repeated here.

[0076] Another object of the present invention is to provide an electronic device, such as... Figure 7 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor performing the steps of the alternative MRI construction method based on three-dimensional head scans.

[0077] The alternative MRI construction method based on three-dimensional head scanning includes the following steps: Images from the MRI database were preprocessed to obtain head MRI data, and a three-dimensional MRI head model was constructed. Acquire 3D head scan data, and use the 3D head scan data to perform point cloud registration and shape similarity matching with the MRI head 3D model; Based on the point cloud registration and shape similarity matching results, the MRI head 3D model with the highest matching degree to the 3D head scan data was selected as the alternative model.

[0078] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the alternative MRI construction method based on three-dimensional head scanning.

[0079] The alternative MRI construction method based on three-dimensional head scanning includes the following steps: Images from the MRI database were preprocessed to obtain head MRI data, and a three-dimensional MRI head model was constructed. Acquire 3D head scan data, and use the 3D head scan data to perform point cloud registration and shape similarity matching with the MRI head 3D model; Based on the point cloud registration and shape similarity matching results, the MRI head 3D model with the highest matching degree to the 3D head scan data was selected as the alternative model.

[0080] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0081] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0082] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0083] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for constructing an alternative MRI based on three-dimensional head scanning, characterized in that, include: Images from the MRI database were preprocessed to obtain head MRI data, and a three-dimensional MRI head model was constructed. Acquire 3D head scan data, and use the 3D head scan data to perform point cloud registration and shape similarity matching with the MRI head 3D model; Based on the point cloud registration and shape similarity matching results, the MRI head 3D model with the highest matching degree to the 3D head scan data was selected as the alternative model.

2. The alternative MRI construction method based on three-dimensional head scanning according to claim 1, characterized in that, Image preprocessing for MRI databases includes binarization of MRI images, hole filling of MRI images, boundary extraction of MRI images, and segmentation of MRI images.

3. The alternative MRI construction method based on three-dimensional head scanning according to claim 2, characterized in that, The binarization process of MRI images includes: Find a threshold based on the grayscale of MRI images to maximize the inter-class variance between the foreground and background; After performing automatic thresholding on the MRI image, regions with voxel values ​​higher than the threshold are set to 1, and regions with voxel values ​​lower than the threshold are set to 0, thereby obtaining the binarized result of the image; Hole filling in MRI images involves filling the background area completely surrounded by the foreground along the X, Y, and Z axes of each two-dimensional slice after binarization, thereby eliminating internal holes. Boundary extraction in MRI images includes: Perform dilation and erosion operations on the binary image of the head, respectively; After completing the expansion and corrosion processes, perform an XOR operation on the results of the two processes, extract the difference between the expanded and corroded regions, and obtain the corresponding head boundary region. MRI image segmentation includes: Browse MRI slices in three directions—axial, sagittal, and coronal—using a visual image user interface; Based on the selected direction, specific slice locations are selected for cropping, retaining the MRI data above the target area and removing the rest.

4. The alternative MRI construction method based on three-dimensional head scanning according to claim 3, characterized in that, The inter-class variance between the foreground and background is: In the formula, The inter-class variance between foreground and background. For the threshold, and These represent the pixel ratios of the background and foreground, respectively. grayscale value Histogram probability, The number of gray levels. and These are the average gray values ​​for the background and foreground, respectively.

5. The alternative MRI construction method based on three-dimensional head scanning according to claim 1, characterized in that, Methods for constructing 3D models of the MRI head include: Isosurfaces are extracted from the preprocessed MRI data. By traversing the voxel units in the volume data, the position of the isosurface in the voxel is determined by linear interpolation based on the relationship between the voxel vertices and a set threshold, and the three-dimensional voxel data is converted into a continuous triangular mesh model.

6. The alternative MRI construction method based on three-dimensional head scanning according to claim 1, characterized in that, Methods for point cloud registration and shape similarity matching include: Uniform sampling is performed on the surface of the MRI head 3D model, and the data is converted into point cloud data. The iterative nearest point algorithm is used to rigidly register the point cloud. The rotation and translation parameters are iteratively optimized to make the source point cloud and the target point cloud coincide in space. Calculate the Chamfer distance between the head point cloud from the 3D head scan data and the head point cloud generated from each MRI image, and perform a traversal comparison of all 3D head models from MRI.

7. The alternative MRI construction method based on three-dimensional head scanning according to claim 1, characterized in that, The MRI head 3D model with the highest matching degree to the 3D head scan data is the MRI head 3D model with the smallest Chamfer distance, which is: In the formula, and This represents two sets of 3D point clouds that need to be compared. Represents a set Any point in, Represents a set Any point in, and Representing sets and set The total number of midpoints and Point and points The square of the Euclidean distance between them.

8. An alternative MRI reconstruction system based on three-dimensional head scanning, characterized in that, include: The preprocessing module is used to preprocess images from the MRI database to obtain head MRI data and construct a three-dimensional MRI head model. The matching module is used to acquire 3D head scan data and perform point cloud registration and shape similarity matching with the 3D head MRI model. The selection module is used to select the MRI head 3D model with the highest matching degree with the 3D head scan data as the alternative model based on the point cloud registration and shape similarity matching results.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images