A photoacoustic imaging guided surgical robot system and a control method thereof

Abstract

The present application belongs to the field of medical devices, and particularly relates to a photoacoustic imaging guided surgical robot system and a control method thereof. The photoacoustic imaging guided surgical robot system comprises an input module, a signal preprocessing module, a motion compensation processor and an AI reconstruction engine. The present application dynamically adjusts the array element delay by using a curvature adaptive focusing algorithm; the optical tracking and IMU data are used to compensate the photoacoustic signal, and the motion artifacts of the photoacoustic image are eliminated in cooperation with the image domain compensation. At the same time, the VPC and U-Net++ modules are used to generate a tumor boundary and clear blood vessels, and the photoacoustic image containing an electronic fence which can update the coordinates in real time according to the change of the intraoperative photoacoustic hemoglobin, so that the misoperation rate of the robot surgery is significantly reduced, and the present application has practical popularization and application value.

Description

Technical Field

[0001] This invention belongs to the field of medical devices, specifically relating to a photoacoustic imaging-guided surgical robot system and its control method. Background Technology

[0002] Surgical robotic systems are an advanced medical technology that originated in the late 20th century, initially funded by NASA and DARPA to provide telemedicine services to astronauts and soldiers. It combines the advantages of minimally invasive surgery and robotic control, aiming to improve the precision and safety of procedures. With advancements in science and technology, surgical robots have evolved from image-guided precision task robots to master-slave configured robots.

[0003] Existing surgical robot systems, such as the da Vinci system, largely rely on the registration of preoperative CT / MRI images with intraoperative optical or ultrasound images for navigation. However, this technology has significant limitations: First, preoperative images cannot reflect changes in anatomical structures caused by respiration, heartbeat, and tissue displacement during surgery in real time, resulting in a "time lag" problem; second, traditional ultrasound images have limited resolution for microvessels (especially tumor-feeding vessels) and insufficient contrast in distinguishing tumor boundaries from normal tissue, which can easily lead to accidental injury to blood vessels or incomplete tumor resection during surgery.

[0004] Photoacoustic imaging technology combines the high contrast of optical imaging with the high penetration depth of ultrasound imaging, enabling it to specifically identify hemoglobin and thus clearly display vascular networks. However, integrating photoacoustic imaging into surgical robots and achieving real-time, precise intraoperative guidance still faces challenges: (1) Patient physiological movements (such as breathing and heartbeat) can cause severe motion artifacts in photoacoustic images; (2) Rigid probes are difficult to fit the curved surfaces of complex organs, resulting in incomplete signal acquisition; (3) There is a lack of an intelligent decision-making mechanism that can convert imaging information into robot motion constraints in real time to achieve proactive safety protection.

[0005] Therefore, there is an urgent need for an intelligent system that can generate high-precision, artifact-free photoacoustic images in real time, and can automatically define surgical no-go zones and dynamically guide robot operations. Summary of the Invention

[0006] To address the above problems, the present invention provides a photoacoustic imaging-guided surgical robot system, comprising: The input module is configured to input an electrical signal converted from a photoacoustic signal generated by the target tissue in the surgical area, and the contact curvature radius between the photoacoustic probe used to acquire the photoacoustic signal and the target tissue. The signal preprocessing module is configured to perform time-delay alignment on the electrical signals converted from all photoacoustic signals; The motion compensation processor is configured to perform motion compensation on the time-delay aligned electrical signal to obtain a motion-corrected time-domain electrical signal. The AI ​​reconstruction engine is configured to: generate an initial motion-corrected image based on the motion-corrected temporal electrical signal; register the initial motion-corrected image with a preoperative MRI image, extract the vascular skeleton, add structural constraints to the intensity values ​​of image pixels, and obtain a reconstructed photoacoustic image; fuse the reconstructed photoacoustic image with the initial motion-corrected image to obtain a fused image; segment the tumor boundary and blood vessels in the fused image based on VPC; the real-time calculation formula for VPC is shown in Formula VII. Formula VII in, P(f) :yes p(t) Fourier transform, p(t) It is the hemoglobin pulsation signal extracted from photoacoustic signals; Of) Fourier transform of the tissue motion acceleration vector a(t) measured in real time by the IMU inertial unit; A (f) : Of) The complex conjugate; f min ,f max The frequency range corresponding to heart rate is 0.8Hz~2.5Hz.

[0007] Furthermore, the input module includes a laser emission module, a multi-wavelength fiber array, a flexible photoacoustic probe, a robot arm joint, and a six-dimensional force sensor; The laser emitting module is configured to emit a laser. A multi-wavelength fiber array is configured to transmit laser light to the surgical area of ​​the body to excite the target tissue to generate photoacoustic signals. A flexible photoacoustic probe is configured to conform to the curved surface of the surgical area tissue, and uses an integrated ultrasonic transducer array therein to collect photoacoustic signals generated by the tissue and convert them into electrical signals; The robot arm joint is configured to connect to a flexible photoacoustic probe for controlling and adjusting the position of the flexible photoacoustic probe; A six-dimensional force sensor is configured to be integrated inside the robot arm joint for real-time measurement of the radius of curvature of the contact between the probe and the tissue.

[0008] Furthermore, the laser emitted by the laser emitting module has a wavelength range of 750nm to 850nm; The ultrasonic transducer array is a spiral gradient density ultrasonic transducer array with polydimethylsiloxane as a flexible substrate. The spiral gradient density ultrasonic transducer array has an element density of 12 elements / mm² in the central region of the probe. 2 The element density in the edge region of the probe gradually changes to 4 elements / mm. 2 ; Furthermore, the signal preprocessing module calculates the delay time in real time using formula I; Formula I in, : Acoustic delay time of the i-th element in the ultrasonic transducer array; r: radius of curvature of the contact between the photoacoustic probe and the target tissue; d i θ is the distance from the i-th array element to the central reference point of the photoacoustic probe. i The azimuth angle of the i-th element relative to the surface normal of the target tissue surface; C The average speed of sound in the target tissue.

[0009] Furthermore, the motion compensation processor uses a wavelet domain compensation module to correct the electrical signal output by the signal preprocessing module using data provided by the optical tracker and the IMU inertial unit, thereby obtaining a motion-corrected time-domain electrical signal. The wavelet domain compensation module uses the electrical signal output from the Symlet-5 wavelet function decomposition signal preprocessing module to suppress noise and motion artifacts in the decomposed electrical signal by applying IMU weights. The formula for calculating the IMU weights is as follows: Formula II Among them, W IMF :IMU weights; λ: attenuation coefficient 0.5; a( t ): The acceleration vector of tissue motion measured in real time by the IMU inertial unit; e: natural constant 2.71828.

[0010] Furthermore, the optical tracker is used to collect the patient's pose data; the IMU inertial unit is used to measure tissue motion data.

[0011] Furthermore, the AI ​​reconstruction engine includes an image reconstruction unit, an image domain compensation module, a dual-domain fusion module, and an image segmentation module; The image reconstruction unit is configured to receive the motion-corrected time-domain electrical signal and generate an initial motion-corrected image using a time-delay superposition algorithm. The image domain compensation module is configured to use the Demons algorithm to register the initial motion-corrected image with the preoperative MRI image, extract the vascular skeleton, and use Formula III to add structural constraints to the intensity values ​​of the image pixels to obtain the reconstructed photoacoustic image. Formula III Among them, Ifinal( x,y ) : The pixel value of the final reconstructed photoacoustic image at coordinates (x, y); I PA (x,y): The pixel value of the initial motion-corrected image at coordinates (x,y); I MRI (x,y): Pixel value at coordinates (x,y) in the registered preoperative MRI image; S(x,y): Binarized mask of the vascular skeleton extracted from the preoperative MRI, with 1 for vascular areas and 0 for non-vascular areas; α: Fusion intensity coefficient, ranging from 0.2 to 0.5, used to control the degree of introduction of prior MRI information; The dual-domain fusion module is configured to use an adaptive weighted fusion algorithm based on signal-to-noise ratio to fuse the reconstructed photoacoustic image output by the image domain compensation module with the initial motion-corrected image output by the image reconstruction unit. The image segmentation module is configured to use the fused image output by the VPC and dual-domain fusion module as the input channel of the U-Net++ model, segment the tumor boundary and blood vessels in the fused image, and mark the risk areas of the segmentation results based on the real-time dynamics of VPC: when 0.15 ≤ VPC < 0.3, it is marked as a high-risk area of ​​the electronic fence; when VPC < 0.15, it is marked as a no-go area of ​​the electronic fence.

[0012] Furthermore, the steps of the adaptive weighted fusion algorithm based on signal-to-noise ratio are as follows: ① Based on the signal-to-noise ratio of each region in the initial motion-corrected image and the reconstructed photoacoustic image, weights are dynamically allocated using formula IV; Formula IV in, W As weight, SNR domain The signal-to-noise ratio of the two image domain data sources involved in the fusion; SNR domain1 The signal-to-noise ratio of the initial motion-corrected image is calculated within an N×N local window centered on the target pixel. SNR domain2 The signal-to-noise ratio is calculated for the corresponding region in the reconstructed photoacoustic image; N can be 7 or 9. ②The images after weighting in step ① are fused using formula VI. Formula VI in, W 1(x,y) and W 2(x,y) is the weight calculated based on formula IV for the position (x,y), and satisfies W 1+ W 2=1 to ensure the stability of the intensity range of the fused image; I motion_corrected(x,y) are the pixel values ​​of the initial motion-corrected image; I structure_constrained (x,y) are the pixel values ​​of the reconstructed photoacoustic image; I fused (x,y) are the pixel values ​​of the final fused image; The signal-to-noise ratio is calculated according to formula V: Formula V, Where SNR is the signal-to-noise ratio; SignalPower is taken from the square of the mean of the pixels within the window, and NoisePower is taken from the square of the standard deviation of the pixels within the window.

[0013] Furthermore, the surgical robot system also includes a surgical navigation controller and a robot motion actuator; The surgical navigation controller is configured to convert the electronic fence into robot joint torque limiting commands, forcing the robot speed to ≤1mm / s in high-risk areas of the electronic fence and 0mm / s in no-go areas of the electronic fence.

[0014] The present invention also provides a control method for the aforementioned surgical robot system, comprising: S1: A laser is emitted through a laser emission module; a multi-wavelength fiber array excites photoacoustic signals in the target tissue of the surgical area; a flexible photoacoustic probe converts the photoacoustic signals generated by the tissue into electrical signals using an integrated ultrasonic transducer array; S2: The contact curvature radius between the flexible photoacoustic probe and the tissue is measured in real time using a six-dimensional force sensor, and then the electrical signal converted by the ultrasonic transducer array is delayed and aligned using a signal preprocessing module. S3: The electrical signal output by the signal preprocessing module is corrected by the motion compensation processor; S4: Using an AI reconstruction engine, tumor boundaries and blood vessels are segmented based on VPC and U-Net++ models to generate an electronic fence with real-time updated coordinates. S5: Update the VPC and iterate through S4.

[0015] Furthermore, the VPC described in step S5 is updated every 100ms.

[0016] This invention discloses a photoacoustic imaging-guided surgical robot system and its control method. The system dynamically adjusts array element delay using a curvature adaptive focusing algorithm in the signal preprocessing module. In the motion compensation processor, motion compensation of the photoacoustic signal is performed using optical tracking and IMU data, and motion artifacts in the photoacoustic image are eliminated with the assistance of image domain compensation. Simultaneously, based on the vascular pulsatility coherence coefficient (VPC) and the U-Net++ module, photoacoustic images of clearly defined tumor boundaries and blood vessels are generated, containing an electronic fence for robot movement restricted areas whose coordinates are updated in real-time according to intraoperative photoacoustic hemoglobin changes. This significantly reduces the error rate in robotic surgery and has practical application value.

[0017] Obviously, based on the above description of the present invention, and according to common technical knowledge and conventional methods in the field, various other modifications, substitutions or alterations can be made without departing from the basic technical concept of the present invention.

[0018] The following detailed embodiments further illustrate the above-described content of the present invention. However, this should not be construed as limiting the scope of the present invention to the following examples. All technologies implemented based on the above-described content of the present invention fall within the scope of the present invention. Attached Figure Description

[0019] Figure 1 Structural diagram of a photoacoustic imaging-guided surgical robot system; Figure 2 A flowchart for two-domain motion artifact removal. Detailed Implementation

[0020] Example 1: Photoacoustic Imaging Guided Surgical Robot System of the Present Invention A: Laser emitting module: used to emit lasers with a wavelength range of 750nm ~ 850nm; B: Multi-wavelength fiber array: used to transmit time-division multi-wavelength pulsed laser to the surgical area of ​​the body to irradiate the target tissue, thereby exciting photoacoustic signals that can be used for imaging; C: Flexible photoacoustic probe: It integrates a spiral gradient density ultrasonic transducer array (F) with polydimethylsiloxane (PDMS) as a flexible substrate; the array has an element density of 12 elements / mm² in the central region of the probe and a gradually decreasing element density of 4 elements / mm² in the edge region of the probe, which is used to conform to the curved surface of the surgical area, collect the photoacoustic signals generated by the tissue, and convert them into electrical signals; D: Robot arm joint: Connected to the flexible photoacoustic probe, used to control and adjust the position of the flexible photoacoustic probe to achieve adaptive fitting and omnidirectional scanning of its sensing surface to the tissue surface, thereby collecting photoacoustic signals generated by the tissue without blind spots; E: Six-dimensional force sensor: integrated inside the robot arm joint, used to measure the contact curvature radius r between the probe and the tissue in real time; G: Signal preprocessing module: Dynamically calculates the delay time of each element in the ultrasonic transducer array using formula I. T i Furthermore, all electrical signals are delayed and aligned to eliminate the phase difference caused by curved surface detection, thereby improving the clarity of image reconstruction. Formula I in, T i : Acoustic delay time of the i-th element; r: Radius of curvature of the probe-tissue contact point measured in real time by the six-dimensional force sensor (unit: mm); d i θ: Distance from the i-th element to the center reference point of the probe (unit: mm); i : The azimuth angle of the i-th element relative to the surface normal (unit: degrees); C The average speed of sound in tissue (1540 m / s) H: Motion Compensation Processor: Utilizes motion data collected by an optical tracker and an IMU inertial unit, and corrects the preprocessed electrical signal through a wavelet domain compensation module to obtain a motion-corrected time-domain electrical signal; Optical tracker: used to collect patient pose data; IMU (Inertial Measurement Unit): Used to measure tissue motion data; Wavelet domain compensation module: The preprocessed electrical signal is decomposed using the Symlet-5 wavelet function. IMU weights are applied to suppress noise and motion artifacts in the decomposed electrical signal. The formula for calculating the IMU weights is as follows: Formula II Among them, W IMF : IMU weights; λ: attenuation coefficient, empirical value 0.5, used to adjust the influence of motion acceleration on the weights; a( t Tissue motion acceleration vector measured in real time by the IMU inertial unit (unit: m / s²) 2 e: natural constant, with a value of 2.71828; I:AI Reconstruction Engine: It has a built-in image reconstruction unit, image domain compensation module and dual domain fusion module to generate photoacoustic images that eliminate motion artifacts; and a built-in image segmentation module to obtain three-dimensional photoacoustic images with clear tumor boundaries and blood vessels. The three-dimensional photoacoustic images also contain electronic fences for robot movement restricted areas that can be updated in real time according to changes in photoacoustic hemoglobin during surgery. Image reconstruction unit: Receives the motion-corrected time-domain electrical signal and generates an initial motion-corrected image using a time-delay superposition algorithm; Image domain compensation module: The Demons algorithm is used to register the initial motion-corrected image with the preoperative MRI image, extract the vascular skeleton, and use Formula III to add structural constraints to the intensity values ​​of image pixels to obtain the reconstructed photoacoustic image; Formula III Among them, I final( x,y ) : The pixel value of the final reconstructed photoacoustic image at coordinates (x, y); I PA (x,y): The pixel value of the initial motion-corrected image at coordinates (x,y); I MRI (x,y): Pixel value at coordinates (x,y) in the registered preoperative MRI image; S(x,y): Binarized mask of the vascular skeleton extracted from the preoperative MRI, 1 for vascular areas and 0 for non-vascular areas; a: Fusion intensity coefficient, ranging from 0.2 to 0.5, used to control the degree of introduction of prior MRI information; Dual-domain fusion module: It adopts an adaptive weighted fusion algorithm based on signal-to-noise ratio to fuse the reconstructed photoacoustic image output by the image domain compensation module with the initial motion correction image output by the image reconstruction unit; The steps of the adaptive weighted fusion algorithm based on signal-to-noise ratio are as follows: ① Based on the signal-to-noise ratio (SNR) of each region in the initial motion-corrected image and the reconstructed photoacoustic image, weights are dynamically allocated using formula IV; Formula IV in, W As weight, SNR domain The signal-to-noise ratio of the two image domain data sources involved in the fusion; SNR domain1 The signal-to-noise ratio of the initial motion-corrected image is calculated within an N×N local window centered on the target pixel. SNR domain2 To reconstruct the photoacoustic image, the signal-to-noise ratio (SNR) is calculated for the corresponding region; N is either 7 or 9, and this window size strikes a good balance between statistical robustness and spatial detail preservation; the SNR is calculated using formula V: Formula V SignalPower is taken from the square of the mean of the pixels within the window, and NoisePower is taken from the square of the standard deviation of the pixels within the window.

[0021] ② For the image after weighting in step ①, use formula VI to fuse it. Formula VI in, W 1(x,y) and W 2(x,y) is the weight calculated based on formula IV for the position (x,y), and satisfies W 1+ W 2=1 to ensure the stability of the intensity range of the fused image; I motion_corrected (x,y) are the pixel values ​​of the initial motion-corrected image; I structure_constrained (x,y) are the pixel values ​​of the reconstructed photoacoustic image; I fused (x,y) are the pixel values ​​of the final merged image.

[0022] Image segmentation module: The vascular pulsatility coherence coefficient (VPC) is calculated in real time using formula VII. The fused image of VPC and the output of the dual-domain fusion module is used as the input channel of the U-Net++ model. The tumor boundary and blood vessels are segmented in the fused image. At the same time, based on the real-time dynamics of VPC, the segmentation results are marked with risk areas: when 0.15 ≤ VPC < 0.3, it is marked as a high-risk area of ​​the electronic fence; when VPC < 0.15, it is marked as a no-risk area of ​​the electronic fence.

[0023] Formula VII in, P(f) :yes p(t) Fourier transform, p(t) It is a hemoglobin pulsation signal extracted from the photoacoustic signal excited by multi-wavelength lasers, reflecting the change of hemoglobin concentration over time; Of) Fourier transform of the tissue motion acceleration vector a(t) measured in real time by the IMU inertial unit; A (f) : Of) The complex conjugate; f min ,f max The corresponding heart rate frequency range is 0.8Hz~2.5Hz, which is 48 beats / minute to 150 beats / minute. J: Surgical Navigation Controller: Converts the electronic fence into robot joint torque limiting commands. In high-risk areas of the electronic fence, the robot speed is forced to be ≤1mm / s; in no-risk areas of the electronic fence, the robot speed is 0mm / s. K: Robotic motion actuator: Receives instructions from the surgical navigation controller.

[0024] The following experimental examples illustrate the beneficial effects of the present invention.

[0025] Experiment 1: Performance Verification of the Core Algorithm 1. Experimental Objective Verify the effect of the curvature adaptive focusing algorithm (Formula I) in the signal preprocessing module of this invention on improving the resolution of curved surface imaging.

[0026] 2. Experimental Methods Experimental platform: Photoacoustic imaging experimental platform (not connected to the robot actuator) Experimental sample: Custom photoacoustic phantom (containing a tungsten wire target with a diameter of 50μm, and adjustable surface curvature with radii of curvature of ∞, 50mm, and 20mm respectively) Control group design: Control group: Traditional uniform array ultrasonic transducer + conventional delay superposition algorithm Experimental group: Gradient density array of this invention + curvature adaptive focusing algorithm (Formula I) Evaluation metric: Imaging resolution, obtained by measuring the point spread function (PSF) half-width (FWHM) of the target image.

[0027] 3. Experimental Results Table 1 Comparison of imaging resolution under different curvature conditions 4. Conclusion Experimental results show that on flat surfaces, the proposed solution has optimized the resolution by 12.5%. As the radius of curvature decreases (the curvature of the surface increases), the resolution of traditional uniform arrays deteriorates sharply (dropping to 150 μm at a radius of 20 mm). However, the proposed solution, through the synergistic effect of a gradient density array and a curvature adaptive focusing algorithm, successfully maintains the resolution at a high level of 50 μm, representing a 66.7% improvement over traditional solutions. This demonstrates that the proposed solution can effectively overcome the problem of image quality degradation when a flexible probe is fitted to a curved surface.

[0028] Experimental Example 2: Evaluation of Overall System Imaging Performance 1. Experimental Objective The overall resolution of the photoacoustic images generated by the surgical robot system of this invention was evaluated and compared with commonly used preoperative CT images in clinical practice.

[0029] 2. Experimental Methods Experimental sample: Multipurpose imaging phantom (Model 040GSE, CIRS), which is compatible with both ultrasound / photoacoustic imaging and CT imaging, and contains a grid target group with spatial frequencies from 5 lp / mm to 20 lp / mm.

[0030] Imaging scheme: The system of this invention performs photoacoustic imaging on a phantom body according to the method described in Example 1 to acquire a photoacoustic image. Traditional CT navigation system: Performs clinical-grade CT scans (0.625mm slice thickness, 0.3mm reconstruction interval) on the same phantom to acquire CT images. Evaluation metric: Modulation transfer function (MTF), which measures the contrast transfer capability of two images at different spatial frequencies, and records the spatial frequency (i.e., the limit resolution) at which the MTF drops to 50%. 3. Experimental Results The system of this invention: the MTF of the photoacoustic image decreases to 50% when the spatial frequency reaches 15 lp / mm. Traditional CT navigation systems: The MTF of CT images drops to 50% when the spatial frequency reaches 8 lp / mm. Comparison results: 15 lp / mm > 8 lp / mm. The intraoperative real-time photoacoustic images generated by the system of this invention have a significantly better limiting spatial resolution than traditional preoperative CT images. 3. Conclusion The higher spatial frequency means that the system of the present invention can distinguish finer anatomical structures (such as tumor-feeding vessels with a diameter of less than 100 μm), providing a better imaging basis for high-precision intraoperative navigation.

[0031] Experiment 3: Performance Validation of Autonomous Decision-Making Surgery 1. Method Twelve fresh bovine renal artery samples were taken and randomly divided into two groups of six each.

[0032] Experimental group (surgical robot system of Example 1, Figure 1 Follow these steps S1-S5: S1: Excite tissues with multi-wavelength lasers and collect photoacoustic signals; S2: Motion compensation is achieved by fusing optical tracking and IMU data; S3: Segmentation of tumor boundaries and blood vessels based on the U-Net++ model; S4: Dynamically generate electronic fences and limit the robot's range of motion; S5: Update the photoacoustic hemoglobin spectrum every 100ms and iterate through S4.

[0033] Control group (traditional surgical robot system): The da Vinci surgical robot system, based on preoperative static CT image navigation, lacks real-time photoacoustic imaging and electronic fence decision-making capabilities. Its operation procedure is supplemented as follows: S1': Preoperative CT scan of bovine renal artery samples to obtain a three-dimensional vascular model.

[0034] S2': Import the 3D model into the da Vinci system for surgical path planning.

[0035] S3': The robotic arm performs cutting operations according to a preset path, lacking real-time intraoperative image feedback and dynamic safety boundary protection.

[0036] 2. Results 2.1 Accuracy of blood vessel segmentation in three-dimensional photoacoustic images After completing step S3 (segmenting tumor boundaries and blood vessels based on the U-Net++ model), the segmentation accuracy of blood vessels in the 3D images generated by the two surgical robot systems was evaluated. The mean intersection-over-union ratio (mIoU) was used as the evaluation index, and the pathological slide results were used as the gold standard.

[0037] The results show: The mIoU for vessel segmentation in traditional surgical robot systems (based on preoperative CT image navigation, without real-time photoacoustic feedback) is 0.82. The surgical robot system of this invention (based on a U-Net++ model guided by VPC coefficients) achieves a vessel segmentation mIoU of 0.91. The increase in mIoU from 0.82 to 0.91 signifies a reduction in segmentation error rate of over 50%. This indicates that the VPC parameter-guided segmentation method employed in this invention can more accurately distinguish vascular structures from background tissue, resulting in a more reliable "electronic fence" that can more effectively prevent the robot from accidentally damaging critical blood vessels.

[0038] 2.2 Blood vessel miscut rate After completing steps S4 (dynamically generating an electronic fence and limiting the robot's range of motion) and S5 (the robot performs the cutting), an anatomical examination is performed on the cut bovine renal artery samples to count the number of vessels that were mistakenly cut (i.e., non-target vessels were cut or damaged).

[0039] The results show: Traditional surgical robot systems have a blood vessel mis-cutting rate of 83%. The vascular mis-cut rate of the surgical robot system of this invention is 3%. This invention reduces the rate of accidental vascular rupture from 83% to 3%, a reduction of 96%. This fully demonstrates that using real-time updated electronic fences for robot motion constraints can effectively avoid accidental damage to critical blood vessels during surgery and significantly improve surgical safety.

[0040] Experimental Case 4: Verification via Liver Tumor Resection Surgery 1. Method Eighteen liver tissue samples from the VX2 tumor model of New Zealand white rabbits (a commonly used animal model for liver tumor research) were taken and divided into two groups of nine samples each.

[0041] Experimental group (surgical robot system of Example 1): Liver tumor was cut according to the following procedure: 1) Initialization: A six-dimensional force sensor was used to detect the radius of curvature of the liver surface, r=25mm; adaptive curvature focusing was activated, α=0.023.

[0042] 2) Artifact Removal: Dual-domain motion artifact elimination (DDMC) workflow simultaneously processes respiratory motion artifacts (IMU weighted attenuation) and MRI structural constraint reconstruction. Figure 2 ).

[0043] 3) Vascular decision-making: The portal vein VPC was calculated to be 0.25 (<0.3), the electronic fence was marked as a high-risk area, and the robot speed was reduced to 1 mm / s.

[0044] 4) Real-time closed-loop surgical execution: The system continuously updates the photoacoustic hemoglobin map at 100ms intervals and iteratively executes steps 1) to 3) (i.e., curvature adaptive focusing, dual-domain motion artifact elimination, VPC-based vascular decision-making, and electronic fence updating). Guided by the real-time updated electronic fence, the robotic motion actuator performs tumor resection. After resection, pathological section analysis is performed on the excised tumor tissue to verify the surgical margins (i.e., determine whether cancer cells are present at the margins) and assess the volume of damage to surrounding normal liver tissue.

[0045] Control group (traditional surgical robot system): The operation procedure is supplemented as follows: 1') Initialization: Set static surgical boundaries based on preoperative MRI images.

[0046] 2') Navigation execution: The robotic arm performs tumor resection within the pre-planned static boundary and cannot perceive or adapt to tissue deformation and displacement caused by physiological activities such as breathing during the operation.

[0047] 3') Postoperative verification: After the excision is completed, the excised tissue is subjected to pathological analysis to verify the resection margin.

[0048] Results determination: The following results were primarily obtained through postoperative pathological section analysis of the resected tumor tissue, rather than directly testing the three-dimensional photoacoustic images obtained in step S2. The core evaluation indicators were the tumor margin positivity rate (i.e., the presence of cancer cells at the margin observed under a microscope) and the volume of damage to normal liver tissue.

[0049] 2. Results Microscopic examination was performed to observe the presence of cancer cells at the surgical margins and the volume of damage to surrounding normal liver tissue. Pathological analysis showed that in the experimental group using the surgical robot system of this invention, the positive rate of tumor margins was 5%, and the average volume of damage to surrounding normal liver tissue was 25 mm³. In contrast, in the control group using the traditional surgical robot system, the positive rate of tumor margins was 30%, and the average volume of damage to surrounding normal liver tissue was 80 mm³.

[0050] The results above demonstrate that the system of this invention, with its real-time updated photoacoustic hemoglobin mapping and dynamic electronic fence technology, can continuously track changes in tumor boundaries during surgery. This allows the robot to completely remove the tumor (with a low margin positivity rate) while maximally protecting healthy liver tissue (with minimal damage volume). In contrast, traditional systems, based on static image navigation, cannot cope with dynamic changes during surgery, easily leading to incomplete margins or over-resection, proving the significant superiority of this invention in complex and dynamic surgical environments.

Claims

1. A photoacoustic imaging-guided surgical robot system, characterized in that: include: The input module is configured to input an electrical signal converted from a photoacoustic signal generated by the target tissue in the surgical area, and the contact curvature radius between the photoacoustic probe used to acquire the photoacoustic signal and the target tissue. The signal preprocessing module is configured to perform time-delay alignment on the electrical signals converted from all photoacoustic signals; The motion compensation processor is configured to perform motion compensation on the time-delay aligned electrical signal to obtain a motion-corrected time-domain electrical signal. The AI ​​reconstruction engine is configured to: generate an initial motion-corrected image based on the motion-corrected temporal electrical signal; register the initial motion-corrected image with a preoperative MRI image, extract the vascular skeleton, add structural constraints to the intensity values ​​of image pixels, and obtain a reconstructed photoacoustic image; fuse the reconstructed photoacoustic image with the initial motion-corrected image to obtain a fused image; segment the tumor boundary and blood vessels in the fused image based on the vascular pulsatility coherence coefficient (VPC); the real-time calculation formula for the VPC is shown in Formula VII. Formula VII in, P(f) :yes p(t) Fourier transform, p(t) It is the hemoglobin pulsation signal extracted from photoacoustic signals; A (f) Fourier transform of the tissue motion acceleration vector a(t) measured in real time by the IMU inertial unit; A (f) : A(f) The complex conjugate; f min ,f max Heart rate corresponds to a frequency range of 0.8Hz to 2.5Hz; The AI ​​reconstruction engine includes an image reconstruction unit, an image domain compensation module, a dual-domain fusion module, and an image segmentation module; The image reconstruction unit is configured to receive the motion-corrected time-domain electrical signal and generate an initial motion-corrected image using a time-delay superposition algorithm. The image domain compensation module is configured to use the Demons algorithm to register the initial motion-corrected image with the preoperative MRI image, extract the vascular skeleton, and use Formula III to add structural constraints to the intensity values ​​of the image pixels to obtain the reconstructed photoacoustic image. Official III Among them, I final( x,y ) : The pixel value of the final reconstructed photoacoustic image at coordinates (x, y); I PA (x,y): The pixel value of the initial motion-corrected image at coordinates (x,y); I MRI (x,y): Pixel value at coordinates (x,y) in the registered preoperative MRI image; S(x,y): Binarized mask of the vascular skeleton extracted from the preoperative MRI, with 1 for vascular areas and 0 for non-vascular areas; α: Fusion intensity coefficient, ranging from 0.2 to 0.5, used to control the degree of introduction of prior MRI information; The dual-domain fusion module is configured to use an adaptive weighted fusion algorithm based on signal-to-noise ratio to fuse the reconstructed photoacoustic image output by the image domain compensation module with the initial motion-corrected image output by the image reconstruction unit. The image segmentation module is configured to use the fused image of the vascular pulsatility coherence coefficient (VPC) and the output of the dual-domain fusion module as the input channel of the U-Net++ model. It segments the tumor boundary and blood vessels in the fused image and marks the segmentation results as risk areas based on the real-time dynamics of the VPC: when 0.15 ≤ VPC < 0.3, it is marked as a high-risk area of ​​the electronic fence; when VPC < 0.15, it is marked as a no-go area of ​​the electronic fence.

2. The surgical robot system according to claim 1, characterized in that: The input module includes a laser emission module, a multi-wavelength fiber array, a flexible photoacoustic probe, a robot arm joint, and a six-dimensional force sensor; The laser emitting module is configured to emit a laser. A multi-wavelength fiber array is configured to transmit laser light to the surgical area of ​​the body to excite the target tissue to generate photoacoustic signals. A flexible photoacoustic probe is configured to conform to the curved surface of the surgical area tissue, and uses an integrated ultrasonic transducer array therein to collect photoacoustic signals generated by the tissue and convert them into electrical signals; The robot arm joint is configured to connect to a flexible photoacoustic probe for controlling and adjusting the position of the flexible photoacoustic probe; A six-dimensional force sensor is configured to be integrated inside the robot arm joint for real-time measurement of the radius of curvature of the contact between the probe and the tissue.

3. The surgical robot system according to claim 2, characterized in that: The laser emitted by the laser emitting module has a wavelength range of 750nm to 850nm; The ultrasonic transducer array is a spiral gradient density ultrasonic transducer array with polydimethylsiloxane as a flexible substrate. The spiral gradient density ultrasonic transducer array has an element density of 12 elements / mm² in the central region of the probe. 2 The element density in the edge region of the probe gradually changes to 4 elements / mm. 2 .

4. The surgical robot system according to claim 3, characterized in that: The signal preprocessing module calculates the delay time in real time using formula I; Formula I in, : Acoustic delay time of the i-th element in the ultrasonic transducer array; r: radius of curvature of the contact between the photoacoustic probe and the target tissue; d i θ is the distance from the i-th array element to the central reference point of the photoacoustic probe. i The azimuth angle of the i-th element relative to the surface normal of the target tissue surface; C The average speed of sound in the target tissue.

5. The surgical robot system according to claim 1, characterized in that: The motion compensation processor uses a wavelet domain compensation module to correct the electrical signal output by the signal preprocessing module using data provided by the optical tracker and the IMU inertial unit, thereby obtaining a motion-corrected time-domain electrical signal. The wavelet domain compensation module uses the electrical signal output from the Symlet-5 wavelet function decomposition signal preprocessing module to suppress noise and motion artifacts in the decomposed electrical signal by applying IMU weights. The formula for calculating the IMU weights is as follows: Official II Among them, W IMF :IMU weights; λ: attenuation coefficient 0.5; a( t ): The acceleration vector of tissue motion measured in real time by the IMU inertial unit; e: natural constant 2.71828.

6. The surgical robot system according to claim 5, characterized in that: The optical tracker is used to collect the patient's pose data; the IMU inertial unit is used to measure tissue motion data.

7. The surgical robot system according to claim 1, characterized in that: The steps of the adaptive weighted fusion algorithm based on signal-to-noise ratio are as follows: Step ①: Based on the signal-to-noise ratio of each region in the initial motion-corrected image and the reconstructed photoacoustic image, weights are dynamically allocated using formula IV; Formula IV in, W As weight, SNR domain The signal-to-noise ratio of the two image domain data sources involved in the fusion; SNR domain1 The signal-to-noise ratio of the initial motion-corrected image is calculated within an N×N local window centered on the target pixel. SNR domain2 The signal-to-noise ratio is calculated for the corresponding region in the reconstructed photoacoustic image; N can be 7 or 9. Step ②: The images after weighting in Step ① are fused using Formula VI. Formula VI in, W 1(x,y) and W 2(x,y) is the weight calculated based on formula IV for the position (x,y), and satisfies W 1+ W 2=1 to ensure the stability of the intensity range of the fused image; I motion_corrected (x,y) are the pixel values ​​of the initial motion-corrected image; I structure_constrained (x,y) are the pixel values ​​of the reconstructed photoacoustic image; I fused (x,y) are the pixel values ​​of the final fused image; The signal-to-noise ratio is calculated according to formula V: Formula V, Where SNR is the signal-to-noise ratio; SignalPower is taken from the square of the mean of the pixels within the window, and NoisePower is taken from the square of the standard deviation of the pixels within the window.

8. The surgical robot system according to claim 1, characterized in that: The surgical robot system also includes a surgical navigation controller and a robot motion actuator; The surgical navigation controller is configured to convert the electronic fence into robot joint torque limiting commands, forcing the robot speed to ≤1mm / s in high-risk areas of the electronic fence and 0mm / s in no-go areas of the electronic fence.

Images