A Deep Brain Stimulation (DBS) Electrode Planning and Reconstruction System and Method for Pig Brains

Abstract

This invention discloses a system and method for deep brain stimulation (DBS) electrode planning and reconstruction in pigs, belonging to the field of medical image processing and neurosurgical pathway planning technology. The invention includes an automatic registration module based on magnetic resonance (MR) images and standard pig brain atlases for accurate identification of target brain regions; a module for optimal electrode trajectory planning within the target area based on principal component analysis (PCA) and an electric field simulation module based on the Hodgkin-Huxley model for optimizing treatment effects and predicting volume of activated tissue (VTA); and a multimodal hybrid registration module based on postoperative computed tomography (CT) scans and preoperative MR images for accurate reconstruction of the actual electrode positions. This invention, through the combination of automated, data-driven planning and precise postoperative verification, significantly improves the accuracy, reliability, and efficiency of electrode implantation in pig brain DBS research.

Description

Technical Field

[0001] This invention relates to the fields of medical image processing, computer-aided pathway planning, and neuromodulation, and particularly to a system and method for deep brain stimulation (DBS) electrode planning and reconstruction in pigs. Background Technology

[0002] Deep brain stimulation (DBS), as an important neuromodulation therapy, has been widely used to treat various neurological disorders such as Parkinson's disease, essential tremor, epilepsy, and depression. Its core lies in the precise implantation of electrodes into specific nuclei (target sites) deep within the brain, applying electrical impulses to modulate abnormal neural circuit activity. The success of DBS largely depends on the precision of electrode implantation.

[0003] Before applying DBS technology clinically, extensive preclinical studies are typically required in large animal models (such as pigs) to validate its safety and efficacy. The pig brain shares high similarity with the human brain in anatomical structure, size, and gray-white matter ratio, making it an ideal experimental animal. However, existing DBS research procedures for pig models present several challenges:

[0004] 1. Low planning accuracy: Due to the lack of detailed digital maps of the pig brain and efficient registration tools, target selection and electrode trajectory planning largely rely on manual operation by researchers. This is not only time-consuming and labor-intensive, but also highly subjective, making it difficult to guarantee accuracy and repeatability.

[0005] 2. Lack of simulation prediction: Traditional planning methods cannot predict the electric field distribution and the actual range of influence on surrounding nerve tissue, i.e., the volume of activation (VTA), at specific electrode parameters (such as voltage and pulse width) and locations. This makes parameter setting somewhat arbitrary and difficult to optimize treatment effects.

[0006] 3. Inadequate verification methods: After electrode implantation, postoperative images (usually CT) are needed to verify whether the actual position of the electrode matches the plan. Multimodal registration of postoperative CT and preoperative MRI is crucial for verification. However, due to the different imaging principles of the two modalities, metal artifacts exist on CT, making image registration difficult. Manual registration has low accuracy, while fully automatic registration algorithms often fail due to artifact interference.

[0007] 4. Fragmented processes: Currently, from preoperative planning and simulation to postoperative verification, multiple different software and tools are usually required. Data import and export are cumbersome, the processes are fragmented, and efficiency is low. There is a lack of an integrated and automated solution.

[0008] Therefore, there is an urgent need for an integrated and automated electrode planning and reconstruction system specifically designed for pig brain DBS research to solve the above problems and improve the efficiency and quality of preclinical research. Summary of the Invention

[0009] The present invention aims to overcome the shortcomings of the prior art and provide a system and method for planning and reconstructing pig brain DBS electrodes. The main technical problem it solves is: how to achieve fully automated, precise and integrated management of pig brain DBS electrodes from preoperative planning, electric field simulation to postoperative position verification.

[0010] The objective of this invention is achieved through the following technical solution: a deep brain stimulation (DBS) electrode planning and reconstruction system for pigs, comprising:

[0011] The atlas registration module is used to register preoperative MR images of pigs with standard digital atlases of the pig brain to automatically identify one or more target brain regions on the MR images.

[0012] The path planning module is used to calculate the optimal electrode implantation trajectory based on the location and morphological information of the target brain region, and to perform electric field simulation on the electrodes implanted along the trajectory.

[0013] The postoperative reconstruction module is used to register the postoperative computed tomography (CT) images of pigs with the preoperative MR images in order to reconstruct and verify the actual implantation location of the electrodes in the preoperative MR image space.

[0014] Furthermore, the registration process performed by the map registration module includes:

[0015] 1) Initial alignment of preoperative MR images with standard pig brain digital atlas using rigid or affine transformation;

[0016] 2) Perform nonlinear fine registration based on the Demons algorithm for B-spline or contour registration to achieve high-precision matching between the atlas and the anatomical structures of the MR images, in order to accommodate anatomical differences between individuals.

[0017] Furthermore, the path planning module calculates the optimal electrode implantation trajectory to ensure that the multiple stimulation contacts of the electrodes can cover the effective area of ​​the target brain region to the greatest extent. Specific methods include:

[0018] 1) When the target brain region is a single brain region: extract all voxels of the target brain region in three-dimensional space to form a voxel point cloud; perform principal component analysis (PCA) on the coordinates of the voxel point cloud to determine the first principal component direction of the voxel point cloud as the longest principal axis of the target brain region; define the longest principal axis as the direction of the optimal electrode implantation trajectory, and calculate the two endpoints of the voxel point cloud on the longest principal axis as the theoretical entry point and target point of the electrode;

[0019] 2) When there are two target brain regions: Calculate the centroids of the two target brain regions respectively. Based on the size of the target brain regions, determine the relatively smaller deep brain region as the target point and the relatively larger superficial brain region as the entry point. Directly connect the two centroids to form the electrode trajectory, and extend a preset distance outside the centroid of the superficial brain region as the electrode entry point.

[0020] 3) When there are two or more target brain regions: calculate the centroid of each target brain region, construct a centroid point cloud from all centroid coordinates, perform principal component analysis (PCA) on the centroid point cloud, determine the first principal component direction as the optimal trajectory direction, calculate the two extreme points of the centroid point cloud along the optimal trajectory direction, and extend a preset distance from the end closer to the skull as the electrode entry point and target point.

[0021] Furthermore, the method by which the path planning module performs electric field simulation includes:

[0022] 1) Establish a three-dimensional computational mesh around the virtually implanted electrodes;

[0023] 2) Automatically segment brain tissue based on MR image signal intensity characteristics and establish a multi-tissue conductivity model, including gray matter, white matter, cerebrospinal fluid, deep nuclei and blood vessels;

[0024] 3) Construct an anisotropic conductivity tensor for the white matter region. The directional distribution of conductivity is calculated based on the fiber orientation;

[0025] 4) Solve the Poisson equation in inhomogeneous media using the finite element method. Calculate the potential distribution and through Calculate the electric field intensity distribution; where, It is the local conductivity, which is the fiber bundle anisotropic tensor in the white matter region and a scalar in other tissues; V is the electric potential, in V; I is the injected current intensity, in A. yes A function used to introduce a point current source at the electrode location; It is the spatial gradient operator; E is the electric field intensity, in V / m;

[0026] 5) Using the Hodgkin-Huxley neuron model, solve the differential equation response of neuronal membrane potential to the electric field stimulation, and determine the tissue activation volume VTA based on whether the membrane potential reaches the activation threshold.

[0027] Furthermore, the Hodgkin-Huxley neuron model is a set of nonlinear differential equations describing the dynamic changes in neuronal membrane potential. By numerically solving the equations, it accurately simulates whether each neuron around the electrode will generate an action potential under the action of an electric field; the region formed by all neurons that generate action potentials is the tissue activation volume (VTA).

[0028] Furthermore, the registration process performed by the postoperative reconstruction module is a hybrid registration method, including:

[0029] 1) Preprocessing of CT images: Soft tissue contrast is enhanced and skull artifacts are suppressed by setting soft tissue windows and applying gamma correction.

[0030] 2) Automatic registration stage: The postoperative CT images and preoperative MR images are automatically initially registered on the multi-resolution image pyramid using a metric based on mutual information and a limited memory L-BFGS-B optimization algorithm.

[0031] 3) Manual fine-tuning stage: Provides a graphical user interface that allows operators to simultaneously observe and fuse registered CT and MR images in axial, coronal, and sagittal planes, and to manually fine-tune the automatic registration results with six degrees of freedom of rotation and translation to achieve final accurate alignment.

[0032] Furthermore, CT image preprocessing is used to enhance the contrast between brain tissue and electrodes. Specifically, a soft tissue window is used to truncate the CT values ​​and gamma correction is performed to highlight soft tissue details and suppress registration interference caused by parts of the skull.

[0033] Furthermore, the automatic registration stage adopts a multi-resolution registration framework, which performs initial alignment on coarse-resolution images and then gradually transitions to fine matching on high-resolution images; the similarity metric for registration adopts Mutual Information, which is robust to multimodal images, and the optimizer adopts the L-BFGS-B algorithm.

[0034] Furthermore, after the registration process in the postoperative reconstruction module is completed, the electrode model reconstructed on the CT can be accurately superimposed on the preoperative MR image and compared with the planned target point and trajectory to quantitatively assess the path error.

[0035] On the other hand, the present invention also provides a method for planning and reconstructing deep brain stimulation (DBS) electrodes in pigs, the method comprising the following steps:

[0036] (1) Obtain preoperative magnetic resonance MR images of pigs and register them with standard pig brain digital atlases to automatically identify one or more target brain regions on the MR images;

[0037] (2) Based on the location and morphological information of the target brain region, calculate the optimal electrode implantation trajectory and perform electric field simulation on the electrodes implanted along the trajectory;

[0038] (3) Register the postoperative computed tomography (CT) images of pigs with the preoperative MR images to reconstruct and verify the actual implantation location of the electrodes in the preoperative MR image space.

[0039] Compared with the prior art, the present invention has the following significant advantages:

[0040] 1. Improved accuracy and objectivity of planning: Through MR-map registration and PCA-based trajectory optimization algorithm, the automation and data-driven approach to target identification and trajectory planning are realized, eliminating the subjectivity and uncertainty of manual planning and improving the accuracy and repeatability of planning.

[0041] 2. Quantitative prediction of treatment effects was achieved: VTA simulation based on the Hodgkin-Huxley model was introduced, enabling researchers to quantitatively predict the range of neural activation under different stimulation parameters before surgery, providing a scientific basis for optimizing DBS parameter settings, which is something that traditional methods cannot achieve.

[0042] 3. Improved accuracy and reliability of postoperative verification: The innovative "automatic + manual" hybrid registration strategy makes full use of the computer's efficient computing power and the human eye's precise recognition ability, effectively solving the challenges caused by modal differences and metal artifacts in postoperative CT-MR registration, and achieving accurate reconstruction of the true position of the electrodes.

[0043] 4. An integrated research process has been established: This invention integrates preoperative planning, simulation prediction and postoperative verification into a unified system platform, realizing seamless data flow and high functional integration, greatly simplifying the operation process and improving the overall efficiency of preclinical DBS research. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a general flowchart of a method for planning and reconstructing pig brain DBS electrodes according to an embodiment of the present invention.

[0046] Figure 2This is a detailed flowchart of the atlas registration module according to an embodiment of the present invention, showing the specific steps for registering MR images with standard pig brain atlases;

[0047] Figure 3 This is a flowchart of the optimal electrode trajectory calculation according to an embodiment of the present invention, illustrating different processing strategies for a single brain region, two brain regions, and multiple brain regions;

[0048] Figure 4 This is a flowchart of an electric field simulation module according to an embodiment of the present invention, showing the detailed process of multi-tissue conductivity modeling and VTA calculation of the Hodgkin-Huxley neuron model;

[0049] Figure 5 This is a flowchart of a postoperative reconstruction module according to an embodiment of the present invention, illustrating the specific steps of CT-MR hybrid registration and electrode position verification. Detailed Implementation

[0050] 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, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0051] Reference Figures 1-5 This invention provides a complete embodiment of DBS electrode planning and reconstruction in the pig brain. For example... Figure 1 The overall workflow shown includes three main stages: map registration, pathway planning, and postoperative reconstruction. The detailed steps for each stage are as follows: Figures 2-5 As shown, the detailed steps are as follows:

[0052] Example: A complete workflow from electrode planning and design to post-implantation target reconstruction

[0053] Preparation

[0054] Data acquisition: High-resolution T1-weighted (T1w) and T2-weighted (T2w) MR image sequences of the experimental pig heads were acquired using a 3.0T medical magnetic resonance scanner. The images were saved in NIfTI format (.nii.gz).

[0055] System preparation: Start the porcine brain DBS electrode planning and reconstruction system described in this invention. A standardized three-dimensional digital atlas of the porcine brain (e.g., a standard 12-week-old porcine brain atlas) is pre-loaded into the system. This atlas is also in NIfTI format, where different integer values ​​represent different brain regions.

[0056] Step 1: Map registration and target determination

[0057] like Figure 2 As shown, the operator loads the preoperative T1w MR images of the experimental pig into the system.

[0058] The system starts the map registration module and automatically performs MR map registration.

[0059] Initial alignment: First, the system calculates and aligns the geometric center of the MR image with that of the atlas, and then performs a 12-parameter affine transformation to eliminate global translation, rotation, scaling and shear differences between the two.

[0060] Fine-grained registration: Building upon affine registration, the system initiates a B-spline-based nonlinear registration algorithm. This algorithm divides the image into a grid and fits the deformation of local anatomical structures by optimizing the positions of control points, ultimately achieving precise matching between the atlas and the MR image.

[0061] After registration, the segmentation labels for all brain regions in the atlas were converted and applied to the MR image space of the experimental pig, thereby automatically segmenting target brain regions such as the subthalamic nucleus (STN) and the globus pallidus (GPi). For example, the automatically segmented subthalamic nucleus (STN) can be clearly seen in the system's 3D and 2D slice views. The operator selected the left STN as the target brain region for this simulated implantation.

[0062] Step 2: Optimal Trajectory Planning and Electrophysiological Simulation

[0063] The operator confirms STN as the target in the system and starts the simulated path planning module.

[0064] Optimal trajectory calculation (e.g.) Figure 3 (as shown)

[0065] For a single brain region (taking the STN as an example): The system automatically extracts the coordinates of all voxels in the STN from the MR image, forming a 3D point cloud. The system performs principal component analysis (PCA) on this point cloud coordinates. The calculation results show that the first principal component vector is (0.85, -0.49, 0.21), which represents the longest axis direction of the porcine STN structure in space. The system calculates the two endpoints of the STN point cloud along this direction, automatically generating an electrode trajectory that passes through the longest axis of the STN.

[0066] Dual brain region scenario (taking STN and GPi as an example): When the operator simultaneously selects the STN and the globus pallidus (GPi) as target brain regions, the system calculates the centroids of the two brain regions separately. By comparing the number of voxels in the brain regions, the system determines the STN (a smaller, deep nucleus) as the target point and the GPi (a larger, superficial nucleus) as the entry point direction. The system directly connects the two centroids and extends 40 mm outside the GPi centroid to form a complete electrode trajectory, ensuring that the electrode can safely enter from outside the skull and pass through both target brain regions.

[0067] For multiple brain regions (using STN, GPi, and VIM of the ventral thalamus as examples): When the operator selects three or more brain regions, the system calculates the centroid of each region separately, constructing a centroid point cloud from all centroid coordinates. Principal component analysis is performed on this centroid point cloud to determine the direction of the first principal component as the optimal trajectory direction. The system calculates two extreme points along this direction of the centroid point cloud, extending 40 mm outward from the shallower end as the electrode entry point, and the deeper extreme point as the target point, thereby generating an electrode trajectory that maximizes coverage of all target brain regions.

[0068] In either case, the generated trajectory is displayed in a 3D view, with the target point located near the geometric center of the target brain region and the entry point located in a safe area on the outer side of the skull. This data-driven approach, based on adaptive brain region number, ensures that the electrode trajectory closely matches the anatomical morphology of the target brain region, achieving maximum treatment coverage.

[0069] Electric field and VTA simulation (e.g.) Figure 4 (as shown)

[0070] The operator sets the electrode parameters (e.g., Medtronic 3389 electrode, 4 contacts, each 1.5 mm long, 0.5 mm apart) and stimulation parameters (e.g., selecting the most distal contact 2 as the cathode, voltage 3.0 V, pulse width 90 μs, frequency 130 Hz, tissue conductivity 0.2 S / m) in the system.

[0071] Step 1: Electric Field Distribution Calculation

[0072] The system starts the simulation calculation and performs the following detailed calculation steps:

[0073] (1) Computational Grid Establishment: Based on the finite element method and multi-tissue conductivity model, the system establishes a three-dimensional rectangular coordinate computational grid with a side length of 40 mm (±20 mm) using the electrode target coordinates as the geometric center. A uniform grid division strategy is adopted, with 50 nodes in each of the X, Y, and Z directions, forming a computational domain of 50×50×50=125,000 grid points. The grid spacing is approximately 0.8 mm to ensure sufficient spatial resolution at the brain tissue scale.

[0074] (2) Contact point position calculation: The system calculates the precise three-dimensional coordinates of the active contact point based on the electrode geometric parameters. Let the electrode entry point be P_entry, the target point be P_target, and the electrode direction vector be:

[0075]

[0076] For the i-th contact point (i∈the set of active contact points), its position is calculated as follows:

[0077]

[0078] Where L_spacing is the contact point spacing and L_contact is the contact point length.

[0079] (3) Establishment of multi-tissue conductivity model: The system first automatically segments different brain tissue types based on the signal intensity characteristics of MR images. Using a multi-threshold segmentation algorithm combined with morphological post-processing, brain tissue is divided into the following main types:

[0080] a) Gray matter region: mainly contains neuronal cell bodies, with an electrical conductivity σ_gray = 0.33 S / m;

[0081] b) White matter region: mainly contains myelinated axons, with anisotropic conductivity characteristics, longitudinal conductivity σ_white_parallel = 0.6 S / m, and transverse conductivity σ_white_perpendicular = 0.08 S / m;

[0082] c) Cerebrospinal fluid region: mainly in the ventricles and subarachnoid space, with a conductivity σ_CSF = 1.79 S / m;

[0083] d) Deep nuclei region: including the thalamus, basal ganglia, etc., with an electrical conductivity σ_nuclei = 0.23 S / m;

[0084] e) Vascular region: Based on TOF-MRA or DSA image recognition, conductivity σ_vessel = 0.7 S / m.

[0085] (4) Calculation of anisotropic conductivity tensor: For the white matter region, the system uses diffusion tensor imaging (DTI) data (if available) or a statistical model based on the white matter fiber orientation to determine the conductivity tensor. Let the main fiber orientation be the unit vector v, then the white matter anisotropic conductivity tensor is:

[0086]

[0087] Where I is the identity matrix. It is the tensor outer product.

[0088] (5) Electric potential field calculation: The system calculates the electric potential value for each grid point (x,y,z) according to its tissue type. Based on the point charge model, the electric potential distribution generated by a single contact point in a non-uniform tissue needs to consider the local conductivity:

[0089]

[0090] Where U is the stimulation voltage amplitude, σ_local(x,y,z) is the local tissue conductivity at grid point (x,y,z), and r_i is the Euclidean distance from the grid point to the i-th contact point.

[0091]

[0092] To avoid numerical singularities, the system sets a minimum distance threshold r_min = 0.1 mm. For the anisotropic case of the white matter region, the system uses the finite element method to solve the Poisson equation. To obtain a more accurate potential distribution. Among them, is the spatial gradient operator for the electric potential V, representing the rate of change of the electric potential in each coordinate direction in three-dimensional space. I is the current intensity injected at the electrode, referring to the constant or time-varying current applied in the model as a point source. yes A function is used to strictly confine the point current source to the geometric center of the electrode; the total potential is the superposition of all active contact points:

[0093] .

[0094] (6) Electric field strength calculation: The system uses the finite difference method to calculate the electric field strength. For isotropic regions, the electric field strength is calculated based on the relationship between electric field and electric potential. Calculate the electric field components in three directions at each grid point:

[0095]

[0096]

[0097]

[0098] in The grid spacing is for the corresponding direction. For anisotropic white matter regions, the system uses tensor form for calculation: This is to accurately reflect the influence of the fiber orientation on the propagation of the electric field. The electric field strength amplitude is calculated as follows: .

[0099] Step 2: VTA calculation using the Hodgkin-Huxley neuron model

[0100] The system invokes a complete Hodgkin-Huxley neuron model for neural activation simulation. This model includes the following key parameters: membrane capacitance Cm = 1.0 μF / cm², sodium ion channel conductance gNa = 120 mS / cm², potassium ion channel conductance gK = 36 mS / cm², leakage current conductance gL = 0.3 mS / cm², and corresponding equilibrium potentials ENa = 50 mV, EK = -77 mV, and EL = -54.4 mV. The system solves a system of four coupled differential equations:

[0101] Membrane potential equation: dV / dt = (1 / Cm)×(-INa - IK - IL + Istim)

[0102] Sodium channel activation gating: dm / dt = αm×(1-m) - βm×m

[0103] Sodium channel inactivation gating: dh / dt = αh×(1-h) - βh×h

[0104] Potassium channel gating: dn / dt = αn×(1-n) - βn×n

[0105] The rate constants of each gated variable are as follows:

[0106] αm = 0.1×(V+40) / (1-exp(-(V+40) / 10))

[0107] βm = 4.0×exp(-(V+65) / 18)

[0108] αh = 0.07×exp(-(V+65) / 20)

[0109] βh = 1.0 / (1+exp(-(V+35) / 10))

[0110] αn = 0.01×(V+55) / (1-exp(-(V+55) / 10))

[0111] βn = 0.125×exp(-(V+65) / 80)

[0112] Ion current calculation: m is the sodium channel activation gating variable, representing the probability that the sodium channel activation gate is open; h is the sodium channel inactivation gating variable, representing the probability that the sodium channel inactivation gate is not yet closed; n is the potassium channel activation gating variable, representing the probability that the potassium channel activation gate is open.

[0113] Stimulation current: Istim(t) = U×exp(-t / τ), where τ is the pulse width parameter and t is the simulation time variable, representing the current time point in the model calculation.

[0114] The system uses the fourth-order Runge-Kutta method (RK45) to solve the system of differential equations within a 10-millisecond time window, with a time step of 10 μs, and calculates a total of 1000 time points. For each spatial location within the electric field coverage area, the system determines whether the neuron membrane potential exceeds the activation threshold of -50 mV, and all activation locations constitute the tissue activation volume (VTA).

[0115] The Hodgkin-Huxley neuron model is a set of nonlinear differential equations describing the dynamic changes in neuronal membrane potential. By numerically solving these equations, it is possible to accurately simulate whether each neuron around the electrode will generate an action potential under the influence of the aforementioned electric field. The region comprised of all neurons generating action potentials is the tissue activation volume (VTA). This simulation can help researchers predict the therapeutic range under different stimulation parameters, thereby enabling optimization.

[0116] Step 3: Current Distribution Calculation

[0117] Based on Ohm's law J = σE, the system calculates the current density distribution in the tissue. Twenty seed points are uniformly distributed within a 5 mm radius around the electrodes. A streamline tracing algorithm (maximum propagation distance 20 mm, integration step size 0.1 mm, bidirectional tracing) is used to generate current streamlines. The system uses a pipe filter (0.1 mm radius, 12-sided cross-section) to visualize the streamlines as a blue semi-transparent tubular structure, visually demonstrating the three-dimensional flow path of current in brain tissue.

[0118] Step 4: Comprehensive Analysis of Simulation Results

[0119] The system automatically calculates the VTA radius based on the stimulus parameters: Where U is the voltage amplitude and τ is the pulse width. In this example, a 3.0V voltage and a 90μs pulse width produce a spherical VTA with a radius of approximately 3.67 mm. This VTA effectively covers the dorsolateral portion of the STN under the current parameters, and the electric field strength remains within the effective stimulation range of 200-600 V / m in the target area, achieving the expected therapeutic effect.

[0120] 4. Step 3: Postoperative electrode reconstruction and verification

[0121] like Figure 5 As shown, data acquisition: assuming the planned DBS electrodes were successfully implanted into the pig's brain, CT images of the head containing electrode artifacts were acquired during a CT scan of the pig and saved in NIfTI format.

[0122] The operator loads the pig's postoperative CT images and preoperative MR images into the system.

[0123] The system initiates the postoperative reconstruction module to perform CT-MR hybrid registration. The postoperative reconstruction module is used for precise three-dimensional reconstruction and verification of the actual location of the implanted electrodes. This module employs an innovative "automatic + manual" hybrid registration strategy to register postoperative CT images (clearly displaying high-density electrodes) and preoperative MR images (clearly displaying soft tissue anatomy) of the experimental pigs.

[0124] CT preprocessing: The system first processes the CT images to enhance the contrast between brain tissue and electrodes, adjusts the CT value window width / window level to 40 / 100 HU, and applies a gamma correction of 0.8 to make the soft tissue structure clearer and suppress registration interference caused by some skull bones.

[0125] Automatic Registration: The system initiates a 5-level multi-resolution registration process. Initial alignment is performed on coarse-resolution images, gradually transitioning to fine-tuning on high-resolution images. Using the L-BFGS-B optimizer, with mutual information as the metric, automatic registration is performed on CT and MR images, starting with images downsampled by 16x and gradually moving to the original resolution. After approximately 30 seconds, the system reports that automatic registration is complete, and the electrodes are roughly in the correct positions within the brain in the 3D view. This step can quickly provide an initial registration result close to the optimal solution.

[0126] Manual Fine-tuning: The system automatically switches to the manual fine-tuning interface. The system provides a graphical user interface. On this interface, users can simultaneously browse images in axial, coronal, and sagittal planes, and compare and observe the registered CT and MR images in fused, checkerboard, or side-by-side configurations. Users can perform six-degree-of-freedom fine-tuning of the CT images using the mouse and keyboard for rotation and translation. For example, if an operator finds a slight deviation of approximately 2 degrees in the axial rotation of the electrodes in the coronal view, rotating the CT image by -2 degrees along the Z-axis and translating it forward by 0.5 mm, the skull contour on the CT image perfectly matches the scalp contour on the MR image in the fused view, confirming that the registration accuracy meets the requirements. This "expert-in-the-loop" mode combines computer speed with human visual judgment, overcoming the interference of CT metal artifacts on the fully automated algorithm and achieving high-precision alignment.

[0127] Once registration is complete, the electrode model reconstructed on CT can be precisely superimposed on the preoperative MR image and compared with the planned target and trajectory to quantitatively assess the path error.

[0128] Result verification:

[0129] The system automatically segments and reconstructs a three-dimensional model of the DBS electrode based on the high-density metal signal in the CT image.

[0130] The electrode model was precisely overlaid on the preoperative MR images and the automatically segmented STN target area.

[0131] The system automatically calculates and reports: the Euclidean distance between the actual electrode target point and the planned target point is 1.2 mm, and the angle between the actual trajectory and the planned trajectory is 3.1 degrees.

[0132] The results indicate that the accuracy of the DBS implantation was within an acceptable range, validating the success of the planning. The entire closed-loop process from planning to validation is now complete.

[0133] The above embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A pig deep brain stimulation (DBS) electrode planning and reconstruction system, characterized in that, The system comprises: a atlas registration module for registering a preoperative magnetic resonance (MR) image of a pig with a standard pig brain digital atlas to automatically identify one or more target brain regions on the MR image; a trajectory planning module for calculating an optimal electrode implantation trajectory based on position and shape information of the target brain regions and performing electric field simulation on electrodes implanted along the trajectory; the trajectory planning module calculates an optimal electrode implantation trajectory to enable a plurality of stimulating contacts of the electrodes to cover an effective area of the target brain regions to the greatest extent, and the specific method comprises: 1) when the target brain regions are single brain regions: all voxels of the target brain regions in three-dimensional space are extracted to form a voxel point cloud; principal component analysis (PCA) is performed on coordinates of the voxel point cloud to determine a first principal component direction of the voxel point cloud as a longest principal axis of the target brain regions; the longest principal axis is defined as a direction of the optimal electrode implantation trajectory, and two end points of the voxel point cloud on the longest principal axis are calculated as a theoretical entry point and a target point of the electrodes; 2) when the target brain regions are two brain regions: centroids of the two target brain regions are calculated, a deep brain region with a relatively smaller volume is determined as the target point and a superficial brain region with a relatively larger volume is determined as the entry point direction according to the volume of the target brain regions, and an electrode trajectory is formed by directly connecting the two centroids, and an electrode entry point is extended by a preset distance outside the centroid of the superficial brain region; 3) when the target brain regions are more than two brain regions: centroids of the target brain regions are calculated, and coordinates of all the centroids form a centroid point cloud; principal component analysis (PCA) is performed on the centroid point cloud to determine a first principal component direction as an optimal trajectory direction, two extreme points of the centroid point cloud along the optimal trajectory direction are calculated, and an electrode entry point and a target point are extended by a preset distance outside the end closer to the skull; a postoperative reconstruction module for registering a postoperative computed tomography (CT) image of the pig with the preoperative MR image to reconstruct and verify an actual implantation position of the electrodes in the preoperative MR image space.

2. The pig deep brain stimulation (DBS) electrode planning and reconstruction system of claim 1, wherein, The registration process performed by the atlas registration module comprises: 1) initial alignment of the preoperative MR image and the standard pig brain digital atlas through rigid or affine transformation; 2) performing non-linear fine registration based on B-spline or contour line registration Demons algorithm to achieve high-precision matching of the atlas and the anatomical structure of the MR image to adapt to anatomical differences between individuals.

3. The pig deep brain stimulation (DBS) electrode planning and reconstruction system of claim 1, wherein, The method for electric field simulation performed by the trajectory planning module comprises: 1) establishing a three-dimensional calculation grid around the virtually implanted electrodes; 2) automatically segmenting brain tissues based on MR image signal intensity characteristics, establishing a multi-tissue conductivity model, including gray matter, white matter, cerebrospinal fluid, deep nuclei and blood vessels; 3) Constructing anisotropic conductivity tensors for white matter regions based on the fiber directions to compute the directional distribution of conductivity; 4) solving Poisson equation in inhomogeneous medium by finite element method calculating the electric potential distribution, and through calculating the electric field intensity distribution; wherein, is the local conductivity, which is the fiber bundle anisotropy tensor in white matter area and scalar in other tissues; V is the electric potential, unit V; I is the injected current intensity, unit A; is a function for introducing a point current source at the electrode position; is the spatial gradient operator; E is the electric field intensity, unit V / m; 5) using the Hodgkin-Huxley neuron model to solve the differential equation response of the neuron membrane potential to the electric field stimulation, and determining the tissue activation volume (VTA) according to whether the membrane potential reaches the activation threshold.

4. The pig deep brain stimulation (DBS) electrode planning and reconstruction system of claim 3, wherein, The Hodgkin-Huxley neuron model is a set of nonlinear differential equations describing the dynamic changes of the membrane potential of neurons. By numerically solving the equations, it is possible to accurately simulate whether each neuron in the electric field will generate an action potential. The area formed by all the neurons that generate action potentials is the volume of tissue activation (VTA).

5. The pig deep brain stimulation (DBS) electrode planning and reconstruction system of claim 1, wherein, The registration process performed by the postoperative reconstruction module is a hybrid registration method, including: 1) CT image preprocessing: By setting a soft tissue window and applying gamma correction, the contrast of soft tissue is enhanced and skull artifacts are suppressed; 2) Automatic registration stage: Using a Mutual Information-based metric and a limited memory L-BFGS-B optimization algorithm, the postoperative CT image and the preoperative MR image are automatically initially registered on a multi-resolution image pyramid; 3) Manual fine-tuning stage: A graphical user interface is provided to allow the operator to simultaneously observe and display the registered CT and MR images in axial, coronal and sagittal views, and to perform six-degree-of-freedom manual fine-tuning of the automatic registration results, including rotation and translation, to achieve the final accurate alignment.

6. The pig deep brain stimulation (DBS) electrode planning and reconstruction system of claim 5, wherein, The preprocessing of the CT image is used to enhance the contrast between the brain tissue and the electrode. Specifically, a soft tissue window is used to truncate the CT values, and gamma correction is performed to highlight the soft tissue details and suppress the registration interference caused by part of the skull.

7. The pig deep brain stimulation (DBS) electrode planning and reconstruction system of claim 5, wherein, The automatic registration stage uses a multi-resolution registration framework that performs initial alignment on coarse resolution images and gradually transitions to high resolution images for fine matching. The similarity metric for registration uses Mutual Information, which is robust to multi-modal images, and the optimizer uses the L-BFGS-B algorithm.

8. The pig deep brain stimulation (DBS) electrode planning and reconstruction system of claim 1, wherein, After the registration process in the postoperative reconstruction module is completed, the electrode model reconstructed on the CT can be accurately superimposed and displayed on the preoperative MR image, compared with the planned target and trajectory, and the path error can be quantitatively evaluated.

9. A method of pig deep brain stimulation (DBS) electrode planning and reconstruction based on the system of any one of claims 1-8, characterized in that, The method comprises the following steps: (1) Registering the preoperative MR image of the pig with a standard pig brain digital atlas to automatically identify one or more target brain regions on the MR image; (2) Based on the location and shape information of the target brain region, calculate the optimal electrode implantation trajectory and perform electric field simulation on the electrodes implanted along the trajectory; (3) Register the postoperative CT image of the pig with the preoperative MR image to reconstruct and verify the actual implantation position of the electrodes in the preoperative MR image space.

Images