A method and system for precise positioning of individualized transcranial magnetic stimulation target points based on handheld three-dimensional scanning without MRI

Abstract

This invention discloses a method and system for precise target localization of transcranial magnetic stimulation (TMS) without MRI, based on handheld 3D scanning, belonging to the field of medical device technology. The method includes: acquiring scalp point cloud data of a subject using a handheld 3D scanner to reconstruct a personalized scalp 3D mesh model; mapping standard spatial target points onto the surface of the personalized scalp model to generate personalized target coordinates; acquiring facial point clouds and registering them with the scalp model to establish a first transformation relationship between the visual sensor and the model space; establishing a second transformation relationship between the visual sensor and the treatment device using a calibration tool; and converting the target coordinates to the physical space of the treatment device based on the two transformation relationships to perform localization. This invention enables precise personalized target localization without the need for large imaging equipment such as MRI / CT, ​​and is suitable for limited scenarios such as bedside and primary care clinics.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to a method and system for precise localization of individualized transcranial magnetic stimulation targets without MRI based on handheld three-dimensional scanning. Background Technology

[0002] Transcranial magnetic stimulation (TMS) is an effective non-invasive treatment for functional brain disorders such as major depressive disorder and chronic neuropathic pain. In TMS treatment, the precise positioning of the stimulation coil on the scalp directly determines whether the stimulation electric field can effectively act on the target brain region (such as the dorsolateral prefrontal cortex (DLPFC)), and is one of the most crucial factors affecting clinical efficacy.

[0003] Currently, stereotactic neuronavigation systems based on individualized magnetic resonance imaging (MRI) are widely recognized as the "gold standard" for target localization. However, this technology faces significant clinical implementation bottlenecks: the purchase and maintenance costs of MRI equipment are extremely high, the appointment process is lengthy, and some subjects cannot undergo scanning due to contraindications such as having metal implants, claustrophobia, or being pregnant. More importantly, high-precision neuronavigation is completely impossible in limited scenarios such as bedside emergency care, community clinics, and disaster relief in the wild where large imaging equipment cannot be deployed. Therefore, there is an urgent clinical need for an individualized target localization solution that can completely eliminate reliance on large imaging equipment such as MRI, while also offering high precision and good accessibility.

[0004] For target localization in the absence of MRI, existing technologies are mainly divided into two categories: sparse geometry measurement method and three-dimensional scanning-assisted method.

[0005] The first category is the cranial surface geometric localization method based on sparse measurements. Valter et al. proposed an MRI-free virtual navigation technique based on standard head model transformation. This technique measures three arc lengths of the subject's scalp using a measuring tape, approximates the skull size using an ellipsoidal model, and then performs a global affine transformation on the standard head model from the Montreal Neurological Institute (MNI) to generate a "pseudo-individualized" head model. However, this method relies solely on three sparse data points for global ellipsoid fitting, completely failing to recreate the complex local surface features of an individual skull (such as skull protrusions, depressions, and asymmetries). Its average target localization error is only 2.75 mm, which is insufficient to meet the demands of high-precision treatment. In addition, Jiang et al. systematically compared various localization methods based on measuring tape, such as electroencephalography (EEG) localization cap, Beam F3, and Continuous Proportional Coordinate (CPC) F3. These methods rely entirely on manual measurement and calculation by the operator. The average localization error of the best method (CPC F3) is still as high as 4.16 mm, and it has inherent defects such as poor repeatability and lack of visual guidance.

[0006] The second category comprises head data acquisition and assisted localization technologies based on 3D scanning. Koessler et al. disclosed a method for EEG sensor localization using a handheld 3D laser scanner, capable of acquiring dense point clouds and electrode positions on the scalp surface. However, the invention's purpose is limited to identifying and locating existing physical electrodes on the scalp for brain power imaging analysis after registration with MRI images. It not only relies on MRI data as a spatial registration benchmark but also completely ignores the issue of how to independently generate therapeutic targets in the absence of imaging—without MRI, the scan data itself lacks independent navigational value. Furthermore, Yang Rongqian et al. proposed a 3D visualization craniotomy localization method based on computed tomography (CT), but its data source relies entirely on cranial CT image segmentation, posing a risk of ionizing radiation (especially unsuitable for sensitive populations such as children and pregnant women). Moreover, its application is limited to outlining the craniotomy incision, which is fundamentally different from the localization of functional neuromodulation targets such as TMS in both principle and objective.

[0007] In summary, existing technologies exhibit significant limitations in MRI-free conditions: low-cost methods like the measuring tape method and standard head mold method suffer from severe accuracy deficiencies due to extremely sparse measurement information; while high-precision 3D scanning technology is confined to the inherent framework of "must be used in conjunction with MRI / CT images" or "only for locating existing physical electrodes." Those skilled in the art have not yet recognized the possibility of using handheld 3D scanning technology independently as the basis for individualized localization in MRI-free conditions. How to fully utilize high-density 3D point cloud data to reconstruct the true geometry of an individual's scalp without the need for large imaging equipment, and how to achieve precise mapping of standard functional target points on this individualized model to guide the treatment device in physical spatial positioning, is a pressing technical challenge. Therefore, this invention proposes a novel MRI-free individualized target point precise localization method and system based on handheld 3D scanning. Summary of the Invention

[0008] To address this, embodiments of the present invention provide a method and system for precise localization of individualized transcranial magnetic stimulation targets without MRI based on handheld 3D scanning. This method solves the problems of low target localization accuracy in the absence of medical images such as MRI / CT due to reliance on sparse manual measurement or scaling of standard templates, and the fact that existing 3D scanning technology only serves the identification of electrodes registered with MRI and cannot independently generate individualized treatment targets.

[0009] To address the aforementioned technical problems, embodiments of the present invention provide a method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning, comprising the following steps: Step S1: Use a handheld 3D scanner to perform a non-contact scan on the subject's head to obtain high-density 3D point cloud data of the subject's scalp surface, and reconstruct an individualized 3D mesh model of the subject's scalp based on the point cloud data. Step S2: Obtain the target coordinate set of the EEG localization system in standard space and the corresponding standard scalp model, register the standard scalp model with the individualized scalp 3D mesh model, and map the target coordinates in standard space to the surface of the individualized scalp 3D mesh model according to the registration result to generate the target coordinate set in the individualized scalp coordinate system. Step S3: Use a visual sensor to collect three-dimensional point cloud data of the subject's face, register the facial point cloud data with the individualized scalp three-dimensional mesh model, and establish a first spatial transformation relationship between the visual sensor coordinate system and the model coordinate system where the individualized scalp three-dimensional mesh model is located; Step S4: Keep the relative position of the treatment device and the vision sensor fixed, use the vision sensor to identify the calibration tool set on the treatment device, calculate the pose of the coordinate system of the calibration tool relative to the coordinate system of the vision sensor, and combine the kinematic data of the treatment device to establish a second spatial transformation relationship between the coordinate system of the vision sensor and the base coordinate system of the treatment device. Step S5: Based on the first spatial transformation relationship and the second spatial transformation relationship, calculate the third spatial transformation relationship between the model coordinate system and the treatment device base coordinate system, convert the target coordinate set under the individualized scalp coordinate system into physical space coordinates under the treatment device base coordinate system based on the third spatial transformation relationship, and control the treatment device to perform target positioning according to the physical space coordinates.

[0010] Preferably, in step S1, before performing a non-contact scan of the subject's head using a handheld 3D scanner, the following steps are also included: Subjects were instructed to wear hair-pressure caps to eliminate interference from hair on scanning accuracy; the handheld 3D scanner was a structured light scanner or a laser scanner.

[0011] Preferably, in step S2, registering the standard scalp model with the individualized scalp 3D mesh model specifically includes: The Super4PCS algorithm is used for global coarse registration to estimate the initial spatial transformation matrix; then the ICP algorithm is introduced for local fine registration. By iteratively optimizing the Euclidean distance between corresponding point sets, the accurate registration of the standard scalp model and the individualized scalp model is completed; the target point coordinate set of the EEG localization system in the standard space is the MNI standard space 10-20 system EEG target point coordinate set.

[0012] Preferably, in step S2, after mapping the target point coordinates in the standard space to the surface of the individualized scalp 3D mesh model, the method further includes: The mapped target coordinates are displayed on the individualized scalp 3D mesh model through a visual interactive interface, and users can pick and / or adjust the target position on the model surface using interactive tools to generate an individualized target coordinate set confirmed by the user.

[0013] Preferably, in step S3, registering the facial point cloud data with the individualized scalp 3D mesh model specifically includes: The Super4PCS algorithm is used to perform global coarse registration between the facial point cloud data and the point cloud data of the individualized scalp 3D mesh model to estimate the initial spatial transformation matrix. Then, the ICP algorithm is introduced for local fine registration. Through iterative optimization, high-precision alignment between the facial point cloud and the individualized scalp model is achieved, and the first spatial transformation relationship between the visual sensor coordinate system and the model coordinate system is established.

[0014] Preferably, the visual sensor is a binocular near-infrared camera with structured light function; the calibration tool is a calibration plate with reflective markers of known geometric dimensions, wherein the reflective markers are reflective spheres or high-contrast colored markers.

[0015] Preferably, in step S4, the process of using the visual sensor to identify a calibration tool mounted on the treatment device and calculating the pose of the calibration tool's coordinate system relative to the visual sensor's coordinate system specifically includes: The vision sensor identifies reflective markers of known geometric dimensions on the calibration tool, and calculates the rotation matrix and translation vector of the calibration tool coordinate system relative to the vision sensor coordinate system based on the spatial distribution of the reflective markers; the kinematic data of the treatment device is the TCP data of the end effector of the treatment device.

[0016] Preferably, in step S5, after controlling the treatment device to perform target localization according to the physical space coordinates, the method further includes: During the treatment, the visual sensor collects the position information of the marker on the subject's head in real time and monitors the subject's head movement. When head displacement is detected, the displacement compensation amount is calculated and fed back to the treatment device, which then controls the treatment device to adjust the target position in real time to follow the head movement.

[0017] This invention also provides an MRI-free personalized transcranial magnetic stimulation target precision localization system based on handheld three-dimensional scanning, comprising: The handheld 3D scanning module is used to perform non-contact scanning of the subject's head and obtain high-density 3D point cloud data of the subject's scalp surface; The modeling module is used to receive point cloud data collected by the handheld 3D scanning module and reconstruct an individualized 3D mesh model of the subject's scalp. The target point generation module is used to obtain the target point coordinate set of the EEG localization system in standard space and the corresponding standard scalp model, register the standard scalp model with the individualized scalp 3D mesh model, and map the target point coordinates in standard space to the surface of the individualized scalp 3D mesh model according to the registration result, thereby generating the target point coordinate set in the individualized scalp coordinate system. The visual sensing module is used to acquire three-dimensional point cloud data of the subject's face; The patient registration module is used to register the facial point cloud data with the individualized scalp 3D mesh model and establish a first spatial transformation relationship between the coordinate system of the visual sensing module and the model coordinate system where the individualized scalp 3D mesh model is located. The treatment device registration module is used to identify a calibration tool set on the treatment device using the visual sensing module while keeping the relative position of the treatment device and the visual sensing module fixed, calculate the pose of the coordinate system of the calibration tool relative to the coordinate system of the visual sensing module, and establish a second spatial transformation relationship between the coordinate system of the visual sensing module and the base coordinate system of the treatment device by combining the kinematic data of the treatment device. The positioning guidance module is used to calculate a third spatial transformation relationship between the model coordinate system and the treatment device base coordinate system based on the first spatial transformation relationship and the second spatial transformation relationship, convert the target coordinate set under the individualized scalp coordinate system into physical space coordinates under the treatment device base coordinate system based on the third spatial transformation relationship, and control the treatment device to perform target positioning according to the physical space coordinates.

[0018] Preferably, it also includes a real-time head movement tracking module, used to collect the position information of the subject's head markers in real time through the visual sensing module during the treatment process, monitor the subject's head movement, calculate the displacement compensation amount when head displacement is detected and feed it back to the treatment device, and control the treatment device to adjust the target position in real time to follow the head movement; the handheld three-dimensional scanning module is a structured light scanner or a laser scanner; the visual sensing module is a binocular near-infrared camera with structured light function; the calibration tool is a calibration plate with reflective markers of known geometric dimensions.

[0019] As can be seen from the above technical solutions, this invention application has the following beneficial effects: First, this invention uses a handheld 3D scanner to acquire scalp point clouds and reconstruct individualized models without the need for large imaging equipment such as MRI / CT. It can be flexibly deployed in limited scenarios such as bedside, primary clinics, and field emergency rescue, avoiding contraindications such as metal implants and claustrophobia, significantly reducing the implementation cost and threshold of precise neuromodulation, and greatly improving the accessibility of the technology.

[0020] Second, existing sparse measurement methods rely on only a very small amount of arc length data for ellipsoidal approximation, which cannot reconstruct the complex local curved surfaces of an individual's skull, resulting in positioning errors of over 4mm. This invention acquires a dense point cloud of the scalp through handheld scanning and reconstructs a high-fidelity mesh model, fully preserving the personalized geometric features of the skull. It combines a hybrid algorithm of Super4PCS global coarse registration and ICP local fine registration to accurately map standard spatial target points onto the individualized scalp surface. Dual registration through patient registration and robotic arm registration ensures high-precision alignment between the virtual model and physical space, enabling target point positioning accuracy to approach that of image-guided navigation even without imaging, effectively solving the core problem of large positioning errors in constrained scenarios.

[0021] Third, this invention automates the entire process from scanning, modeling, registration to positioning. Users only need to intuitively view and fine-tune the target position on a visual interactive interface, without requiring professional algorithm knowledge, thus reducing the operational threshold and human error. During treatment, the visual sensor tracks the head marker in real time and automatically feeds back the displacement compensation amount to the robotic arm to achieve head movement following, ensuring that the coil is always aligned with the target. Ordinary medical staff can operate it proficiently after simple training, significantly improving treatment efficiency and safety while ensuring accuracy. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Referring to the drawings will make the features and advantages of the present invention clearer. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of a method for precise target localization of individualized transcranial magnetic stimulation based on handheld three-dimensional scanning without MRI, provided by the present invention. Figure 2 This is a block diagram of a handheld 3D scanning-based MRI-free personalized transcranial magnetic stimulation target precision localization system provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. 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.

[0024] Example 1: To address the issues of low target localization accuracy in the absence of medical imaging such as MRI / CT due to reliance on sparse manual measurement or standard template scaling, and the limitations of existing 3D scanning technology which only serves for electrode identification registered with MRI and cannot independently generate individualized treatment targets, this invention proposes a method for precise MRI-free individualized transcranial magnetic stimulation target localization based on handheld 3D scanning. Figure 1 As shown, the method includes the following steps: Step S1: Use a handheld 3D scanner to perform a non-contact scan on the subject's head to obtain high-density 3D point cloud data of the subject's scalp surface, and reconstruct an individualized 3D mesh model of the subject's scalp based on the point cloud data. Step S2: Obtain the target coordinate set of the EEG localization system in standard space and the corresponding standard scalp model, register the standard scalp model with the individualized scalp 3D mesh model, and map the target coordinates in standard space to the surface of the individualized scalp 3D mesh model according to the registration result to generate the target coordinate set in the individualized scalp coordinate system. Step S3: Use a visual sensor to collect three-dimensional point cloud data of the subject's face, register the facial point cloud data with the individualized scalp three-dimensional mesh model, and establish a first spatial transformation relationship between the visual sensor coordinate system and the model coordinate system where the individualized scalp three-dimensional mesh model is located; Step S4: Keep the relative position of the treatment device and the vision sensor fixed, use the vision sensor to identify the calibration tool set on the treatment device, calculate the pose of the coordinate system of the calibration tool relative to the coordinate system of the vision sensor, and combine the kinematic data of the treatment device to establish a second spatial transformation relationship between the coordinate system of the vision sensor and the base coordinate system of the treatment device. Step S5: Based on the first spatial transformation relationship and the second spatial transformation relationship, calculate the third spatial transformation relationship between the model coordinate system and the treatment device base coordinate system, convert the target coordinate set under the individualized scalp coordinate system into physical space coordinates under the treatment device base coordinate system based on the third spatial transformation relationship, and control the treatment device to perform target positioning according to the physical space coordinates.

[0025] As can be seen from the above technical solution, this invention proposes a method for precise target localization of transcranial magnetic stimulation (TMS) without MRI based on handheld 3D scanning. First, a handheld 3D scanner is used to acquire high-density point clouds on the subject's scalp surface and reconstruct an individualized scalp 3D mesh model, providing a precise individualized geometric data foundation for subsequent registration and mapping. Then, the target coordinate set in standard space is mapped to the surface of the individualized scalp model through registration between the standard scalp model and the individualized scalp model, generating individualized target coordinates highly adapted to the subject's skull features. Subsequently, a visual sensor is used to acquire facial point clouds and register them with the scalp model. The process involves establishing a first transformation relationship between the visual sensor and the model space, achieving a preliminary connection between the virtual model and the physical space; then, establishing a second transformation relationship between the visual sensor and the base coordinate system of the treatment device through a calibration tool fixed on the treatment device; finally, combining the first and second transformation relationships to calculate a third transformation relationship between the model space and the treatment device space, converting the individualized target coordinates into the physical space coordinates of the treatment device and driving the treatment device to perform precise positioning, forming a complete closed loop from individualized modeling, target mapping to physical positioning, achieving high-precision individualized target positioning without relying on large imaging equipment such as MRI / CT.

[0026] Furthermore, in step S1, individualized scalp acquisition and modeling are first performed: a handheld 3D scanner is used to perform a non-contact scan of the subject's head to obtain high-density 3D point cloud data of the subject's scalp surface, and an individualized 3D mesh model of the subject's scalp is reconstructed based on the point cloud data.

[0027] Specifically, before scanning, the subject wears a hair-pressing cap. The cap compresses and gathers the subject's hair within the cap, eliminating interference from hair on scanning accuracy. Hair itself has irregular geometric shapes and can obscure the underlying scalp surface; without processing, the point cloud acquired by the scanner will contain a large amount of noise data generated by hair, resulting in an inability to obtain a clean scalp surface point cloud. With the hair-pressing cap on, the scanner can directly acquire point cloud data reflecting the true geometric contours of the skull through the surface morphology of the scalp. The handheld 3D scanner is either a structured light scanner or a laser scanner. A structured light scanner projects a specifically coded structured light pattern (such as stripes or grids) onto the target surface, uses a camera to capture the deformation of the pattern on the object's surface, and calculates the 3D coordinates of each pixel based on triangulation principles. A laser scanner, on the other hand, emits a laser beam and receives the reflected signal, measuring distance based on time-of-flight or phase difference principles. Both scanners are non-contact measurement devices, completing data acquisition without direct contact with the subject's skin.

[0028] During the scanning process, the operator slowly moves the handheld scanner around the subject's head, while the scanner continuously acquires high-density 3D point cloud data of the scalp surface. Since scanners typically have built-in real-time positioning and mapping capabilities (achieved by identifying geometric features of the scanned object's surface or pre-applied reference markers during the scanning process), point cloud data acquired from different perspectives can be automatically fused into the same coordinate system, forming a complete scalp surface point cloud. The navigation software receives the 3D scalp mesh model point cloud generated by the handheld 3D scanner's built-in software—which has already performed preprocessing such as noise reduction, smoothing, and triangulation of the point cloud, outputting a directly usable 3D mesh model (e.g., STL, OBJ, or PLY format). This model consists of numerous triangular facets, with each facet's vertex representing a 3D point cloud data point. The point cloud density typically reaches hundreds to thousands of points per square centimeter, sufficient to finely depict the undulations of the scalp surface, the skull contour, and other individual geometric features.

[0029] After the above processing, an individualized 3D mesh model of the subject's scalp was reconstructed. This model fully preserves the individualized geometric features of the subject's skull—including the overall size of the skull, the ratio of anteroposterior diameter to lateral diameter, the degree of forehead prominence, the curvature of the occipital region, and local details such as temporal depression—providing a solid data foundation for subsequent high-precision registration and target mapping.

[0030] Further, in step S2, individualized EEG target point generation and mapping are performed: the target point coordinate set of the EEG localization system in standard space and the corresponding standard scalp model are obtained, the standard scalp model is registered with the individualized scalp three-dimensional mesh model, and the target point coordinates in the standard space are mapped to the surface of the individualized scalp three-dimensional mesh model according to the registration result, thereby generating a target point coordinate set in the individualized scalp coordinate system.

[0031] The target coordinate set of the EEG localization system in the standard space is the MNI (Montreal Neurological Institute) standard space 10-20 system EEG target coordinate set. The 10-20 system is an internationally recognized standard for EEG electrode placement. This system divides the scalp into specific proportional locations based on scalp anatomical landmarks (nasal root, external occipital protuberance, and left and right preauricular points). Target points such as F3, Cz, and Pz are clearly defined in this system. The MNI standard space provides a standardized coordinate reference framework, allowing brain regions to be compared between different individuals.

[0032] The registration between the standard scalp model and the individualized scalp 3D mesh model adopts a hybrid registration strategy of "linear + nonlinear", which includes the following sub-steps: (I) Global Coarse Registration Based on the Super4PCS Algorithm. Super4PCS (Super 4-Points CongruentSets) is a global point cloud registration algorithm that can quickly complete coarse registration between two point clouds without providing initial pose estimation. The core principle of this algorithm is as follows: First, select a coplanar four-point basis (i.e., four points located in the same plane, usually a combination of four approximately coplanar points) in the source point cloud (standard scalp model point cloud), and calculate the affine invariants of the four-point basis—that is, the two proportional parameters between the four points (derived from the geometric constraints of the four points being coplanar, and these parameters remain unchanged under rigid body transformation).

[0033] Secondly, in the target point cloud (individualized scalp model point cloud), a spatial indexing structure (such as KDTree) is used to quickly search for all four-point combinations that have the same affine invariants as the aforementioned four-point basis. These four-point combinations are called "congruent four-point sets". Since affine invariants are invariant under rigid body transformations, there is a unique rigid body transformation relationship between the four-point basis in the source point cloud and the matching congruent four-point sets in the target point cloud.

[0034] Then, for each set of matched four-point pairs, calculate the corresponding rigid body transformation matrix (rotation matrix R and translation vector t), and apply the transformation to all points of the source point cloud. Calculate the degree of overlap between the source point cloud and the target point cloud after the transformation (i.e., the number of points that fall within a certain distance threshold of the target point cloud).

[0035] Finally, the transformation matrix with the highest overlap is selected as the initial spatial transformation matrix for global coarse registration. The Super4PCS algorithm, through a four-point basis affine invariant matching strategy, reduces the search space of the registration problem from O(N) to O(N). 4 The time complexity is reduced to O(N), with linear time complexity, and it can quickly find the near-optimal initial alignment in a large amount of point cloud data.

[0036] (II) Local Fine Registration Based on ICP Algorithm. The ICP (Iterative Closest Point) algorithm is a classic point cloud fine registration method used to calculate the optimal rotation and translation transformations between two point cloud datasets to achieve accurate alignment. This algorithm is based on the least squares method and iteratively optimizes the transformation parameters by performing the following two steps: The first step is to find the nearest corresponding point: For each point in the source point cloud (the standard scalp model point cloud after coarse registration transformation), search for the point with the closest Euclidean distance in the target point cloud (the individualized scalp model point cloud) and establish a one-to-one corresponding point pair relationship.

[0037] The second step is to solve for the optimal transformation: Based on all the corresponding point pairs established in the previous step, construct the objective function—that is, the sum of the squares of the Euclidean distances between all corresponding point pairs. Solve for the rigid body transformation matrix (rotation matrix R and translation vector t) that minimizes the objective function using the least squares method.

[0038] The obtained transformation matrix is ​​applied to the source point cloud to update its position. These two steps are then repeated until the convergence condition is met—that is, the mean square error is less than a preset threshold, the change in the transformation is less than a preset threshold, or the maximum number of iterations is reached. Since the ICP algorithm is sensitive to the initial pose, if the initial pose deviates significantly from the true value, it is prone to getting trapped in local optima. In this step, Super4PCS is first used for global coarse registration, providing a good initial transformation matrix for ICP, thus effectively avoiding the problem of ICP getting trapped in local optima and ensuring high-precision alignment of the two point clouds.

[0039] After completing the precise registration between the standard scalp model and the individualized scalp model, the MNI target coordinates in the standard space are mapped to the surface of the individualized scalp 3D mesh model based on the registration result (i.e., the final spatial transformation matrix). Specifically, for each target coordinate in the MNI standard space (such as the DLPFC target [-38,44,26]), multiplying it by the spatial transformation matrix obtained from the registration yields the corresponding coordinates of the target in the individualized scalp model space. Since the target may be located below the scalp surface (i.e., in a brain region), it is also necessary to project the target onto the scalp mesh model surface along the normal direction of that point (or along the ray direction from the center of the skull to that point) to obtain the target position on the scalp surface—this position is the stimulation position of the subsequent TMS coil.

[0040] Building upon the aforementioned automatic mapping, this embodiment also provides an interactive target setting method: The mapped target coordinates are displayed on a personalized scalp 3D mesh model via a visual interactive interface. Users can pick and / or adjust target positions on the model surface using interactive tools, generating a user-confirmed personalized target coordinate set. Specifically, the navigation software renders the personalized scalp model and mapped target points in a 3D visualization. Users can directly pick target positions on the model surface through mouse clicks, drags, and other interactive operations, or fine-tune the mapped target points. This interactive picking method provides users with flexible means of target confirmation and adjustment, and is particularly suitable for scenarios where clinicians manually modify target points based on anatomical landmarks or clinical experience.

[0041] Finally, a set of precise target coordinates in an individualized scalp coordinate system is generated.

[0042] Further, in step S3, patient spatial registration is performed: three-dimensional point cloud data of the subject's face is collected using a visual sensor, and the facial point cloud data is registered with the individualized scalp three-dimensional mesh model to establish a first spatial transformation relationship between the visual sensor coordinate system and the model coordinate system where the individualized scalp three-dimensional mesh model is located.

[0043] The visual sensor is a binocular near-infrared camera with structured light functionality. This camera consists of two near-infrared cameras and a structured light projection module: the structured light projection module projects a near-infrared structured light pattern (such as speckle or stripes) onto the target surface; the two near-infrared cameras simultaneously acquire images of the structured light pattern on the target surface from different angles; and the three-dimensional coordinates of each pixel on the target surface are calculated based on the principle of binocular stereo vision and triangulation. The advantage of using the near-infrared band is that it is insensitive to ambient visible light, can operate normally under normal indoor lighting conditions, and does not irritate the subject's eyes.

[0044] In practice, the camera performs a full-face scan of the patient—typically covering the entire facial area from the hairline to the jawline and from the left ear to the right ear—capturing dense 3D point cloud data of the patient's face. After acquiring the facial point cloud, the ROI (Region of Interest) is selected to remove point cloud data from non-facial areas such as the neck and shoulders, resulting in clean, real-life facial point cloud data.

[0045] For the real-life facial point cloud data and the individualized scalp 3D mesh model point cloud data generated in step S1, the same "coarse registration + fine registration" hybrid strategy as in step S2 is used for registration: First, the Super4PCS algorithm is used to perform global coarse registration between the facial point cloud data and the point cloud data of the individualized scalp 3D mesh model. Since there is a significant geometric overlap between the facial point cloud and the scalp model point cloud (the face is part of the scalp model), the Super4PCS algorithm can quickly find an approximate alignment transformation between the two sets of point clouds by utilizing the geometric features of these overlapping regions (such as prominent geometric structures like the bridge of the nose, eye sockets, and forehead). The specific principle of the algorithm is consistent with that described in step S2: coplanar four-point basis points are selected in the facial point cloud, affine invariants are calculated, congruent four-point sets are searched in the scalp model point cloud, the optimal initial transformation matrix is ​​selected through a voting mechanism, and the initial spatial transformation matrix is ​​estimated.

[0046] Then, based on the coarse registration, the ICP algorithm is introduced for local fine registration. Using the transformation matrix obtained from the coarse registration as the initial value for ICP, the transformation parameters are continuously optimized through iterative execution of two steps: "finding the nearest corresponding point" and "solving the optimal transformation," achieving high-precision alignment between the facial point cloud and the individualized scalp model. Since there are numerous overlapping points (the entire facial region) between the facial point cloud and the scalp model point cloud, the ICP algorithm can achieve sub-millimeter-level accuracy in the overlapping area through iterative optimization.

[0047] Through the above registration process, a first spatial transformation relationship is established between the visual sensor coordinate system and the model coordinate system where the individualized scalp 3D mesh model is located—that is, the rotation matrix R1 and translation vector t1 (or its inverse transformation) from the model coordinate system to the visual sensor coordinate system. This transformation relationship allows any physical space point subsequently acquired by the visual sensor to be transformed into the model coordinate system, and vice versa.

[0048] Further, in step S4, the registration of the treatment device (robotic arm registration) is completed: the relative position of the treatment device and the vision sensor is kept fixed, the calibration tool set on the treatment device is identified by the vision sensor, the pose of the coordinate system of the calibration tool relative to the coordinate system of the vision sensor is calculated, and the second spatial transformation relationship between the coordinate system of the vision sensor and the base coordinate system of the treatment device is established by combining the kinematic data of the treatment device.

[0049] This step employs a single-position fixed calibration strategy, meaning that the relative positions of the treatment device (such as a robotic arm) and the binocular near-infrared camera remain fixed throughout the entire calibration and use process—both are fixed to the same base or bracket and will not be moved once installed.

[0050] The calibration tool is a calibration plate with reflective markers of known geometric dimensions. These markers can be reflective spheres or high-contrast colored dots. The surface of the reflective spheres is coated with a high-reflectivity material (such as a glass microsphere reflective film), which produces strong echo reflections under near-infrared light illumination, facilitating rapid and accurate identification and positioning by the camera. The spatial distances and relative positions of the reflective spheres on the calibration plate are precisely known in advance—these geometric parameters have been calibrated and recorded using high-precision measuring equipment at the time the calibration tool leaves the factory.

[0051] The specific calibration process is as follows: First, a calibration plate with a reflective ball of known geometry is fixed to the end effector of the treatment device (robotic arm). The calibration plate can be securely installed using methods such as threaded connection, magnetic adsorption, or clamps, ensuring no relative displacement between the calibration plate and the end effector during the calibration process.

[0052] Secondly, a binocular near-infrared camera is used to photograph and identify the calibration board. The camera identifies the spatial position of each reflective ball on the calibration board—since the reflective balls appear as high-brightness circular spots in the near-infrared image, the pixel coordinates of the center of each ball in the image can be accurately calculated using image processing algorithms (such as threshold segmentation, centroid extraction, etc.); combined with the principle of binocular stereo vision, the three-dimensional spatial coordinates of the center of each ball in the visual sensor coordinate system can be further calculated.

[0053] Then, based on the three-dimensional coordinates of each reflective ball in the visual sensor coordinate system and its known coordinates in the calibration plate's own coordinate system (i.e., the relative positions of each ball determined during the calibration plate design), the pose of the calibration plate coordinate system relative to the visual sensor coordinate system is solved using the least squares method or singular value decomposition (SVD) algorithm. —that is, rotation matrix Translation vector This pose describes "where the calibration plate (i.e., the end effector of the robotic arm) is and in which direction it is facing".

[0054] Next, the TCP (Tool Center Point) data of the robotic arm's end effector is acquired. TCP data is the real-time position and orientation of the end effector in the robotic arm's base coordinate system, provided by the robotic arm control system—that is, the end-effector pose calculated from the angles of each joint of the robotic arm using forward kinematics. This data reflects the current kinematic state of the robotic arm.

[0055] Finally, combining the pose X1 of the calibration board coordinate system relative to the vision sensor coordinate system and the TCP data of the robotic arm end effector, the fixed transformation matrix of the vision sensor coordinate system relative to the robotic arm base coordinate system is calculated. Specifically, since the calibration plate is fixed to the end effector of the robotic arm, there is a fixed and known transformation relationship between the coordinate system of the calibration plate and the coordinate system of the end effector of the robotic arm (determined by the mounting method of the calibration plate); the pose of the vision sensor coordinate system relative to the coordinate system of the calibration plate is... The inverse of the coordinate system; the pose of the robotic arm's end effector coordinate system relative to the robotic arm's base coordinate system is given by TCP data. Through the composite operations of the coordinate system transformation chain, the fixed transformation relationship between the vision sensor coordinate system and the robotic arm's base coordinate system can be calculated—that is, the second spatial transformation relationship (rotation matrix). Translation vector Once this conversion relationship is established, it remains valid as long as the relative positions of the camera and the robotic arm remain unchanged, without the need for repeated calibration.

[0056] Further, in step S5, target physical space localization and treatment guidance are performed: based on the first spatial transformation relationship and the second spatial transformation relationship, a third spatial transformation relationship between the model coordinate system and the treatment device base coordinate system is calculated; based on the third spatial transformation relationship, the target coordinate set under the individualized scalp coordinate system is converted into physical space coordinates under the treatment device base coordinate system; and the treatment device is controlled to perform target localization according to the physical space coordinates.

[0057] Specifically, step S3 establishes the first transformation relationship between the model coordinate system and the visual sensor coordinate system. Step S4 establishes the second transformation relationship between the vision sensor coordinate system and the robotic arm base coordinate system. By performing composite operations of coordinate system transformations, the third transformation relationship between the model coordinate system and the robot arm's base coordinate system can be calculated. : , .

[0058] This transformation relationship describes "where any point in the individualized scalp 3D model is located in the robotic arm's base coordinate system".

[0059] Based on the third-space transformation relationship, the target point coordinate set (target points located on the surface of the scalp model, i.e., the stimulation positions of the TMS coils) generated in step S2 under the individualized scalp coordinate system is converted into physical space coordinates under the robotic arm base coordinate system. Specifically, for each target point coordinate P_model (three-dimensional coordinates in the model coordinate system), its corresponding physical space coordinates in the robotic arm base coordinate system are... for: .

[0060] After obtaining the physical spatial coordinates of the target point in the base coordinate system of the robotic arm, the robotic arm control system drives the joints of the robotic arm to move according to the coordinates, so that the TMS coil installed at the end of the robotic arm moves precisely to the target point position - that is, the stimulation center of the coil is aligned with the target point on the scalp surface, and the coil plane is tangent to the scalp surface or at a preset angle (usually 45 degrees).

[0061] Furthermore, during treatment, a visual sensor (binocular near-infrared camera) collects real-time positional information of the markers on the subject's head, monitoring the subject's head movements. Specifically, several reflective markers (usually 3-5, distributed on the forehead, temples, etc.) are applied to the subject's head, and the camera continuously tracks the three-dimensional spatial position of these markers at a high frequency (e.g., 30-60Hz). When the subject moves their head (e.g., slightly turning or nodding), the spatial coordinates of each marker change accordingly. By calculating the positional offset of the marker set, the magnitude and direction of the head displacement can be accurately quantified. When the detected head displacement exceeds a preset threshold (e.g., 1mm), the displacement compensation is calculated and fed back to the robotic arm control system, which controls the robotic arm to adjust the target position in real time to follow the head movement—that is, the robotic arm drives the TMS coil to move synchronously by the same amount of displacement, ensuring that the coil is always aligned with the original target position, avoiding off-target problems caused by head movement, and improving the continuity and effectiveness of treatment.

[0062] Example 2: like Figure 2 As shown, this invention provides a handheld 3D scanning-based MRI-free personalized transcranial magnetic stimulation (TMS) target precision localization system. This system is used to implement the MRI-free personalized TMS target precision localization method based on handheld 3D scanning described in Embodiment 1 above, specifically including: The handheld 3D scanning module is used to perform non-contact scanning of the subject's head, acquiring high-density 3D point cloud data of the subject's scalp surface. This module is either a structured light scanner or a laser scanner. Before scanning, the subject must wear a hair-pressing cap to eliminate interference from hair on the scanning accuracy.

[0063] The modeling module receives point cloud data acquired by the handheld 3D scanning module and reconstructs an individualized 3D mesh model of the subject's scalp. This module can be built into the handheld scanner's built-in software system, or it can be a standalone general-purpose 3D point cloud processing software or a customized modeling software.

[0064] The target point generation module acquires the target point coordinate set of the EEG localization system in standard space and the corresponding standard scalp model. It then registers the standard scalp model with the individualized scalp 3D mesh model. Based on the registration result, it maps the target point coordinates in standard space onto the surface of the individualized scalp 3D mesh model, generating a target point coordinate set in the individualized scalp coordinate system. This module employs a hybrid registration strategy using the Super4PCS algorithm for global coarse registration and the ICP algorithm for local fine registration. Furthermore, the module provides a visual interactive interface, allowing users to pick and / or adjust target point positions on the 3D model surface using interactive tools.

[0065] The visual sensing module is used to acquire 3D point cloud data of the subject's face. This module is a binocular near-infrared camera with structured light capability.

[0066] The patient registration module is used to register the facial point cloud data with the individualized scalp 3D mesh model, establishing a first spatial transformation relationship between the coordinate system of the visual sensing module and the coordinate system of the individualized scalp 3D mesh model. This module also employs a hybrid registration strategy combining Super4PCS global coarse registration and ICP local fine registration.

[0067] The treatment device registration module is used to, while keeping the relative position of the treatment device and the visual sensing module fixed, utilize the visual sensing module to identify a calibration tool set on the treatment device, calculate the pose of the calibration tool's coordinate system relative to the visual sensing module's coordinate system, and, combined with the kinematic data of the treatment device, establish a second spatial transformation relationship between the visual sensing module's coordinate system and the treatment device's base coordinate system. The calibration tool is a calibration plate with reflective markers of known geometric dimensions; the reflective markers are reflective spheres or high-contrast colored dots.

[0068] The positioning guidance module is used to calculate a third spatial transformation relationship between the model coordinate system and the treatment device base coordinate system based on the first spatial transformation relationship and the second spatial transformation relationship, convert the target coordinate set under the individualized scalp coordinate system into physical space coordinates under the treatment device base coordinate system based on the third spatial transformation relationship, and control the treatment device to perform target positioning according to the physical space coordinates.

[0069] The real-time head movement tracking module is used to collect the position information of the subject's head markers in real time through the visual sensing module during the treatment process, monitor the subject's head movement, calculate the displacement compensation amount when head displacement is detected and feed it back to the treatment device, and control the treatment device to adjust the target position in real time to follow the head movement.

[0070] This embodiment presents a handheld 3D scanning-based MRI-free personalized transcranial magnetic stimulation (TMS) target precision localization system, used to implement the aforementioned handheld 3D scanning-based MRI-free personalized TMS target precision localization method. Therefore, the specific implementation of the handheld 3D scanning-based MRI-free personalized TMS target precision localization system can be found in the previous embodiment section of the handheld 3D scanning-based MRI-free personalized TMS target precision localization method. To avoid redundancy, it will not be repeated here.

[0071] Supplementary Explanation of Algorithm Principles To enable those skilled in the art to better understand the technical principles of the key algorithms used in this invention, the Super4PCS algorithm and ICP algorithm involved in steps S2 and S3 will be further described in detail below.

[0072] The mathematical principle of the Super4PCS algorithm: The core idea of ​​the Super4PCS algorithm is based on the invariance of four coplanar affine invariants under rigid body transformation. Let the source point cloud P contain four coplanar points. Its affine invariant is defined as: , ; Where 'a' is the reference point among the four points, and 'b', 'c', and 'd' are the other three points. Under rigid body transformations (rotation and translation), the Euclidean distance between the points remains unchanged, therefore... and It remains unchanged before and after the transformation.

[0073] In the target point cloud Q, for any four coplanar points If its affine invariants are the same as the four-point basis in the source point cloud (i.e., and If the four-point set is identical to the source four-point basis, then the four-point set is called a four-point set "congruent" to the source four-point basis. For each pair of matching four-point bases, the rigid body transformation matrix can be obtained by solving the following system of linear equations: For the three non-collinear points in the source four-point basis (such as a, b, c) and the corresponding three points in the target four-point basis ( , , There exists a unique rigid body transformation. Make: , , ; This system of equations can be solved by SVD decomposition or orthogonal Procrustes analysis.

[0074] The Super4PCS algorithm achieves global coarse registration with linear time complexity by using a spatial index structure to quickly search for all four-point combinations that are identical to the source four-point basis in the target point cloud, and by using a voting mechanism (i.e., counting the number of overlapping points between the source and target point clouds after each candidate transformation) to select the optimal transformation.

[0075] The mathematical principles of the ICP algorithm: The objective of the ICP algorithm is to minimize the following objective function: ; in The first point cloud in the source cloud One point, In the target point cloud The nearest corresponding point, For rotation matrix, It is a translation vector.

[0076] The iterative process of the algorithm is as follows: Step 1 (Corresponding Point Search): For each point in the source point cloud Search for the point in the target point cloud with the closest Euclidean distance. Establish corresponding point pairs The search process is typically accelerated using spatial indexing structures such as KDTree.

[0077] Step 2 (Transformation Estimation): Based on all corresponding point pairs, solve the objective function using the least squares method. Minimize the rigid body transformation matrix. This problem can be solved as follows: First, calculate the centroids of the two sets of point clouds to center the point clouds. Then, perform SVD decomposition on the covariance matrix to obtain the optimal rotation matrix. Finally, calculate the optimal translation vector based on the relationship between the rotation matrix and the translation vector.

[0078] Step 3 (Transformation Application): Apply the obtained transformation matrix to all points in the source point cloud to update the position of the source point cloud: .

[0079] Step 4 (Convergence Judgment): Calculate the mean square error between the updated source point cloud and the target point cloud. If the mean square error is less than the preset threshold, or the change in the transformation is less than the preset threshold, or the maximum number of iterations is reached, then terminate the iteration; otherwise, return to step 1 to continue the iteration.

[0080] The ICP algorithm ultimately outputs a rigid body transformation matrix that achieves optimal alignment between the source point cloud and the target point cloud in the overlapping region.

[0081] By employing the "coarse-to-fine" combination strategy of Super4PCS and ICP—Super4PCS providing globally optimal initial alignment, and ICP performing local fine-tuning based on this—high accuracy and robustness of point cloud registration are achieved. Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application 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.

[0082] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. 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... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0083] 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 functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus 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] 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 method for precise target localization of individualized transcranial magnetic stimulation based on handheld 3D scanning without MRI, characterized in that, Includes the following steps: Step S1: Use a handheld 3D scanner to perform a non-contact scan on the subject's head to obtain high-density 3D point cloud data of the subject's scalp surface, and reconstruct an individualized 3D mesh model of the subject's scalp based on the point cloud data. Step S2: Obtain the target coordinate set of the EEG localization system in standard space and the corresponding standard scalp model, register the standard scalp model with the individualized scalp 3D mesh model, and map the target coordinates in standard space to the surface of the individualized scalp 3D mesh model according to the registration result to generate the target coordinate set in the individualized scalp coordinate system. Step S3: Use a visual sensor to collect three-dimensional point cloud data of the subject's face, register the facial point cloud data with the individualized scalp three-dimensional mesh model, and establish a first spatial transformation relationship between the visual sensor coordinate system and the model coordinate system where the individualized scalp three-dimensional mesh model is located; Step S4: Keep the relative position of the treatment device and the vision sensor fixed, use the vision sensor to identify the calibration tool set on the treatment device, calculate the pose of the coordinate system of the calibration tool relative to the coordinate system of the vision sensor, and combine the kinematic data of the treatment device to establish a second spatial transformation relationship between the coordinate system of the vision sensor and the base coordinate system of the treatment device. Step S5: Based on the first spatial transformation relationship and the second spatial transformation relationship, calculate the third spatial transformation relationship between the model coordinate system and the treatment device base coordinate system, convert the target coordinate set under the individualized scalp coordinate system into physical space coordinates under the treatment device base coordinate system based on the third spatial transformation relationship, and control the treatment device to perform target positioning according to the physical space coordinates.

2. The method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning according to claim 1, characterized in that, In step S1, before performing a non-contact scan of the subject's head using a handheld 3D scanner, the following steps are also included: Subjects were instructed to wear hair-pressure caps to eliminate interference from hair on scanning accuracy; the handheld 3D scanner was a structured light scanner or a laser scanner.

3. The method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning according to claim 1, characterized in that, In step S2, the standard scalp model is registered with the individualized scalp 3D mesh model, specifically including: The Super4PCS algorithm is used for global coarse registration to estimate the initial spatial transformation matrix; then the ICP algorithm is introduced for local fine registration. By iteratively optimizing the Euclidean distance between corresponding point sets, the accurate registration of the standard scalp model and the individualized scalp model is completed; the target point coordinate set of the EEG localization system in the standard space is the MNI standard space 10-20 system EEG target point coordinate set.

4. The method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning according to claim 1, characterized in that, In step S2, after mapping the target point coordinates in the standard space to the surface of the individualized scalp 3D mesh model, the method further includes: The mapped target coordinates are displayed on the individualized scalp 3D mesh model through a visual interactive interface, and users can pick and / or adjust the target position on the model surface using interactive tools to generate an individualized target coordinate set confirmed by the user.

5. The method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning according to claim 1, characterized in that, In step S3, the facial point cloud data is registered with the individualized scalp 3D mesh model, specifically including: The Super4PCS algorithm is used to perform global coarse registration between the facial point cloud data and the point cloud data of the individualized scalp 3D mesh model to estimate the initial spatial transformation matrix. Then, the ICP algorithm is introduced for local fine registration. Through iterative optimization, high-precision alignment between the facial point cloud and the individualized scalp model is achieved, and the first spatial transformation relationship between the visual sensor coordinate system and the model coordinate system is established.

6. The method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning according to claim 1, characterized in that, The visual sensor is a binocular near-infrared camera with structured light function; the calibration tool is a calibration plate with reflective markers of known geometric dimensions, wherein the reflective markers are reflective spheres or high-contrast colored dots.

7. The method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning according to claim 1, characterized in that, In step S4, the visual sensor is used to identify the calibration tool set on the treatment device, and the pose of the calibration tool's coordinate system relative to the visual sensor's coordinate system is calculated. Specifically, this includes: The vision sensor identifies reflective markers of known geometric dimensions on the calibration tool, and calculates the rotation matrix and translation vector of the calibration tool coordinate system relative to the vision sensor coordinate system based on the spatial distribution of the reflective markers; the kinematic data of the treatment device is the TCP data of the end effector of the treatment device.

8. The method for precise target localization of MRI-free individualized transcranial magnetic stimulation based on handheld three-dimensional scanning according to claim 1, characterized in that, In step S5, after controlling the treatment device to perform target localization according to the physical space coordinates, the method further includes: During the treatment, the visual sensor collects the position information of the marker on the subject's head in real time and monitors the subject's head movement. When head displacement is detected, the displacement compensation amount is calculated and fed back to the treatment device, which then controls the treatment device to adjust the target position in real time to follow the head movement.

9. A handheld 3D scanning-based MRI-free personalized transcranial magnetic stimulation target localization system, characterized in that, include: The handheld 3D scanning module is used to perform non-contact scanning of the subject's head and obtain high-density 3D point cloud data of the subject's scalp surface; The modeling module is used to receive point cloud data collected by the handheld 3D scanning module and reconstruct an individualized 3D mesh model of the subject's scalp. The target point generation module is used to obtain the target point coordinate set of the EEG localization system in standard space and the corresponding standard scalp model, register the standard scalp model with the individualized scalp 3D mesh model, and map the target point coordinates in standard space to the surface of the individualized scalp 3D mesh model according to the registration result, thereby generating the target point coordinate set in the individualized scalp coordinate system. The visual sensing module is used to acquire three-dimensional point cloud data of the subject's face; The patient registration module is used to register the facial point cloud data with the individualized scalp 3D mesh model and establish a first spatial transformation relationship between the coordinate system of the visual sensing module and the model coordinate system where the individualized scalp 3D mesh model is located. The treatment device registration module is used to identify a calibration tool set on the treatment device using the visual sensing module while keeping the relative position of the treatment device and the visual sensing module fixed, calculate the pose of the coordinate system of the calibration tool relative to the coordinate system of the visual sensing module, and establish a second spatial transformation relationship between the coordinate system of the visual sensing module and the base coordinate system of the treatment device by combining the kinematic data of the treatment device. The positioning guidance module is used to calculate a third spatial transformation relationship between the model coordinate system and the treatment device base coordinate system based on the first spatial transformation relationship and the second spatial transformation relationship, convert the target coordinate set under the individualized scalp coordinate system into physical space coordinates under the treatment device base coordinate system based on the third spatial transformation relationship, and control the treatment device to perform target positioning according to the physical space coordinates.

10. The MRI-free personalized transcranial magnetic stimulation target precision localization system based on handheld three-dimensional scanning according to claim 9, characterized in that, It also includes a real-time head movement tracking module, which is used to collect the position information of the subject's head markers in real time through the visual sensing module during the treatment process, monitor the subject's head movement, calculate the displacement compensation amount when head displacement is detected and feed it back to the treatment device, and control the treatment device to adjust the target position in real time to follow the head movement; the handheld three-dimensional scanning module is a structured light scanner or a laser scanner; the visual sensing module is a binocular near-infrared camera with structured light function; the calibration tool is a calibration plate with reflective markers of known geometric dimensions.

Images