Intraoperative brain tissue real-time imaging method and system based on broadband ultrasound array

Abstract

This invention discloses a method and system for real-time intraoperative brain tissue imaging based on a broadband ultrasound array, belonging to the fields of medical imaging and neurosurgical navigation. It addresses problems such as navigation inaccuracies caused by brain drift, poor probe compatibility, and insufficient imaging accuracy. The system includes a patch-type flexible broadband ultrasound probe, a communication module, a data processing center, and a display device. The probe can be stably attached to the surgical field, accommodating both deep probe exploration and superficial imaging; communication supports wired / wireless anti-interference transmission. The data processing center generates a voxel-level three-dimensional displacement field, which, after non-rigid spatial correction, is registered and fused with preoperative MRI / CT images to generate precise navigation images. This invention can achieve quantification and sub-millimeter-level correction of brain drift movement, is adaptable to various craniocerebral surgeries, improves navigation accuracy and surgical safety, has strong clinical adaptability, and good application value.

Description

Technical Field

[0001] This invention relates to the fields of medical imaging and neurosurgical navigation technology. Specifically, it relates to a method and system for real-time intraoperative brain tissue imaging based on a broadband ultrasound array, which can achieve high-resolution, continuous dynamic real-time imaging of brain tissue during surgery and perform dynamic correction of brain drift throughout the procedure, adapting to various intraoperative navigation-related application scenarios in neurosurgery. Background Technology

[0002] Neuronavigation technology is a crucial support for the precision and minimally invasive nature of neurosurgery. In surgeries involving the resection of brain tumors, cerebral hemorrhage, and cerebral vascular malformations, it provides technical support for the operation by locating the lesion and protecting surrounding normal neurovascular structures. Current neuronavigation systems have evolved from traditional framed stereotactic systems to frameless stereotactic systems primarily based on optical and electromagnetic navigation, and have also spawned new technologies such as virtual navigation and surgical robots. The positioning accuracy of various navigation systems can reach the millimeter level, providing positioning support for most neurosurgical procedures.

[0003] However, intraoperative brain tissue displacement (i.e., "brain drift") is a core technical factor affecting further improvements in the accuracy of neuronavigation. This displacement is mainly caused by factors such as cerebrospinal fluid loss, brain tissue traction, tumor resection, and dynamic changes in intracranial pressure during surgery. It manifests in various forms of displacement, such as brain tissue rebound, collapse, and traction, and is characterized by continuity, dynamism, and subtlety. This can cause deviations between preoperative static images such as MRI and CT and the actual anatomical morphology of the brain during surgery, making it difficult to continuously adapt the preoperatively planned navigation information and potentially increasing the risk of surgical instruments contacting functional areas of the brain.

[0004] To address the navigation information deviation caused by brain drift, intraoperative imaging technology has become a key means to dynamically correct brain drift and update anatomical information during surgery. Among them, intraoperative ultrasound navigation, with its real-time imaging, portable equipment, and no ionizing radiation, has become an important supplementary navigation method in clinical practice. It is widely used in intraoperative localization of intracranial lesions and monitoring of lesion resection process, and is one of the important technical directions for brain drift correction.

[0005] Currently, some progress has been made in the research on brain drift correction and flexible ultrasound imaging, among which US7072705B2 and US7878977B2 are two core technologies in this field. US7072705B2 discloses a brain drift compensation system based on preoperative MRI. It updates navigation images by measuring brain displacement intraoperatively and constructing a displacement model using statistical or finite element models. This is the first time a technical approach of quantifying brain drift using displacement models has been proposed. However, this technology uses an indirect modeling method and lacks real-time intraoperative displacement data, resulting in limited correction effects. Furthermore, it is not integrated with ultrasound imaging technology to form an integrated correction scheme, making it difficult to achieve dynamic real-time correction of brain drift. US7878977B2 discloses a flexible ultrasound array technology. It uses a flexible structure design to solve the compatibility problem of traditional rigid ultrasound probes not being able to fit the irregular curved surfaces of the human body. However, it does not design an attachment structure for the characteristics of the neurosurgical surgical field, nor does it combine broadband imaging, high-density array elements, and other technologies. This makes it difficult to capture the micron-level subtle displacements caused by brain drift, and it has not been applied to brain drift detection and brain navigation scenarios. There is a disconnect between hardware design and clinical needs.

[0006] In addition to the two core patented technologies mentioned above, although related research in this field has proposed registering intraoperative ultrasound with MRI and achieving brain drift correction through 3D ultrasound image registration, and existing neurosurgical navigation systems are gradually supporting real-time navigation modes with ultrasound + MRI registration, there are still many areas for improvement. There is still room for improvement in adapting to the clinical needs of precise neurosurgical procedures for high-precision dynamic correction of brain drift throughout the entire process. 1. The rigid probes used in traditional intraoperative ultrasound navigation are difficult to fit into irregular cranial sutures or soft brain tissue surfaces, limiting the operation within the surgical field. Furthermore, the limited number of probes results in low imaging resolution and makes it difficult to capture subtle displacements of brain drift. 2. Signal transmission mostly uses conventional wired methods. Cables are easily tangled in surgical instruments and obstruct the surgical field. Pulling the cable can also cause the probe to shift, affecting the continuity of brain drift detection. Wireless transmission solutions also lack targeted designs for low latency, high bandwidth, and resistance to intraoperative electromagnetic interference. 3. The probe's emission frequency is fixed, which cannot meet the low-frequency requirements of skull penetration and the high-frequency requirements of superficial fine structure recognition, and its ability to capture subtle brain drift displacement is insufficient. 4. The brain drift correction algorithm is only described in terms of functionality, without disclosing the specific implementation steps and core mathematical model. This results in insufficient disclosure and cannot provide effective support for the claims. Furthermore, the algorithm's compatibility with ultrasound hardware needs to be improved, making it difficult to achieve sub-millimeter level correction effects. 5. The integration of brain drift correction and multimodal fusion is low. The fusion of intraoperative ultrasound data and preoperative MRI / CT images is mostly done offline, which takes a long time. Moreover, the fusion results do not incorporate real-time brain drift data, resulting in large registration bias. The navigation-related effects need to be optimized.

[0007] In summary, current technologies lack a comprehensive intraoperative brain tissue imaging solution that organically combines attachable flexible ultrasound acquisition structures, broadband segmented imaging, and brain drift correction based on intraoperative measured data. Especially in neurosurgical settings, existing technologies still have room for improvement in areas such as surgical field attachment stability, imaging of both superficial and deep tissues, quantitative modeling of brain drift, and real-time fusion of preoperative and intraoperative images.

[0008] Therefore, this invention provides a method and system for real-time intraoperative brain tissue imaging based on a broadband ultrasound array. By constructing a patch-type flexible broadband ultrasound array probe, a unified coordinate system processing mechanism for ultrasound echoes of different frequency bands, and a non-rigid spatial correction and multimodal fusion process for brain tissue based on voxel-level displacement field, the method improves the ability of intraoperative brain tissue imaging to track and compensate for brain drift.

[0009] The technical solution of this invention revolves around the following technical ideas: First, a flexible broadband ultrasound array probe that can be attached to the surface of the surgical field is used to improve the stability of intraoperative signal acquisition; second, deep tissue displacement reference and superficial tissue local deformation information are obtained based on ultrasound signals of different frequency bands, and fused in a unified voxel coordinate system; third, a voxel-level three-dimensional displacement field is generated based on continuous frame ultrasound data, and non-rigid spatial correction is performed on the three-dimensional image of brain tissue accordingly, and then registered and fused with preoperative images to generate intraoperative navigation images. Summary of the Invention Technical problems to be solved

[0010] This invention aims to overcome the shortcomings of existing technologies and solve technical problems in current intraoperative ultrasound navigation for neurosurgery, such as poor probe compatibility, interference of wired transmission with surgical operations, insufficient disclosure of brain drift correction algorithms with limited correction effects, lack of an integrated processing system for brain drift quantification and correction, single imaging frequency, and insufficient real-time performance of multimodal fusion. It provides a method for real-time intraoperative brain tissue imaging based on a broadband ultrasound array and constructs a corresponding imaging system to implement this method. This system forms an integrated solution for full-process brain drift correction from multiple levels, including hardware adaptation, signal transmission, and processing logic design. By capturing and quantifying brain drift and achieving real-time compensation of the imaging coordinate system, it improves the navigation accuracy reduction caused by brain drift. Furthermore, it supplements the core algorithm for brain drift correction with specific implementation steps and key mathematical models, providing sufficient technical support for the claims. Technical solution

[0011] To achieve the above objectives, this invention provides an intraoperative real-time brain tissue imaging system based on a broadband ultrasound array, and also provides an imaging method using this system. The technical solution is designed around the entire process of brain drift correction. From hardware attachment and signal acquisition to data processing and image fusion, each link provides technical support for brain drift correction, forming an integrated technical system in which hardware and software are deeply integrated and work together.

[0012] A real-time intraoperative brain tissue imaging system based on a broadband ultrasound array includes a patch-type flexible broadband ultrasound array probe, a communication module, a data processing center, and a display device.

[0013] The patch-type flexible broadband ultrasound array probe includes a flexible substrate and a high-density ultrasound element array disposed on the flexible substrate. The edge of the flexible substrate is provided with a fixation structure for fixing the probe to the cranial sutures within the surgical field or to the surface of exposed brain tissue. The flexible substrate can adapt to the shape changes of the surgical field surface to improve the adhesion stability between the probe and the tissue during intraoperative acquisition.

[0014] The high-density ultrasound array includes multiple independent transmit and receive channels, employing either piezoelectric micromechanical ultrasound transducer (pMUT) elements or capacitive micromechanical ultrasound transducer (cMUT) elements. The high-density ultrasound array includes a first-band element and a second-band element. The first-band element is used to transmit and receive low-frequency ultrasound signals, while the second-band element is used to transmit and receive high-frequency ultrasound signals, thus accommodating both deep tissue signal acquisition and superficial tissue detail imaging.

[0015] In some implementations, the first frequency band array element and the second frequency band array element are disposed on the same flexible substrate and share the same probe coordinate system; the different frequency band array elements can be staggered, grouped, or excited by subarrays to realize the transmission and reception of ultrasonic signals in different frequency bands.

[0016] The communication module is used to send control commands and transmit ultrasonic echo signals to the patch-type flexible broadband ultrasonic array probe. The communication module can employ wireless communication and / or wired communication. Wireless communication may include one or more of 5G, WiFi, Bluetooth, and ultra-wideband (UWB) communication, while wired communication may include flexible cable communication. Preferably, the communication module has time synchronization, electromagnetic interference immunity, and data verification functions.

[0017] The data processing center includes a processor and a memory, the memory storing data processing programs for performing brain tissue motion tracking, image correction, and registration fusion. The data processing center is configured to perform noise reduction on ultrasound echo signals, generate a voxel-level three-dimensional displacement field, perform non-rigid spatial correction on the three-dimensional brain tissue image based on the voxel-level three-dimensional displacement field, and register and fuse the corrected three-dimensional brain tissue image with preoperative images.

[0018] The display device is used to display the corrected and fused intraoperative navigation images. The display device may be a medical monitor, a head-mounted display device, or other visualization device suitable for intraoperative observation.

[0019] In some embodiments, the system includes multiple patch-type flexible broadband ultrasonic array probes. The data processing center stores the spatial calibration parameters of each probe and resamples the volume data collected by each probe to a unified voxel coordinate system before performing fusion processing.

[0020] Furthermore, the fixation structure is one or more of medical sterile adhesive tape, auxiliary suture point, or clamping fixation structure; the flexible substrate is made of silicone rubber or polyimide material with a thickness of 0.2mm to 1.0mm, which has both flexibility and support, and can adapt to the irregular shape of the surgical field surface.

[0021] Furthermore, the high-density ultrasound array integrates multiple independently transmitting and receiving ultrasound elements, with more than 1,000 elements, to improve spatial sampling density and provide high signal-to-noise ratio raw data for voxel-level displacement field calculation; the first frequency band element operates at a frequency of 2MHz to 5MHz, which can be used for skull penetration and deep brain tissue signal acquisition; the second frequency band element operates at a frequency of 5MHz to 12MHz, which can be used for superficial high-resolution imaging of the cerebral cortex and lesion boundaries.

[0022] Furthermore, the communication module includes a data frame encapsulation module and an anti-electromagnetic interference link module. The data frame contains a frame header, probe ID, timestamp, array element number, RF echo data block, and checksum. The timestamp is used for time alignment of data acquired by different probes and / or different array elements. The anti-electromagnetic interference link module employs one or more of differential transmission, shielding isolation, and forward error correction coding (FEC) to ensure stable transmission of the ultrasound echo signal during surgery. The wired communication method uses an ultra-fine flexible cable with a diameter ≤1.0mm to reduce the possibility of probe displacement due to traction.

[0023] In some embodiments, the data processing center can perform brain drift correction and registration fusion according to the following processing flow: including voxel-by-voxel correlation calculation, non-rigid spatial correction based on displacement field, and registration fusion of preoperative and intraoperative images. The relevant processing procedures are described below with reference to one embodiment.

[0024] The voxel-by-voxel correlation calculation is used to generate a voxel-level three-dimensional displacement field based on the echo signal after noise reduction of consecutive frames. As one implementation, a correlation window for the corresponding voxel position can be selected in the current frame and reference frame images. A similarity metric is calculated within a preset three-dimensional search range, and voxel displacement estimation is determined through sub-pixel interpolation. Then, the displacement field is smoothed and constrained to obtain a voxel-level three-dimensional displacement field describing brain tissue deformation. The specific steps are as follows: ① Relevant window selection: For the echo images of the current frame and the previous frame, select an N×N×N cubic relevant window (N is an odd number from 3 to 7) at the position of the voxel to be detected. Adopt an equally spaced sliding design, slide along the image voxel grid at a preset step size (step size ≤ N / 2) to retain the spatial correlation information of brain tissue deformation and ensure the continuity of voxel-level detection. ② Normalized Cross-Correlation Coefficient Calculation: Within the preset 3D search range, the normalized cross-correlation coefficient (NCC) between the current frame window and the reference frame window is calculated position by position to quantify the similarity between the two windows and overcome the influence of the grayscale variation of the ultrasonic echo signal. The NCC calculation formula is as follows: In the formula, W represents the selected N×N×N cubic correlation window; (u,v,w) are the displacement vectors of the window in the x, y, and z directions, representing the displacement components of the current frame window relative to the reference frame window; (x, y, z) are the coordinate positions of the voxels within the window; I_t and I_{t-1} are the echo amplitude values ​​at the corresponding coordinate positions in the current and reference frame images, respectively; \bar{I}_t and \bar{I}_{t-1} are the mean echo amplitudes of all voxels within the current and reference frame windows, respectively. The closer the NCC value is to 1, the higher the similarity between the two windows, and its maximum value corresponds to the initial displacement estimate of the voxels. ③ Subpixel interpolation to obtain displacement estimation: Select the integer pixel displacement vector corresponding to the maximum NCC value, perform cubic polynomial interpolation on the NCC values ​​of its surrounding neighborhood, fit a continuous distribution surface, and obtain the subpixel level displacement vector corresponding to the peak of the surface, thereby improving the displacement detection accuracy from the pixel level to the subpixel level. ④ Three-dimensional Gaussian smoothing regularization constraint: Apply three-dimensional Gaussian smoothing filter to the original displacement field composed of sub-pixel displacement vectors of all voxels. While preserving the real deformation characteristics, suppress ultrasonic speckle noise and isolated noise points caused by calculation errors, and ensure the spatial smoothness and continuity of the displacement field. ⑤ Output voxel-level three-dimensional displacement field: Integrate the three-dimensional displacement vectors of each voxel after Gaussian smoothing to generate a voxel-level three-dimensional displacement field covering the entire surgical field. Each vector value corresponds to the three-dimensional spatial displacement of the voxel, directly quantifying the deformation state of brain tissue.

[0025] The non-rigid spatial correction is used to perform geometric correction on the three-dimensional image of brain tissue based on the voxel-level three-dimensional displacement field. As one implementation method, a control point mesh can be constructed based on the voxel-level three-dimensional displacement field, the control point displacements can be solved, and a free deformation model can be used to interpolate the control point displacements to the voxel positions of the brain tissue image to obtain the corrected three-dimensional image of brain tissue. The specific steps are as follows: ① Constructing a control point grid: Using a voxel-level three-dimensional displacement field as input, a uniform three-dimensional control point grid is constructed in the global coordinate system of the three-dimensional brain tissue image. The grid spacing is set to 8 to 16 voxels according to the image resolution. The control points cover the entire imaging area to ensure the global effectiveness of the non-rigid transformation. ② Solve for control point displacement: Using the reference frame brain tissue image as the registration target, solve for the optimal displacement vector for each control point to optimize the consistency between the transformed image and the reference image, and ensure the continuity of the anatomical structure in the transformed image; ③ Global Coordinate Transformation Based on B-Spline Free Deformation: B-spline free deformation (FFD) is a smooth transformation method commonly used for non-rigid image registration, which can balance the accuracy of local deformation and the continuity of global transformation. This invention uses a third-order B-spline free deformation (FFD) model to interpolate the displacement of control points to each voxel point of the brain tissue image, thereby achieving global coordinate transformation and balancing the accuracy of local deformation and the smoothness of global transformation. The third-order B-spline basis functions are: B0(t)=(1-t) 3 / 6 B1(t)=(3t 3 -6t 2 +4) / 6 B2(t)=(-3t 3 +3t 2 +3t+1) / 6 B3(t)=t 3 / 6 Based on the above basis functions, the displacement of the control points is smoothly interpolated to all voxel points to obtain the transformed coordinates of each voxel; ④ Generate geometrically corrected image: Based on the coordinate transformation results of the B-spline free deformation, resample and interpolate the current frame brain tissue image to generate a geometrically corrected image that matches the actual deformation of the brain tissue, thus completing the non-rigid spatial correction of brain drift.

[0026] The registration and fusion is used to fuse the corrected three-dimensional brain tissue image with the preoperative image. As one implementation method, the anatomical feature point set from the preoperative image and the dynamic feature point set from the corrected intraoperative ultrasound image can be extracted first to solve the initial registration relationship. Then, the preoperative image is non-rigidly registered using the voxel-level three-dimensional displacement field to generate an intraoperative navigation image. The specific steps are as follows: ① Extraction of preoperative anatomical feature point set: Image enhancement and edge detection are performed on preoperative MRI / CT images to extract a unique and easily matched set of anatomical feature points, including vascular bifurcation points, sulcus intersections, anatomical landmarks, etc. The feature points are evenly distributed in space and cover the brain tissue areas related to the surgery. ② Extraction of dynamic feature point set of intraoperative ultrasound: Perform lesion boundary segmentation and contour extraction on the corrected intraoperative ultrasound three-dimensional image, and extract dynamic feature point set, including lesion boundary corner points, blood vessel centerline points, resection cavity contour points, etc., to ensure the robustness and matching of feature points; ③ RANSAC algorithm for matching feature point sets: Calculate the feature descriptors of two sets of feature points and obtain the initial matching pairs based on Euclidean distance; then use the RANSAC algorithm to remove outlier matching pairs, retain inlier matching pairs and solve the rigid registration transformation matrix to achieve coarse registration between preoperative images and intraoperative ultrasound images. ④ Deformation field driven non-rigid registration: Using the rigid registration transformation matrix as the initial transformation, the voxel-level three-dimensional displacement field is used as the deformation field constraint to perform non-rigid registration on the preoperative images, optimize the local mutual information between the preoperative images and the intraoperative ultrasound images, and make the anatomical structure of the preoperative images accurately match the actual morphology during the operation, thus completing the fine registration. ⑤ Generate intraoperative navigation images: The preoperative images after non-rigid registration and the corrected intraoperative ultrasound images are weighted and fused in a unified voxel coordinate system to generate intraoperative navigation images that combine detailed preoperative anatomical details with real-time intraoperative dynamic information.

[0027] It should be understood that the above-described voxel-by-voxel displacement estimation, non-rigid spatial correction, and registration fusion process is one of the preferred embodiments of the present invention. Those skilled in the art may also employ other similarity measurement methods, displacement estimation methods, spatial transformation methods, or registration methods adapted to the above processing flow without departing from the technical concept of the present invention.

[0028] Furthermore, the data processing center is configured to: acquire a global displacement reference for deep brain tissue based on low-frequency ultrasound signals; acquire local deformation information of superficial brain tissue based on high-frequency ultrasound signals; and perform multi-scale fusion of low-frequency and high-frequency ultrasound signals in a unified voxel coordinate system, using the low-frequency displacement reference as a global constraint and the high-frequency deformation information as a basis for detail correction, to generate a layered corrected voxel-level three-dimensional displacement field. In some embodiments, the data processing center is equipped with a GPU accelerator card, enabling real-time storage and traceability of brain drift data throughout the surgical procedure.

[0029] Furthermore, the spatial calibration parameters are obtained in the following way: the intrinsic parameter matrix of each probe is calibrated in advance using a calibration phantom containing multiple known reflection points; the relative extrinsic parameter matrix between each probe is obtained through the identification marks on the probe edges or cross-correlation matching of overlapping fields of view; the intrinsic parameter matrix and the extrinsic parameter matrix are combined to obtain the calibration parameter matrix of each probe to the unified voxel coordinate system. The data processing center is further configured to: perform periodic cross-correlation verification on the ultrasound images of the overlapping field of view of multiple probes; when the registration error in the overlapping area exceeds a preset threshold, automatically trigger a local optimization update of the relative extrinsic parameter matrix between probes to maintain the continuous validity of the unified voxel coordinate system.

[0030] Furthermore, the display device includes a medical high-definition display, smart glasses, or a head-mounted mixed reality device, supporting interactive operations such as navigation image magnification, three-dimensional rotation, and brain drift data overlay display, adapting to the clinical observation needs of different neurosurgical procedures.

[0031] A method for real-time intraoperative brain tissue imaging based on a broadband ultrasound array, utilizing the aforementioned real-time intraoperative brain tissue imaging system based on a broadband ultrasound array, includes the following steps: S1, Attachment and Positioning

[0032] One or more patch-type flexible broadband ultrasound array probes are fixed to the cranial sutures or exposed brain tissue surface within the patient's surgical field to improve the stability and continuity of intraoperative ultrasound signal acquisition. S2, Wideband Excitation and Signal Acquisition

[0033] The communication module sends control commands to the patch-type flexible broadband ultrasonic array probe, driving the ultrasonic array elements inside the probe to collect ultrasonic echo signals of different frequency bands, and transmits the ultrasonic echo signals to the data processing center.

[0034] In some embodiments, the ultrasound echo signals of different frequency bands include low-frequency ultrasound echo signals and high-frequency ultrasound echo signals; the low-frequency ultrasound echo signals are used to obtain overall displacement reference of deep brain tissue, and the high-frequency ultrasound echo signals are used to obtain local deformation information of superficial brain tissue.

[0035] Furthermore, the ultrasonic echo signals of different frequency bands include low-frequency ultrasonic echo signals of 2MHz to 5MHz and high-frequency ultrasonic echo signals of 5MHz to 12MHz, which are obtained through a time-division transmission strategy or a group transmission strategy. The time-division transmission strategy is to drive different array elements to emit ultrasonic waves of corresponding frequency bands in different time periods within the same imaging cycle. The group transmission strategy is to divide the array elements into a low-frequency transmission group and a high-frequency transmission group to simultaneously emit ultrasonic waves of corresponding frequency bands. S3, Real-time Imaging and Brain Drift Correction

[0036] The data processing center processes the received ultrasonic echo signals, including: (1) S3-1 signal noise reduction: noise reduction processing of echo signal; (2) S3-2 displacement quantization: Based on the continuous frame echo signal, a voxel-level three-dimensional displacement field describing brain tissue deformation is generated (refer to the voxel-by-voxel correlation calculation process described above); (3) S3-3 image correction: Based on the voxel-level three-dimensional displacement field, non-rigid spatial correction is performed on the three-dimensional image of brain tissue (refer to the B-spline free deformation process described above) to obtain a dynamic three-dimensional image corresponding to the intraoperative brain tissue morphology.

[0037] Furthermore, the generation of the voxel-level three-dimensional displacement field in step S3 can refer to the voxel-by-voxel correlation calculation process described above, specifically including sub-steps such as correlation window selection, normalized cross-correlation coefficient calculation, sub-pixel interpolation, and three-dimensional Gaussian smoothing filtering.

[0038] Furthermore, the non-rigid spatial correction in step S3 can refer to the B-spline free deformation process described above, specifically including sub-steps such as control point mesh construction, control point displacement solution, global coordinate transformation, and resampling interpolation. S4, Multimodal Fusion Navigation

[0039] The data processing center will register and fuse the corrected three-dimensional brain tissue image with the preoperative image (refer to the feature point matching and RANSAC process described above) to generate intraoperative navigation image, and output the intraoperative navigation image to the display device for display.

[0040] In some embodiments, the preoperative images include one or more of MRI images, CT images, and three-dimensional ultrasound images.

[0041] Furthermore, in step S4, the set of anatomical feature points in the preoperative image is matched with the set of dynamic feature points in the corrected intraoperative ultrasound image. The initial registration relationship is obtained through the RANSAC algorithm, and the non-rigid registration of the preoperative image and the intraoperative ultrasound image is completed by combining the voxel-level three-dimensional displacement field.

[0042] Furthermore, this method can be applied to craniocerebral tumor resection surgery, cerebral vascular malformation repair surgery, cerebral hemorrhage removal surgery, and deep brain lesion biopsy surgery. In some implementations, it can form corresponding dynamic correction effects for various brain drift types such as collapse type, rebound type, and traction type. Beneficial effects

[0043] Compared with the prior art, the embodiments of the present invention can achieve the following technical effects: (1) The present invention uses a patch-type flexible broadband ultrasound array probe, which can improve the probe's adhesion to the surgical field surface and thus improve the stability of ultrasound signal acquisition during surgery. At the same time, the setting of array elements of different frequency bands can take into account both deep tissue information acquisition and superficial tissue detail imaging. Among them, low frequency array elements can realize skull penetration related signal acquisition, and high frequency array elements can realize superficial detail imaging. (2) The present invention generates a voxel-level three-dimensional displacement field from continuous frame ultrasound data and performs non-rigid spatial correction on the three-dimensional image of brain tissue based on the displacement field, which can improve the continuity of intraoperative brain drift detection and compensation and improve the precision of displacement detection. (3) The present invention registers and fuses the corrected intraoperative ultrasound images with the preoperative images, and can simultaneously present preoperative anatomical information and intraoperative dynamic information in the intraoperative navigation images, thereby providing a reference for surgical operation. (4) The system structure of the present invention can be combined with the existing neurosurgical procedure and has corresponding clinical adaptation potential; the multi-probe collaborative operation and online verification mechanism can expand the surgical field coverage and maintain the stability of long-term monitoring, further enhancing the clinical application value; the flexible switching design between wireless and wired communication can reduce the interference of cables on surgical operations and provide safety-related support for the operation. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments 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 flowchart illustrating the intraoperative real-time brain tissue imaging method based on a broadband ultrasound array in an embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram illustrating the structure and attachment method of the patch-type flexible broadband ultrasonic array probe in an embodiment of the present invention;

[0047] Figure 3 This is a block diagram illustrating the principle of precise brain drift compensation and correction in the data processing center according to an embodiment of the present invention.

[0048] Figure 4 This is a schematic diagram of the overall structure of the intraoperative real-time brain tissue imaging system based on a broadband ultrasound array in an embodiment of the present invention;

[0049] Figure 5 This is a comparison of imaging effects before and after brain drift correction in an embodiment of the present invention;

[0050] Figure 6 This is a schematic diagram of the frame structure and anti-interference design of the communication module in an embodiment of the present invention.

[0051] Explanation of reference numerals in the attached figures: 1-Pattern-type flexible broadband ultrasound array probe; 11-Flexible substrate; 12-High-density ultrasound array; 13-Communication submodule; 14-Medical sterile adhesive patch; 2-Communication module; 3-Data processing center; 31-Real-time noise reduction module; 32-Brain drift correction algorithm module; 33-Multimodal registration and fusion module; 4-Display device; 41-Anti-electromagnetic interference link module; 42-Data frame encapsulation module.

[0052] Figure 1 The core steps are presented sequentially: S1 attachment and localization, S2 broadband excitation and signal acquisition, S3 real-time imaging and brain drift correction (including S3-1 signal denoising, S3-2 displacement quantization, and S3-3 image correction), and S4 multimodal fusion navigation. The integrated core link of "signal denoising-displacement quantization-image correction" in brain drift correction is highlighted, and the logical connection and data flow of each step are clearly presented, highlighting the core position of brain drift correction in the whole method.

[0053] Figure 2 In the diagram, 2a is a schematic diagram of a single probe structure, with the core components such as the flexible base, high-density ultrasound array elements, and communication submodule labeled; 2b is a schematic diagram of a multi-probe splicing array, demonstrating the expandability of the field of view; 2c is a schematic diagram of the probe attachment during surgery, which intuitively shows the attachment status of the probe to the cranial sutures and brain tissue surface during surgery, demonstrating the flexible adaptation advantage of the probe.

[0054] Figure 3 The text is a series of sections that demonstrate the input data, processing procedures, and output results of real-time noise reduction, tissue motion tracking (brain drift quantization), geometric correction, and non-rigid spatial transformation (brain drift correction). It clarifies the processing logic chain and data transformation relationship of brain drift correction and intuitively reflects the working principle of the integrated correction system.

[0055] Figure 4The text demonstrates the physical connections and signal transmission paths of the patch-type flexible broadband ultrasound array probe, communication module, data processing center, and display device, and labels the functional associations of each module with brain drift correction. For example, the probe is labeled "brain drift signal acquisition", the data processing center is labeled "brain drift quantification and correction", and the display device is labeled "correction result output", clearly presenting the modular architecture of the system designed for brain drift correction.

[0056] Figure 5 In the image, 5a is the uncorrected drift image, showing a deviation between the lesion boundary and the location of the brain functional area, resulting in a corresponding deviation in the navigation information; 5b is the image corrected by the method of this invention, showing that the lesion boundary and the actual location of the functional area are more closely related, and the navigation information is more in line with the actual surgical needs.

[0057] Figure 6 The core components of the communication module are shown, including the frame structure of the data frame encapsulation module (including frame header, probe ID, timestamp, etc.) and the design of the anti-electromagnetic interference link module (including FEC, CRC, etc.), demonstrating the anti-interference and data synchronization capabilities of the communication module. Detailed Implementation

[0058] 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. 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. Example 1: Intraoperative Real-Time Brain Tissue Imaging Method Based on Broadband Ultrasound Array (Application in Glioma Resection Surgery, Correction of Collapsed Brain Drift)

[0059] This embodiment applies the method of the present invention to glioma resection surgery, aiming to achieve a dynamic correction effect on the collapsed brain drift caused by tumor resection. The specific process is as follows: Figure 1 As shown, it includes the following steps: S1, Attachment and Positioning

[0060] like Figure 2 As shown in Figure c, after the patient underwent craniotomy and dura mater incision, the surgeon attached two pre-sterilized patch-type flexible broadband ultrasound array probes 1 to the cranial window margin and the exposed brain tissue surface around the glioma within the surgical field; combined with Figure 2 A schematic diagram of a single probe structure is shown. The probe 1 is made of a medical-grade biocompatible silicone rubber flexible substrate 11, with a single probe size of 3cm × 3cm. It integrates 1500 capacitive micromechanical ultrasonic transducer elements 12. The probe edge is equipped with a medical-grade sterile adhesive patch 14 for fixation, supporting the collaborative operation of two probes to form a... Figure 2The imaging array shown in b has a flexible probe that fits closely to the surface of the brain sulci and gyri, eliminating the need for handheld fixation and maintaining signal stability for detecting collapsed brain drift. S2, Wideband Excitation and Signal Acquisition

[0061] After starting the imaging system, as Figure 4 As shown, the data processing center 3 sends control commands to the probe 1 through the 5G communication module 2. This communication module 2 is equipped with a medical-grade electromagnetic interference-resistant link, adopts differential transmission and forward error correction coding (FEC), and has the ability to resist intraoperative electromagnetic interference. The system adopts a time-division transmission strategy. In a 20ms imaging cycle, the array element is driven to emit 2MHz low-frequency ultrasound in the first 10ms to achieve deep brain tissue penetration detection, and the array element is driven to emit 10MHz high-frequency ultrasound in the next 10ms to finely scan the boundary of the glioma lesion and capture the subtle displacement of brain drift. The probe 1 receives echo signals from brain tissue at different depths and transmits the raw radio frequency data back to the data processing center 3 in real time through the 5G communication module 2, which can reduce the interference of cables on the surgical operation and maintain the continuity of brain drift signal acquisition. S3, Real-time Imaging and Brain Drift Correction

[0062] like Figure 3 As shown, the data processing center 3 is equipped with a GPU accelerator card and a built-in brain drift correction algorithm module 32, which processes the received echo signals in real time to achieve a dynamic correction effect for collapse-type brain drift caused by tumor resection. (1) S3-1 signal denoising: The echo signal is filtered and denoised in real time by the real-time denoising processing module 31 of the data processing center 3, the periodic electromagnetic interference noise generated by the electrosurgical unit during the operation is identified and eliminated, the clarity and signal-to-noise ratio of the original echo signal are improved, and a high-quality signal basis is provided for the quantitative detection of collapsed brain drift. (2) S3-2 Displacement Quantization: Based on the denoised continuous frame echo signal, voxel-by-voxel correlation calculation is performed. A 5×5×5 correlation window is selected, and an equally spaced sliding design (step size = 2) is adopted. A dynamic three-dimensional displacement field is generated by normalized cross-correlation coefficient calculation and cubic polynomial subpixel interpolation. In this embodiment, when the surgeon removes a piece of about 1cm 3 After removing glioma tissue, the processing logic can calculate the displacement (0.8 mm) and direction of the surrounding normal brain tissue towards the collapsed resection cavity, thus achieving quantitative detection of collapsed brain drift. (3) S3-3 image correction: Using the generated three-dimensional displacement field as input, a three-dimensional control point grid (grid spacing 12 voxels) is constructed. Global coordinate transformation is achieved based on the third-order B-spline free deformation model. Geometric correction is performed on the initial three-dimensional image of brain tissue. Collapse displacement data is compensated to the imaging coordinate system in real time to generate a dynamic three-dimensional image that matches the real morphology of brain tissue during surgery, thus achieving brain drift dynamic correction throughout the operation. S4, Multimodal Fusion Navigation

[0063] like Figure 4 As shown, the data processing center 3, through the multimodal registration and fusion module 33, performs real-time feature point registration and fusion of the dynamic intraoperative three-dimensional ultrasound image generated in step S3 (corrected for collapsed brain drift) with the pre-loaded high-resolution preoperative MRI image of the patient. It extracts anatomical feature points such as white matter fiber bundles and deep nuclei from the MRI image and dynamic feature points such as tumor resection cavity and vascular displacement from the intraoperative ultrasound. The RANSAC algorithm is used to match feature points and complete non-rigid registration. In this embodiment, the fusion takes 300ms and the registration error is 0.3mm. This index is suitable for the technical requirements of the surgical scenario corresponding to this embodiment. In the fused image, the fine anatomical background provided by MRI and the real-time dynamic boundary corrected for brain drift provided by ultrasound can achieve a good superposition effect. Finally, the navigation image is sent to the medical high-definition monitor in the operating room and the doctor's head-mounted mixed reality device 4 to provide a reference for glioma resection surgery.

[0064] like Figure 5 As shown, 5a is the uncorrected drift image, indicating a deviation between the lesion boundary and the location of the brain functional area, resulting in corresponding deviations in the navigation information; 5b is the image corrected using the method of this embodiment, showing a closer fit between the lesion boundary and the actual location of the functional area, and the navigation information better meets the actual surgical needs. In this embodiment, this method achieves high-resolution, continuous dynamic imaging of brain tissue during glioma resection surgery, providing a dynamic correction effect for brain tissue collapse-type brain drift caused by tumor resection. Surgeons can more clearly observe the positional relationship between the tumor boundary and surrounding functional areas, providing a reference for the operation. Example 2: Intraoperative real-time brain tissue imaging method based on broadband ultrasound array (application in cerebral hemorrhage evacuation surgery, with rebound brain drift correction)

[0065] This embodiment applies the method of the present invention to hypertensive intracerebral hemorrhage evacuation surgery, achieving a dynamic correction effect against the rebound-type brain drift caused by cerebrospinal fluid loss. The specific procedure still follows... Figure 1 The difference between steps S1-S4 shown is that: Figure 2As shown in Figure a, probe 1 is made of polyimide flexible substrate 11 material, with a single probe size of 2cm × 2cm, and integrates 1000 piezoelectric micromechanical ultrasonic transducer array elements 12. Three probes are physically spliced ​​together to form a probe array. Figure 2 The imaging array shown in b is directly attached to the surface of the brain tissue in the area of ​​cerebral hemorrhage; as shown in b. Figure 4 As shown, communication module 2 is a WiFi 6 communication module, employing a packet transmission strategy, simultaneously transmitting a 3MHz low-frequency band to detect deep hematoma and an 8MHz high-frequency band to identify hematoma boundaries and capture brain tissue rebound displacement; as Figure 3 As shown, the data processing center 3 captures the rebound brain drift caused by cerebrospinal fluid loss through voxel-by-voxel correlation calculation. In this embodiment, the displacement detection accuracy can reach 0.09 mm, and geometric correction is completed based on B-spline free deformation. Finally, the intraoperative ultrasound image and the preoperative CT image are fused. In this embodiment, the fusion takes 400 ms and the registration error is 0.4 mm. This indicator can be adapted to the technical requirements of the surgical scenario corresponding to this embodiment, and provides navigation-related references for hematoma minimally invasive removal operation. Example 3: Intraoperative Real-Time Brain Tissue Imaging System Based on Broadband Ultrasound Array

[0066] This embodiment provides an intraoperative real-time brain tissue imaging system based on a broadband ultrasound array, used to implement the methods described in Embodiments 1 and 2. The overall system structure is as follows: Figure 4 As shown, the modules work together to provide complete hardware and software support for the dynamic correction of various types of brain drift, such as collapse, rebound, and traction types. Specifically, they include a patch-type flexible broadband ultrasound array probe 1, a communication module 2, a data processing center 3, and a display device 4. The structure and function of each module are described in detail below:

[0067] (1) Patch-type flexible broadband ultrasonic array probe 1: such as Figure 2 As shown in Figure a, the probe 1 is the core hardware structure of the system, employing a staggered frequency division design. Each probe integrates ≥1000 independent channel array elements 12, with a flexible substrate 11 having a thickness of 0.5 mm. The fixing structure consists of a combination of medical sterile adhesive tape 14 and auxiliary suture points. The high-density ultrasound array element 12 is divided into a first-band array element and a second-band array element. The first-band array element operates at a frequency of 2MHz to 5MHz, and the second-band array element operates at a frequency of 5MHz to 12MHz. This staggered arrangement allows for the simultaneous acquisition of deep brain tissue signals and superficial tissue detail imaging, providing high signal-to-noise ratio raw data for voxel-level displacement field calculations. Figure 2 As shown in c, the probe 1 can be attached to the cranial sutures within the surgical field or to the exposed surface of brain tissue. The flexible base 11 can adapt to the irregular shape of the surgical field surface, ensuring the stability of ultrasound signal acquisition.

[0068] (2) Communication module 2: such as Figure 4As shown, this module is the system's communication transmission unit, supporting 5G / WiFi 6 dual-mode switching. Wired communication uses an ultra-fine flexible cable with a diameter of 0.8mm, which reduces the possibility of probe displacement due to traction. Communication module 2 includes a data frame encapsulation module and an anti-electromagnetic interference link module. The data frame contains a frame header, probe ID, timestamp, array element number, RF echo data block, and checksum. The timestamp is used for time alignment of data collected by different probes and / or different array elements. The anti-electromagnetic interference link module uses a combination of differential transmission, shielding layer isolation, and forward error correction coding (FEC) to ensure stable transmission of intraoperative ultrasound echo signals and maintain the continuity of brain drift detection. Figure 6 As shown, the design details of the data frame structure and anti-interference module are clearly illustrated.

[0069] (3) Data Processing Center 3: such as Figure 4 As shown, this center is the core processing unit of the system, and it has a built-in brain drift correction algorithm module 32 (such as...). Figure 3 As shown, the system can perform a voxel-level 3D displacement field generation process: selecting an N×N×N correlation window (N is an odd number from 3 to 7), calculating the normalized cross-correlation coefficient (NCC), sub-pixel interpolation, and 3D Gaussian smoothing filtering to output a voxel-level 3D displacement field; it can also perform a non-rigid spatial correction process: constructing a uniform 3D control point mesh (grid spacing 8-16 voxels), solving for control point displacements, and implementing global coordinate transformation based on a third-order B-spline free deformation model to generate a corrected 3D image of brain tissue; it can also perform a registration and fusion process: extracting feature point sets, performing coarse registration using the RANSAC algorithm, and fine registration driven by the voxel-level displacement field to generate intraoperative navigation images. For multi-probe scenarios, the data processing center 3 stores the spatial calibration parameters of each probe, and can resample and fuse the volume data collected by each probe to a unified voxel coordinate system; at the same time, it performs periodic cross-correlation verification on the overlapping field of view of multiple probes, and automatically optimizes and updates the relative extrinsic parameter matrix between probes when the registration error exceeds a preset threshold to maintain the continuous effectiveness of the unified voxel coordinate system. In addition, the data processing center 3 is equipped with a GPU accelerator card, which can realize real-time storage and traceability of brain drift data throughout the surgery. In some implementations, it can achieve sub-millimeter displacement detection accuracy and low registration error, which meets the technical requirements of intraoperative dynamic correction.

[0070] (4) Display device 4: such as Figure 4 As shown, this device is the system's visualization output unit, which combines a medical 4K high-definition display with a head-mounted mixed reality device. It supports interactive operations such as navigation image magnification, three-dimensional rotation, and brain drift data overlay display. It can output the intraoperative navigation images generated by the data processing center 3 in real time, providing surgeons with clear and intuitive navigation references and adapting to the clinical observation needs of different neurosurgical procedures.

[0071] The system in this embodiment is easy to operate and can be integrated into the neurosurgical procedure. It is suitable for various craniocerebral lesion surgeries. Its hardware structure and software algorithm are deeply integrated, and the entire process from signal acquisition, transmission to processing and display provides technical support for brain drift correction, providing integrated support for real-time brain tissue imaging and surgical navigation during surgery.

[0072] The intraoperative real-time brain tissue imaging method and system based on broadband ultrasound array described in this invention addresses the technical problems of limited brain drift correction effects, insufficient algorithm disclosure, and lack of full-process dynamic correction capabilities in existing technologies. It forms an integrated solution for full-process brain drift correction from multiple levels, including hardware, algorithms, and transmission. At the same time, it improves traditional technical problems such as probe compatibility, transmission interference, and single imaging frequency, and realizes high-resolution, continuous dynamic real-time imaging of intraoperative brain tissue and dynamic correction of brain drift.

[0073] The method and system of this invention can be applied to various craniocerebral lesion surgeries in neurosurgery departments of hospitals at all levels, and can be adapted to existing neurosurgical procedures; its brain drift correction related effects can improve the accuracy of neurosurgical navigation, reduce the risk of damage to normal neurovascular structures during surgery, provide corresponding support for postoperative recovery of patients, and meet the needs of neurosurgical navigation technology to develop towards precision and minimally invasive procedures.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A real-time intraoperative brain tissue imaging system based on a broadband ultrasound array, characterized in that, include: A patch-type flexible broadband ultrasound array probe includes a flexible substrate and a high-density ultrasound array disposed on the flexible substrate. The edge of the flexible substrate is provided with a fixation structure for fixing the probe to the cranial sutures within the surgical field or to the surface of exposed brain tissue. The high-density ultrasound array integrates multiple independently transmitting and receiving ultrasound elements, preferably no less than 1000, to improve spatial sampling density and provide high signal-to-noise ratio raw data for voxel-level displacement field calculation. A communication module, connected to the probe, is used to send control commands to the probe and receive ultrasonic echo signals collected by the probe; A data processing center includes a processor and a memory, the memory storing a computer program, and the processor being configured to execute the computer program as follows: The ultrasonic echo signal is subjected to noise reduction processing; A voxel-level three-dimensional displacement field describing brain tissue deformation is generated based on the denoised ultrasound echo signal of consecutive frames. Non-rigid spatial correction is performed on the three-dimensional images of brain tissue based on the voxel-level three-dimensional displacement field; The corrected three-dimensional brain tissue image is registered and fused with the preoperative image to generate an intraoperative navigation image; and a display device is used to display the intraoperative navigation image.

2. The system according to claim 1, characterized in that, The fixation structure is one or more of the following: medical sterile adhesive tape, auxiliary suture point, or clamping fixation structure.

3. The system according to claim 1, characterized in that, The high-density ultrasonic array is a piezoelectric micromechanical ultrasonic transducer (pMUT) array or a capacitive micromechanical ultrasonic transducer (cMUT) array; the flexible substrate is made of silicone rubber or polyimide material with a thickness of 0.2 mm to 1.0 mm.

4. The system according to claim 1, characterized in that, The high-density ultrasonic array includes a first frequency band array element and a second frequency band array element. The first frequency band array element is used to transmit and receive low-frequency ultrasonic signals, and the second frequency band array element is used to transmit and receive high-frequency ultrasonic signals. The first frequency band array element and the second frequency band array element are disposed on the same flexible substrate and share the same probe coordinate system.

5. The system according to claim 4, characterized in that, The first frequency band array element operates in the frequency band of 2MHz to 5MHz, and the second frequency band array element operates in the frequency band of 5MHz to 12MHz; the first frequency band array element and the second frequency band array element are arranged in an alternating, partitioned, or sub-array excitation manner.

6. The system according to claim 4 or 5, characterized in that, The data processing center is further configured as follows: The overall displacement reference of deep brain tissue is obtained based on the low-frequency ultrasound signal; Information on local deformation of superficial brain tissue was obtained based on the high-frequency ultrasound signal. Furthermore, low-frequency and high-frequency ultrasonic signals are fused at multiple scales in a unified voxel coordinate system. Low-frequency displacement reference is used as a global constraint, and high-frequency deformation information is used as the basis for detail correction to generate a layered corrected voxel-level three-dimensional displacement field.

7. The system according to claim 1, characterized in that, The system includes multiple patch-type flexible broadband ultrasonic array probes, and the data processing center stores the spatial calibration parameters of each probe and is configured as follows: Based on the spatial calibration parameters of each probe, the three-dimensional volume data acquired by each probe are resampled to a unified voxel coordinate system; The resampled voxel data are then weighted and fused in overlapping regions to generate three-dimensional volumetric data of the complete surgical field. The denoising process, voxel-level displacement field generation, non-rigid spatial correction, and registration fusion are performed on the fused complete surgical field 3D volume data.

8. The system according to claim 7, characterized in that, The spatial calibration parameters are obtained in the following way: The intrinsic parameter matrix of each probe is calibrated in advance using a calibration phantom containing multiple known reflection points; The relative extrinsic parameter matrix between each probe is obtained by marking the probe edges or by cross-correlation matching of overlapping fields of view; By combining the intrinsic and extrinsic parameter matrices, the calibration parameter matrix of each probe to the unified voxel coordinate system is obtained.

9. The system according to claim 7 or 8, characterized in that, The data processing center is further configured as follows: Periodic cross-correlation verification was performed on ultrasound images from overlapping fields of view of multiple probes; When the registration error in the overlapping area exceeds the preset threshold, the local optimization update of the relative extrinsic matrix between the probes is automatically triggered to ensure the continuous effectiveness of the unified voxel coordinate system.

10. The system according to claim 1, characterized in that, The communication module supports wireless communication and / or wired communication. The wireless communication methods include one or more of 5G communication, WiFi 6 and above, Bluetooth 5.0 and above, and ultra-wideband (UWB) communication. The wired communication method uses ultra-fine flexible cable with a cable diameter ≤1.0mm.

11. The system according to claim 10, characterized in that, The communication module includes a data frame encapsulation module and an anti-electromagnetic interference link module. The data frame contains a frame header, probe ID, timestamp, array element number, RF echo data block, and checksum. The timestamp is used to time-align data collected by different probes and / or different array elements. The anti-electromagnetic interference link module is implemented using one or more of differential transmission, shielding layer isolation, and forward error correction coding (FEC) to ensure stable transmission of ultrasound echo signals during surgery.

12. The system according to claim 1, characterized in that, The data processing center is configured to generate the voxel-level three-dimensional displacement field as follows: Select an N×N×N related window (N is an odd number from 3 to 7) at the corresponding voxel position in consecutive frame images; Calculate the similarity metric of the echo signals within the window within the preset three-dimensional search range to obtain the initial displacement estimate; Subpixel interpolation is performed on the initial displacement estimate to obtain a subpixel-level displacement vector; The displacement field is smoothed by applying a three-dimensional Gaussian smoothing filter to suppress noise and ensure the spatial smoothness of the displacement field, and the output is a voxel-level three-dimensional displacement field.

13. The system according to claim 12, characterized in that, The similarity metric is the normalized cross-correlation coefficient (NCC); the sub-pixel interpolation is obtained by performing cubic polynomial interpolation on the similarity metric surface to improve the precision of displacement detection.

14. The system according to claim 1, characterized in that, The data processing center is configured to perform non-rigid spatial correction as follows: A uniform three-dimensional control point mesh is constructed based on the voxel-level three-dimensional displacement field, with a mesh spacing of 8 to 16 voxels. Solve for the displacement vector of each control point to optimize the consistency between the transformed image and the reference image, and ensure the continuity of the anatomical structure in the transformed image; Using a third-order B-spline free deformation (FFD) model, the displacement of control points is interpolated to each voxel point of the three-dimensional image of brain tissue to generate a corrected three-dimensional image that matches the actual deformation of brain tissue.

15. The system according to claim 1, characterized in that, The data processing center is configured to perform registration and fusion as follows: Extract the set of anatomical feature points from preoperative images, including vascular bifurcation points, sulcus intersections, anatomical landmarks, etc. Extract the dynamic feature point set from the corrected intraoperative ultrasound images, including lesion boundary corner points, vessel centerline points, and resection cavity contour points; Initial feature point matching pairs are obtained based on Euclidean distance. The RANSAC algorithm is used to remove outliers and solve the rigid registration transformation matrix to complete coarse registration. By combining the voxel-level three-dimensional displacement field with the preoperative images for non-rigid registration, optimizing local mutual information, and completing fine registration, intraoperative navigation images are generated.

16. The system according to any one of claims 1 to 15, characterized in that, The preoperative images include one or more of MRI images, CT images, and three-dimensional ultrasound images.

17. The system according to any one of claims 1 to 16, characterized in that, The display device includes a medical high-definition display, smart glasses, or a head-mounted mixed reality device, supporting interactive operations such as navigation image magnification, three-dimensional rotation, and brain drift data overlay display; the data processing center is equipped with a GPU accelerator card to support real-time storage and tracing of brain drift data throughout the surgery.

18. A method for real-time intraoperative brain tissue imaging based on a broadband ultrasound array, applied to the system described in any one of claims 1 to 17, characterized in that, include: S1, the patch-type flexible broadband ultrasound array probe is fixed to the cranial sutures or exposed brain tissue surface in the surgical field by a fixing structure to improve the stability and continuity of ultrasound signal acquisition during surgery. s2, send control commands to the probe through the communication module to drive the probe to collect ultrasonic echo signals of different frequency bands and transmit the ultrasonic echo signals to the data processing center; S3, the data processing center performs noise reduction processing on the ultrasound echo signal, generates a voxel-level three-dimensional displacement field based on the continuous frame ultrasound echo signal, and then performs non-rigid spatial correction on the three-dimensional image of brain tissue based on the voxel-level three-dimensional displacement field. S4, the data processing center registers and fuses the corrected three-dimensional brain tissue image with the preoperative image to generate intraoperative navigation images, which are then output and displayed through a display device.

19. The method according to claim 18, characterized in that, In step S2, the ultrasonic echo signals of different frequency bands include low-frequency ultrasonic echo signals of 2MHz to 5MHz and high-frequency ultrasonic echo signals of 5MHz to 12MHz, which are obtained through a time-division transmission strategy or a group transmission strategy. The time-division transmission strategy is to drive different array elements to transmit ultrasonic waves of corresponding frequency bands in different time periods within the same imaging cycle. The group transmission strategy is to divide the array elements into a low-frequency transmission group and a high-frequency transmission group to transmit ultrasonic waves of corresponding frequency bands simultaneously.

20. The method according to claim 18, characterized in that, Step S3, generating the voxel-level three-dimensional displacement field, includes: Select an N×N×N related window (N is an odd number from 3 to 7) at the corresponding voxel position in consecutive frame images; Calculate similarity metrics such as normalized cross-correlation coefficient (NCC) within the preset 3D search range; Voxel displacement is determined by cubic polynomial subpixel interpolation based on similarity metrics; After smoothing and constraint processing using three-dimensional Gaussian smoothing filtering, a voxel-level three-dimensional displacement field is obtained.

21. The method according to claim 18, characterized in that, The non-rigid spatial correction in step S3 includes: A uniform three-dimensional control point mesh is constructed based on the voxel-level three-dimensional displacement field; Solve for the displacement vector of each control point to optimize the consistency between the transformed image and the reference image; The brain tissue 3D image is resampled based on the third-order B-spline free deformation model to generate a corrected brain tissue 3D image.

22. The method according to claim 18, characterized in that, In step S4, the set of anatomical feature points in the preoperative image is matched with the set of dynamic feature points in the corrected intraoperative ultrasound image. The initial registration relationship is obtained through the RANSAC algorithm, and the non-rigid registration of the preoperative image and the intraoperative ultrasound image is completed by combining the voxel-level three-dimensional displacement field.

23. The method according to any one of claims 18 to 22, characterized in that, The preoperative images include one or more of MRI images, CT images, and three-dimensional ultrasound images; the method is suitable for craniocerebral tumor resection surgery, cerebral vascular malformation repair surgery, cerebral hemorrhage removal surgery, and deep brain lesion biopsy surgery. In some embodiments, it can achieve good dynamic correction effects for various brain drift types such as collapse type, rebound type, and traction type.

Images