AI image navigation method, device and equipment for minimally invasive surgery of prostatic hyperplasia and medium of AI image navigation method and device

Abstract

The invention relates to an AI image navigation method, device and equipment for prostatic hyperplasia minimally invasive surgery and a medium thereof. The method comprises the following steps: acquiring a prostate three-dimensional image and an intraoperative real-time image sequence, and fusing the three-dimensional image after preliminary filtering to obtain an optimized and enhanced image; based on the enhanced image and the preoperative three-dimensional image, a three-dimensional offset vector is obtained through a multi-modal registration algorithm combined with an anatomical structure reference, and optimized prostate three-dimensional attitude data is generated in combination with anatomical constraints; combining the attitude data and the three-dimensional image, determining boundary area distribution characteristics, calculating a risk assessment value, mapping a risk level, and generating a path correction instruction to obtain an instrument navigation path if the risk level exceeds a threshold value; and based on the navigation path, constructing a clinical recovery model in combination with the anatomical structure, generating postoperative recovery index simulation data, and performing adjustment and verification until preset requirements are met to obtain a final operation execution scheme. According to the method, precision and individuation of surgical navigation are achieved, and comprehensive technical support is provided for precise implementation of minimally invasive surgery.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of artificial intelligence, and particularly relates to an AI image navigation method, device and equipment for a benign prostatic hyperplasia minimally invasive surgery and a medium thereof. BACKGROUND

[0002] Benign prostatic hyperplasia (BPH) is a common urological disease in middle-aged and elderly men. Minimally invasive surgeries such as transurethral resection of the prostate and laser ablation have become the mainstream treatment method due to their small trauma and fast recovery. However, such surgeries have very high requirements for operation precision, and the surgical path planning, instrument posture control and protection of key adjacent structures such as the urethra and rectum directly affect the surgical effect and the postoperative quality of life of patients. In the prior art, surgical navigation relies on the experience of doctors and the subjective judgment of real-time images during surgery, and there are problems such as limited precision, high risk, lack of personalized planning and lagging behind in postoperative recovery evaluation, which makes it difficult to meet the needs of precise and intelligent surgery. SUMMARY

[0003] Therefore, it is necessary to provide an AI image navigation method, device and equipment for a benign prostatic hyperplasia minimally invasive surgery and a medium thereof, which can improve the precision of prostate posture positioning, realize quantitative judgment of surgical risk, and improve the effectiveness of surgery and the quality of postoperative recovery.

[0004] In a first aspect, the application provides an AI image navigation method for a benign prostatic hyperplasia minimally invasive surgery, comprising:

[0005] obtaining a three-dimensional image of the prostate and a real-time image sequence during surgery, and processing the real-time image sequence after preliminary filtering in combination with the three-dimensional image of the prostate to obtain an optimized enhanced image.

[0006] Based on the enhanced image and the preoperative three-dimensional image of the prostate, a three-dimensional offset vector is obtained by combining a multi-modal image registration of a three-dimensional anatomical structure reference, and a correction parameter is constructed in combination with a prostate anatomical structure constraint to generate optimized three-dimensional posture data of the prostate.

[0007] After determining the boundary region distribution characteristics in combination with the three-dimensional posture data of the prostate and the three-dimensional image of the prostate, a risk assessment value is calculated and a corresponding risk level is mapped, and if the risk level exceeds a preset threshold, a path correction instruction is generated to obtain a surgical instrument navigation path.

[0008] Based on the surgical instrument navigation path, a clinical recovery model is constructed in combination with the prostate anatomical structure to generate postoperative recovery index simulation data, and the recovery index simulation data is adjusted and verified to meet the preset recovery index requirements to obtain a final surgical execution scheme.

[0009] In one embodiment, a three-dimensional prostate image and a real-time image sequence during surgery are acquired, and after preliminary filtering of the real-time image sequence, the three-dimensional prostate image is combined for processing to obtain an optimized enhanced image, including:

[0010] A three-dimensional prostate image and a real-time image sequence during surgery are acquired, and the real-time image sequence is preliminarily filtered to obtain a filtered image sequence; the preliminary filtering includes pixel grayscale threshold screening, inter-frame redundant frame elimination, and Gaussian low-pass filtering.

[0011] Dynamic change features are extracted from the real-time image sequence using a dynamic change detection algorithm based on anatomical structure references in the three-dimensional prostate image to determine a first image of the change region; the dynamic change features include displacement vectors of the prostate relative to the reference, tissue deformation parameters, movement trajectories of surgical instruments, and micro-displacement features caused by physiological movements.

[0012] Noise suppression processing is performed according to the first image to obtain a second image after noise reduction.

[0013] Through time sequence analysis of the second image, it is determined whether the dynamic change exceeds a preset threshold range.

[0014] If the dynamic change exceeds the preset threshold range, the second image is further smoothed to filter out high-frequency noise and retain tissue edge information to obtain an optimized enhanced image.

[0015] In one embodiment, based on the enhanced image and the preoperative three-dimensional prostate image, a three-dimensional position offset vector is obtained by multi-modal image registration combined with three-dimensional anatomical structure references, and a correction parameter is constructed to generate optimized three-dimensional posture data of the prostate based on the anatomical structure constraints of the prostate, including:

[0016] The enhanced image and the preoperative three-dimensional prostate image are aligned by a multi-modal image registration algorithm combined with the three-dimensional anatomical structure references of the prostate to determine the three-dimensional position offset of the prostate.

[0017] The three-dimensional position offset vector corresponding to the three-dimensional position offset is extracted, and a comparison is made with a preset adaptive dynamic threshold to determine whether the three-dimensional position offset vector exceeds the threshold range; the dynamic threshold is adjusted based on the surgical stage and the characteristics of the target region of the prostate.

[0018] If the three-dimensional position offset vector exceeds the dynamic threshold, the posture change trend of the prostate is predicted in real time based on a deep learning model to generate a posture correction instruction containing three-dimensional spatial posture parameters of the prostate.

[0019] The mechanical arm of the surgical robot and the imaging device are driven according to the posture correction instruction to adjust the posture, and the adjusted real-time enhanced image data is acquired synchronously.

[0020] The adjusted real-time enhanced image data is re-extracted for a three-dimensional displacement vector of the prostate using a multi-modal image registration algorithm.

[0021] If the re-extracted three-dimensional displacement vector still exceeds the dynamic threshold, a secondary correction parameter is generated based on the trend of the previous two three-dimensional displacement vectors and the anatomical structure constraints of the prostate to obtain optimized three-dimensional posture data of the prostate; the anatomical structure constraints combine historical data trends and anatomical restrictions.

[0022] In one embodiment, the risk assessment value is calculated and the corresponding risk level is mapped after determining the boundary region distribution characteristics in combination with the three-dimensional posture data of the prostate and the three-dimensional image of the prostate. If the risk level exceeds the preset threshold, a path correction instruction is generated to obtain a surgical instrument navigation path, including:

[0023] The relative position deviation value is calculated based on the three-dimensional posture data of the prostate through coordinate transformation to obtain the relative position deviation value of the surgical instrument and the prostate.

[0024] The relative position deviation value is combined with the three-dimensional image of the prostate, and a boundary segmentation algorithm is used to divide the abnormal hyperplasia region and the normal tissue to determine the boundary region distribution characteristics; the boundary region distribution characteristics include the three-dimensional spatial coordinate range of the abnormal hyperplasia region, the minimum distance from the key anatomical structure, the irregularity parameter of the boundary contour, and the volume proportion of the abnormal region in the prostate.

[0025] According to the boundary region distribution characteristics, the three-dimensional position displacement amount is calculated by quantifying the feature weighted sum to calculate the risk assessment value and map the corresponding risk level.

[0026] If the risk level exceeds the preset threshold, a path correction instruction is generated to obtain the corrected instrument path data.

[0027] A new position displacement amount is calculated through the corrected instrument path data, and the boundary region risk assessment result is updated.

[0028] The updated risk assessment result is input into a path optimization model to generate a final surgical instrument navigation path; the path optimization model is constructed based on a reinforcement learning algorithm.

[0029] In one embodiment, the risk assessment value is calculated by the following formula:

[0030]

[0031] wherein, represents the risk assessment value, represents the minimum distance from the abnormal hyperplasia region to the key anatomical structure, represents the preset safety distance threshold, represents the boundary contour irregularity, , Indicates the length of the outline. Indicates the measurement scale. Indicates the percentage of volume in the abnormal region. , Represents the three-dimensional volume of the abnormally proliferating region. This represents the total three-dimensional volume of the prostate gland. This represents the ratio of the 3D position offset to a preset safety offset threshold. , This represents the real-time three-dimensional position vector of the prostate gland. This represents the preset reference three-dimensional position vector of the prostate. This represents the preset three-dimensional position offset safety threshold vector. This indicates the preset three-dimensional position offset safety threshold. , , , Indicates the feature weights.

[0032] In one embodiment, a clinical recovery model is constructed based on the surgical instrument navigation path and the prostate anatomy to generate simulated postoperative recovery index data. The simulated recovery index data is adjusted and verified to meet the preset recovery index requirements, resulting in the final surgical execution plan, including:

[0033] Acquire surgical operation sequence data based on the updated surgical instrument navigation path; the surgical operation sequence data includes path adjustment parameters and instrument movement timing information.

[0034] A clinical recovery model was constructed based on surgical procedure sequence data and the anatomical characteristics of the prostate, generating simulated data of postoperative recovery indicators. The simulated data of recovery indicators included wound healing period and urinary function recovery score.

[0035] The simulated data of the recovery indicators are compared with the preset graded recovery thresholds, and the preliminary assessment results, including the qualified items, the deviation items and the risk level, are output.

[0036] Key influencing factors were extracted from the preliminary assessment results to determine the optimization direction of the final surgical execution plan. Key influencing factors included path deviation, instrument operation accuracy, and prediction of residual hyperplasia area.

[0037] Based on the optimization direction, the detailed parameters in the surgical operation sequence are adjusted to obtain the updated surgical execution plan data; the detailed parameters include instrument movement speed and positioning accuracy threshold.

[0038] Verify the simulation results of the recovery indicators corresponding to the updated surgical execution plan data. If the simulation results of the recovery indicators meet the preset recovery indicator requirements, then output the final surgical execution plan.

[0039] In a second aspect, the application further provides an AI image navigation device for minimally invasive surgery of prostatic hyperplasia, which comprises:

[0040] An image acquisition enhancement module is configured to acquire a three-dimensional image of the prostate and a real-time image sequence during the surgery, and to obtain an optimized enhanced image by processing the real-time image sequence after preliminary filtering and in combination with the three-dimensional image of the prostate.

[0041] A data registration correction module is configured to obtain a three-dimensional offset vector by combining a three-dimensional anatomical structure reference and a multimodal image registration based on the enhanced image and the preoperative three-dimensional image of the prostate, and to generate optimized three-dimensional posture data of the prostate by constructing a correction parameter in combination with a prostate anatomical structure constraint.

[0042] A risk assessment correction module is configured to determine a risk assessment value and map a corresponding risk level after determining a boundary region distribution feature in combination with the three-dimensional posture data of the prostate and the three-dimensional image of the prostate, and to generate a path correction instruction to obtain a surgical instrument navigation path if the risk level exceeds a preset threshold.

[0043] A surgery scheme optimization module is configured to construct a clinical recovery model to generate postoperative recovery index simulation data based on the surgical instrument navigation path in combination with the prostate anatomical structure, to adjust and verify the recovery index simulation data until a preset recovery index requirement is met, and to obtain a final surgery execution scheme.

[0044] In a third aspect, the application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the foregoing method when executing the computer program.

[0045] In a fourth aspect, the application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the foregoing method.

[0046] The AI image navigation method, device, computer device and storage medium for the prostate hyperplasia minimally invasive surgery provided by the application, obtain a prostate three-dimensional image and a real-time image sequence in surgery, after preliminary filtering of the real-time image sequence, process in combination with the prostate three-dimensional image to obtain an optimized enhanced image; based on the enhanced image and the preoperative prostate three-dimensional image, obtain a three-dimensional offset vector through a multi-modal image registration algorithm combined with a three-dimensional anatomical structure reference, construct a correction parameter in combination with a prostate anatomical structure constraint to generate optimized prostate three-dimensional posture data; in combination with the prostate three-dimensional posture data and the prostate three-dimensional image, determine a boundary region distribution feature to calculate a risk assessment value, map a corresponding risk level, if the risk level exceeds a preset threshold, generate a path correction instruction to obtain a surgical instrument navigation path; based on the surgical instrument navigation path, construct a clinical recovery model in combination with a prostate anatomical structure to generate postoperative recovery index simulation data, adjust and verify the recovery index simulation data until a preset recovery index requirement is met to obtain a final surgical execution scheme. The method realizes the precision and personalization of prostate hyperplasia minimally invasive surgery navigation. Through the combination of multi-modal image registration and prostate anatomical structure constraint, the accuracy of prostate posture positioning is improved, providing a reliable spatial reference for subsequent path planning; through boundary region distribution feature extraction and risk assessment value calculation, quantitative judgment of surgical risk is realized, in combination with the path correction instruction, the safety of instrument navigation is ensured, and the damage risk to key anatomical structures such as the urethra and rectum is reduced; through the construction and simulation data verification of the clinical recovery model, the recovery effect of the surgical scheme can be pre-judged before surgery, the scheme is adjusted and optimized to meet the preset recovery index, the effectiveness of the surgery and the postoperative recovery quality are improved, and comprehensive technical support is provided for the precise implementation of minimally invasive surgery. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the following will briefly introduce the drawings needed to be used in the embodiment or related art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0048] Figure 1 The flow chart of the AI image navigation method for the prostate hyperplasia minimally invasive surgery provided by the embodiment of the present application;

[0049] Figure 2 The structural block diagram of the AI image navigation device for the prostate hyperplasia minimally invasive surgery provided by the embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0051] In one of the embodiments, as shown in Figure 1 The present application provides an AI image navigation method for minimally invasive surgery of prostate hyperplasia, which can include the following steps:

[0052] Step S101, obtain a prostate three-dimensional image and a real-time image sequence during surgery, and process the real-time image sequence after preliminary filtering in combination with the prostate three-dimensional image to obtain an optimized enhanced image.

[0053] Firstly, a prostate three-dimensional image is obtained by a preoperative medical imaging device (such as magnetic resonance imaging, MRI), which contains the anatomical structure details of the prostate (such as the capsule, urethra, hyperplasia area, etc.), serving as an anatomical reference for subsequent processing; at the same time, an image sequence of the surgical area is collected in real time by an intraoperative imaging device (such as an ultrasonic probe, an endoscope), which reflects the dynamic position changes of the prostate during the surgery. Next, the real-time image sequence is preliminarily filtered: pixel gray threshold screening is used to remove background noise (such as surgical room environmental light interference), and frame similarity analysis is used to eliminate redundant frames (such as images with no obvious changes in adjacent frames), and then Gaussian low-pass filtering is used to smooth the image texture and retain the key structure edges, to obtain clear filtered images. Finally, the filtered real-time images are preliminarily registered and fused with the preoperative prostate three-dimensional image, the common anatomical landmark points (such as the apex of the prostate, the bladder neck) of the two are extracted, the spatial position of the real-time images is calibrated, and finally the optimized enhanced image is outputted, which retains the real-time dynamic information during the surgery and fuses the precise anatomical structure before the surgery.

[0054] Step S102, based on the enhanced image and the preoperative prostate three-dimensional image, a three-dimensional offset vector is obtained by combining the multi-modal image registration of the three-dimensional anatomical structure reference, and a correction parameter is constructed in combination with the prostate anatomical structure constraint to generate optimized prostate three-dimensional pose data.

[0055] Illustratively, the enhanced image (intraoperative real-time dynamic image) and the preoperative prostate three-dimensional image (anatomical reference image) are combined; due to the different origins of the two images (multi-modal), a registration algorithm combined with three-dimensional anatomical structure reference is required - taking the clear anatomical landmarks in the preoperative three-dimensional image (such as the urethral axis, the prostate capsule contour) as the reference, the corresponding landmark points in the enhanced image are found, and through the calculation of the coordinate difference of the landmark points in the three-dimensional space, the three-dimensional displacement vector of the prostate (including the displacement amount of x, y, z axes and the rotation angle) is obtained. Subsequently, the correction parameters are constructed in combination with the prostate anatomical structure constraint: the anatomical structure constraint is based on the physiological anatomical range of the prostate (such as the prostate capsule cannot exceed the pelvic bone boundary, the urethral position is relatively fixed), the displacement vector is reasonably checked (such as eliminating abnormal displacement values that exceed the physiological range), and then the checked displacement vector is converted into pose correction parameters through matrix transformation. Finally, the optimized prostate three-dimensional pose data is output, which accurately describes the real-time spatial pose of the intraoperative prostate relative to the preoperative reference position.

[0056] Step S103, after determining the boundary region distribution characteristics by combining the prostate three-dimensional pose data and the prostate three-dimensional image, calculate the risk assessment value and map the corresponding risk level, if the risk level exceeds the preset threshold, generate the path correction instruction to obtain the surgical instrument navigation path.

[0057] Preferably, based on the three-dimensional pose data, the real-time position of the surgical instrument (obtained by the positioning sensor on the instrument) is matched with the real-time position of the prostate through coordinate transformation, and the relative position deviation value (such as the distance between the instrument tip and the hyperplasia region, and the distance from the urethra) is calculated. Subsequently, combined with the preoperative three-dimensional image, the boundary segmentation algorithm (such as the U-Net model based on deep learning) is used to divide the abnormal hyperplasia region and the normal tissue, and the boundary region distribution characteristics are extracted - specifically including the three-dimensional space coordinate range of the abnormal hyperplasia region (determining the position and size of the hyperplasia tissue), the minimum distance from the key anatomical structures such as the urethra / rectum (evaluating the damage risk), the irregularity parameter of the boundary contour (reflecting the morphological complexity of the hyperplasia tissue), and the volume proportion of the abnormal region in the prostate (quantifying the severity of hyperplasia). Then, the risk assessment value is calculated by weighted sum of quantitative features (such as distance weight 0.4, irregularity weight 0.2, volume proportion weight 0.4), and the corresponding risk level is mapped according to the preset risk grading standard (such as low risk ≤3, medium risk 3-6, high risk ≥6). If the risk level exceeds the preset threshold (such as high risk ≥6), the path correction instruction is generated according to the boundary region distribution characteristics (such as adjusting the instrument movement direction to avoid the urethra, reducing the operation range to avoid damage to normal tissues), and finally the surgical instrument navigation path is output - the path clearly shows the movement trajectory of the instrument in the three-dimensional space, ensuring that the instrument accurately acts on the hyperplasia region while avoiding key risk structures.

[0058] Step S104, based on the navigation path of the surgical instrument, a clinical recovery model is constructed to generate postoperative recovery index simulation data. The recovery index simulation data is adjusted and verified to meet the pre-set recovery index requirements, and the final surgical execution plan is obtained.

[0059] Further, based on the navigation path, surgical operation sequence information (such as instrument movement speed, positioning accuracy threshold, and hyperplasia area resection sequence) is extracted, combined with prostate anatomical structure characteristics (such as capsule thickness and blood vessel distribution) and historical clinical recovery data (such as wound healing period of similar cases and urination function recovery score), a clinical recovery model is constructed. The model simulates the degree of damage to the prostate tissue by surgery, the wound healing process, and the change trend of postoperative recovery indexes such as urination function and sexual function. Subsequently, the model outputs postoperative recovery index simulation data, including wound healing period (such as expected 7-day healing), urination function recovery score (such as postoperative 1-week score ≥80 points), and complication incidence (such as bleeding risk ≤5%). Compare the simulation data with the pre-set grading recovery threshold (such as no obvious bleeding on the first day after surgery, smooth urination on the third day, and healing up to standard on the seventh day). If there is a deviation (such as a simulated healing period of 10 days, exceeding the pre-set 7-day threshold), extract the key influencing factors (such as insufficient instrument operation accuracy leading to a large wound, and path deviation damaging blood vessels), and determine the optimization direction (such as improving instrument positioning accuracy and adjusting the path to avoid blood vessel dense areas). Based on the optimization direction, adjust the detailed parameters of the surgical operation sequence (such as reducing the instrument movement speed from 5mm / s to 3mm / s and tightening the positioning accuracy threshold from 0.5mm to 0.3mm), and generate updated surgical execution plan data. The simulation recovery index corresponding to the plan is verified by the clinical recovery model again. If all core indexes meet the pre-set requirements (such as healing period ≤7 days, urination score ≥80 points, and complication incidence ≤3%), the final surgical execution plan is output.

[0060] The AI image navigation method for the above-mentioned minimally invasive surgery of prostate hyperplasia acquires a three-dimensional image of the prostate and a real-time image sequence in surgery, processes the real-time image sequence after preliminary filtering, combines the three-dimensional image of the prostate, and obtains an optimized enhanced image; based on the enhanced image and the three-dimensional image of the prostate before surgery, a three-dimensional offset vector is obtained through a multi-modal image registration algorithm combined with a three-dimensional anatomical structure reference, a correction parameter is constructed combined with the anatomical structure of the prostate, an optimized three-dimensional posture data of the prostate is generated, a risk assessment value is calculated after determining the boundary region distribution characteristics combined with the three-dimensional posture data of the prostate and the three-dimensional image of the prostate, a corresponding risk level is mapped, if the risk level exceeds a preset threshold, a path correction instruction is generated, and a surgical instrument navigation path is obtained; based on the surgical instrument navigation path, a clinical recovery model is constructed combined with the anatomical structure of the prostate, postoperative recovery index simulation data is generated, the recovery index simulation data is adjusted and verified until the preset recovery index requirement is met, and a final surgical execution scheme is obtained. The method realizes the precision and personalization of prostate hyperplasia minimally invasive surgery navigation. Through the combination of multi-modal image registration and prostate anatomical structure constraint, the accuracy of prostate posture positioning is improved, providing a reliable spatial reference for subsequent path planning; through boundary region distribution feature extraction and risk assessment value calculation, the quantitative judgment of surgical risk is realized, combined with the path correction instruction, the safety of instrument navigation is ensured, and the damage risk of key anatomical structures such as urethra and rectum is reduced; through the construction and simulation data verification of the clinical recovery model, the recovery effect of the surgical scheme can be predicted before surgery, the scheme is adjusted and optimized to meet the preset recovery index, and the effectiveness of the surgery and the postoperative recovery quality are improved, providing comprehensive technical support for the precise implementation of minimally invasive surgery.

[0061] In one embodiment, acquiring a three-dimensional image of the prostate and a real-time image sequence in surgery, processing the real-time image sequence after preliminary filtering combined with the three-dimensional image of the prostate to obtain an optimized enhanced image can include the following steps:

[0062] Step S201, acquiring a three-dimensional image of the prostate and a real-time image sequence in surgery, and preliminarily filtering the real-time image sequence to obtain a filtered image sequence; the preliminary filtering includes pixel gray threshold screening, inter-frame redundant frame elimination and Gaussian low-pass filtering.

[0063] Step S202, combining the anatomical structure reference in the three-dimensional image of the prostate to extract dynamic change features from the real-time image sequence using a dynamic change detection algorithm to determine the first image of the change region; the dynamic change features include the displacement vector of the prostate relative to the reference, the tissue deformation parameter, the surgical instrument motion trajectory and the micro-displacement features caused by physiological motion.

[0064] Step S203, noise suppression processing is performed according to the first image to obtain a second image after noise suppression.

[0065] Step S204, whether the dynamic change exceeds a preset threshold range is determined by time series analysis on the second image.

[0066] Step S205, if the dynamic change exceeds the preset threshold range, further smoothing processing is performed on the second image to filter out high-frequency noise and retain tissue edge information to obtain an optimized enhanced image.

[0067] Specifically, a three-dimensional image of the prostate is obtained by a preoperative medical imaging device (such as magnetic resonance imaging, MRI), and a real-time image sequence of the surgical area is collected by an intraoperative imaging device (such as an ultrasonic probe or an endoscope); the real-time image sequence is subjected to pixel gray threshold screening to remove background noise (such as environmental light interference), and inter-frame similarity analysis is performed to remove redundant frames (such as images with no significant change in adjacent frames); then, Gaussian low-pass filtering is used to smooth the image texture and retain the key structure edges to obtain a filtered image sequence; combined with anatomical structure references (such as the urethral axis and the prostate capsule contour) in the three-dimensional image of the prostate, a dynamic change detection algorithm is used to extract dynamic change features (including displacement vectors of the prostate relative to the references, tissue deformation parameters, surgical instrument movement trajectories, and micro-displacement features caused by physiological movements such as respiration) from the above filtered real-time image sequence, and based on the features, a first image reflecting the dynamic changes of the prostate and instruments is determined; noise suppression processing (such as adaptive median filtering) is performed on the first image to reduce the influence of residual noise on image quality, and a second image after noise suppression is obtained; time series analysis (such as calculating the pixel change rate between adjacent frames and the moving distance of feature points) is performed on the second image to determine whether the dynamic change exceeds a preset threshold range (such as a displacement change rate threshold and a deformation amplitude threshold); if the dynamic change exceeds the threshold range, further smoothing processing (such as bilateral filtering) is performed on the second image to filter out high-frequency noise while retaining tissue edge information, and finally an optimized enhanced image is obtained.

[0068] The embodiment realizes the accurate optimization of intraoperative image quality through the organic combination of image processing and dynamic feature extraction, and provides high-quality data support for subsequent surgical navigation. The preliminary filtering and noise suppression steps effectively reduce the interference of background noise and redundant information on the image, and improve the clarity of the image; the dynamic change detection algorithm accurately captures the dynamic characteristics of the prostate and surgical instruments, solving the image offset problem caused by physiological movement and instrument operation during surgery; further smoothing processing removes high-frequency noise while preserving the edges of key anatomical structures, ensuring that the enhanced image has high signal-to-noise ratio and completely retains anatomical details and dynamic information. The enhanced image can provide reliable input data for subsequent multi-modal image registration, prostate posture correction and other steps, significantly improving the accuracy of surgical navigation and laying a solid foundation for the safe implementation of prostate hyperplasia minimally invasive surgery.

[0069] In one of the embodiments, based on the enhanced image and the preoperative prostate three-dimensional image, a three-dimensional offset vector is obtained through multi-modal image registration combined with a three-dimensional anatomical structure reference, and a correction parameter is generated by combining the prostate anatomical structure constraint to construct the optimized prostate three-dimensional posture data, which can include the following steps:

[0070] Step S301, align the enhanced image and the preoperative prostate three-dimensional image by a multi-modal image registration algorithm combined with the three-dimensional anatomical structure reference of the prostate, and determine the three-dimensional position offset of the prostate.

[0071] Step S302, extract the three-dimensional offset vector corresponding to the three-dimensional position offset, and compare it with the preset adaptive dynamic threshold to determine whether the three-dimensional offset vector exceeds the threshold range; the dynamic threshold is adjusted based on the surgical stage and the characteristics of the prostate target region.

[0072] Step S303, if the three-dimensional offset vector exceeds the dynamic threshold, real-time predict the prostate posture change trend based on a deep learning model, and generate a posture correction instruction containing the three-dimensional spatial posture parameter of the prostate.

[0073] Step S304, drive the mechanical arm of the surgical robot and the imaging device to adjust the posture according to the posture correction instruction, and synchronously acquire the adjusted real-time enhanced image data.

[0074] Step S305, re-extract the three-dimensional offset vector of the prostate from the adjusted real-time enhanced image data using the multi-modal image registration algorithm.

[0075] Step S306, if the re-extracted three-dimensional offset vector still exceeds the dynamic threshold, generate a secondary correction parameter based on the change trend of the previous two three-dimensional offset vectors and the anatomical structure constraint of the prostate, and obtain the optimized three-dimensional posture data of the prostate; the anatomical structure constraint combines historical data trends and anatomical restrictions.

[0076] Specifically, by combining the multi-modal image registration algorithm of the prostate three-dimensional anatomical reference (such as the urethral axis, the prostate capsule contour, and other preoperative anatomical landmarks), the enhanced image optimized during the operation is aligned with the preoperative three-dimensional image of the prostate, and the positional offset of the prostate in the three-dimensional space (including x, y, z axis displacement and rotation angle) is calculated; the three-dimensional positional offset corresponding to the three-dimensional offset vector is extracted, and the offset vector is compared with the preset adaptive dynamic threshold to determine whether the offset vector exceeds the threshold range - the dynamic threshold can be adjusted in real time according to the operation stage (such as the puncture stage, the resection stage) and the target region characteristics of the prostate (such as the hyperplastic nodule region, the sensitive region around the urethra); if the three-dimensional offset vector exceeds the dynamic threshold, the deep learning model (input historical offset data, anatomical structure parameters) is used to predict the posture change trend of the prostate in real time, and a posture correction instruction containing the three-dimensional spatial posture parameters of the prostate (such as target position coordinates, rotation angle) is generated; according to the posture correction instruction, the mechanical arm of the surgical robot (adjusting the spatial position of the instrument) and the imaging device (synchronously correcting the shooting angle) are driven to complete the posture adjustment, and the adjusted real-time enhanced image data is obtained; the three-dimensional offset vector of the prostate is extracted again using the above multi-modal image registration algorithm to perform secondary verification; if the re-extracted three-dimensional offset vector still exceeds the dynamic threshold, the secondary correction parameters (including dynamic compensation coefficient, adjustment step) are generated based on the change trend of the first two three-dimensional offset vectors (such as offset increment / decrement rule) and the anatomical structure constraints of the prostate (combined with historical surgical data trends and anatomical restrictions such as prostate capsule and urethra), and the optimized three-dimensional posture data of the prostate is finally obtained.

[0077] The embodiment realizes accurate and real-time correction of the three-dimensional posture of the prostate by organically combining multi-modal registration, adaptive threshold judgment and correction mechanism, and provides a reliable spatial reference for surgical navigation. The multi-modal image registration algorithm combined with anatomical structure improves the accuracy of three-dimensional positional offset calculation and avoids registration errors in a single image modality; the adaptive dynamic threshold is dynamically adjusted according to the surgical scene, which adapts to the accuracy requirements of different stages and different regions, and solves the problem of insufficient adaptability of fixed thresholds; the real-time prediction of the posture change trend by the deep learning model shortens the generation time of the correction instruction and ensures the real-time performance of the operation; it not only ensures the robustness of the posture correction (avoids incomplete correction at a time), but also avoids damage to key structures caused by excessive correction through anatomical restrictions, significantly improving the accuracy and safety of surgical navigation.

[0078] In one of the embodiments, after determining the boundary region distribution characteristics based on the three-dimensional posture data of the prostate and the three-dimensional image of the prostate, the risk assessment value is calculated and the corresponding risk level is mapped, and if the risk level exceeds the preset threshold, the path correction instruction is generated to obtain the surgical instrument navigation path, which can include the following steps:

[0079] Step S401, calculate the relative position of the surgical instrument and the prostate based on the three-dimensional prostate posture data through coordinate transformation, and obtain a relative position deviation value.

[0080] Step S402, combine the relative position deviation value with the three-dimensional image of the prostate, divide the abnormal hyperplasia region and the normal tissue by using a boundary segmentation algorithm, and determine the boundary region distribution characteristics; the boundary region distribution characteristics include the three-dimensional space coordinate range of the abnormal hyperplasia region, the minimum distance from the key anatomical structure, the irregularity parameter of the boundary contour, and the volume proportion of the abnormal region in the prostate.

[0081] Step S403, according to the boundary region distribution characteristics and the three-dimensional position offset, calculate a risk assessment value by weighted summation of quantified characteristics, and map the corresponding risk level.

[0082] Step S404, if the risk level exceeds a preset threshold, generate a path correction instruction, and obtain corrected instrument path data.

[0083] Step S405, calculate a new position offset by using the corrected instrument path data, and update the boundary region risk assessment result.

[0084] Step S406, input the updated risk assessment result into a path optimization model to generate a final surgical instrument navigation path; the path optimization model is constructed based on a reinforcement learning algorithm.

[0085] Based on the three-dimensional posture data of the prostate (real-time spatial position coordinates), the relative position of the tip of the surgical instrument and the key area of the prostate (such as the center of the abnormal hyperplasia area) is calculated through coordinate transformation (such as affine transformation) under the surgical coordinate system, and a quantitative relative position deviation value (unit: mm) is obtained; the deviation value is fused with the preoperative three-dimensional image of the prostate (including complete anatomical structure), and a boundary segmentation algorithm (such as U-Net deep learning model) is used to automatically divide the abnormal hyperplasia area and the normal prostate tissue, and the boundary area distribution characteristics are extracted, including: the three-dimensional spatial coordinate range (x / y / z axis start and end points) of the abnormal hyperplasia area, the minimum straight-line distance from the abnormal hyperplasia area to the urethra / rectum and other key anatomical structures, the irregularity parameter (calculated by fractal dimension) of the boundary contour, and the volume proportion of the abnormal area in the prostate (abnormal area volume / prostate total volume ×100%); based on the above characteristics, combined with the three-dimensional position offset of the prostate (real-time displacement relative to the preoperative reference), the risk assessment value is calculated through a quantitative feature weighted summation formula, and then the corresponding risk level is mapped according to the preset grading standard (such as low risk ≤3, medium risk 3-6, and high risk ≥6); if the risk level exceeds the preset threshold (such as high risk ≥6), path correction instructions (such as adjusting the movement direction of the instrument, increasing the safety distance) are generated for the high-risk features (such as being too close to the urethra), and the corrected instrument path data (including the adjusted motion trajectory coordinate sequence) is obtained; based on the corrected path data, the position offset of the instrument and the prostate is recalculated, and the boundary area risk assessment result is updated (repeat the above risk calculation steps); the updated risk assessment result is input into the path optimization model constructed based on the reinforcement learning algorithm (with "lowest risk, shortest path, and smoothest instrument movement" as the reward function), and the final surgical instrument navigation path (including real-time adjusted speed, acceleration, and positioning accuracy parameters) is output.

[0086] The embodiment realizes the precision and individual optimization of the surgical instrument navigation path. The quantitative calculation of the relative position deviation value provides objective data support for risk assessment, avoiding subjective judgment errors; the boundary segmentation algorithm accurately identifies the boundary between the abnormal hyperplasia area and the key structure, and the risk assessment based on multi-dimensional distribution characteristics (distance, volume, and morphology) comprehensively covers potential risk points (such as urethral injury and hyperplasia residue) in the surgery; the path optimization model driven by reinforcement learning can meet the risk control requirements while considering the efficiency of the path and the stability of the instrument movement, solving the problem of "imbalance between precision and efficiency" in traditional path planning; and ensures that the navigation path can dynamically adapt to the position change of the prostate during the surgery, significantly improving the safety, precision, and robustness of the surgery, and providing a core technical support for the precise implementation of the prostate hyperplasia minimally invasive surgery.

[0087] In one of the embodiments, the risk assessment value can be calculated by the following formula:

[0088]

[0089] wherein, represents a risk assessment value, represents a minimum distance between the hyperplasia region and the key anatomical structure, represents a preset safety distance threshold, represents a boundary contour irregularity, , represents a contour length, represents a measurement scale, represents an abnormal region volume proportion, , represents a three-dimensional volume of the hyperplasia region, represents a total three-dimensional volume of the prostate, represents a ratio of the three-dimensional position offset to the preset safety offset threshold, , represents a real-time three-dimensional position vector of the prostate, represents a preset reference three-dimensional position vector of the prostate, represents a preset three-dimensional position offset safety threshold vector, represents a preset three-dimensional position offset safety threshold, , , , represents a feature weight.

[0090] The embodiment integrates four core risk factors of “key anatomical structure distance, boundary shape irregularity, abnormal region volume proportion, and three-dimensional position dynamic change”, and realizes multi-dimensional risk quantification fusion by combining clinical customized weights: Item accurately describes the safety distance relationship between the abnormal region and the urethra / rectum and other key structures, avoiding the risk of close-range damage; Table (fractal dimension) quantifies boundary irregularity, adapting to the risk differences of different morphological hyperplasia (such as complex boundaries of malignant tendency hyperplasia); directly reflects the influence of hyperplasia severity on surgical difficulty; The offset amount exceeding the safety threshold is exponentially amplified, and the risk accumulation caused by dynamic offset is sensitively captured. The overall formula not only realizes comprehensive coverage of risk factors, but also takes into account clinical priorities (such as key structure protection priority) through nonlinear response and weight adjustment. The output of the quantitative risk assessment value ( ) provides accurate decision-making basis for subsequent path correction and optimization model, effectively improving the objectivity and robustness of surgical risk prediction, and laying a core quantitative foundation for safe and accurate implementation of minimally invasive surgery.

[0091] In one of the embodiments, based on the navigation path of the surgical instrument combined with the prostate anatomical structure, a clinical recovery model is constructed to generate postoperative recovery index simulation data, the recovery index simulation data is adjusted and verified to meet the preset recovery index requirements, and a final surgical execution scheme is obtained, which can include the following steps:

[0092] In step S501, surgical operation sequence data based on the navigation path of the surgical instrument is obtained; the surgical operation sequence data includes path adjustment parameters and instrument action timing information.

[0093] In step S502, a clinical recovery model is constructed according to the surgical operation sequence data combined with the prostate anatomical structure characteristics, and postoperative recovery index simulation data is generated; the recovery index simulation data includes wound healing period and urination function recovery score.

[0094] In step S503, the recovery index simulation data is compared with the preset hierarchical recovery threshold, and a preliminary evaluation result including the compliance item, the deviation item and the risk level is output.

[0095] In step S504, key influence factors are extracted from the preliminary evaluation result to determine the optimization direction of the final surgical execution scheme; the key influence factors include path deviation amplitude, instrument operation accuracy and residual estimation of hyperplasia area.

[0096] In step S505, based on the optimization direction, the detail parameters in the surgical operation sequence are adjusted to obtain updated surgical execution scheme data; the detail parameters include instrument movement speed and positioning accuracy threshold.

[0097] In step S506, the recovery index simulation result corresponding to the updated surgical execution scheme data is verified, and if the recovery index simulation result meets the preset recovery index requirements, the final surgical execution scheme is output.

[0098] Specifically, surgical operation sequence data based on surgical instrument navigation path updates is obtained, which contains path adjustment parameters (such as coordinate correction values of instrument motion trajectories) and instrument action timing information (such as execution time and interval duration of each action); according to the surgical operation sequence data, a clinical recovery model is constructed in combination with prostate anatomical structure characteristics (such as capsule thickness, blood vessel distribution density, and anatomical relationship between the urethra and the hyperplasia region), through simulating the damage range of the prostate tissue, the healing process and the functional impact of the surgery, postoperative recovery index simulation data is generated, including wound healing period (such as expected healing days) and urinary function recovery score (such as functional score at 1 week / 1 month after surgery); the recovery index simulation data is compared with the preset grading recovery threshold (set according to key time nodes such as 1 day, 3 days, 7 days, and 30 days after surgery, such as wound healing period ≤7 days for standard and urinary function recovery score ≥80 points for standard), and the preliminary evaluation result is output, which contains standard items (such as wound healing period standard), deviation items (such as urinary function recovery score below the threshold), and risk levels (such as high risk, medium risk, and low risk); the preliminary evaluation result is analyzed to extract key influencing factors, including path deviation amplitude (such as the maximum deviation distance between the actual path and the planned path), instrument operation precision (such as instrument positioning error value), and hyperplasia region residual estimation (such as the volume proportion of unremoved hyperplastic tissue), based on which the optimization direction of the final surgical execution scheme is determined (such as adjusting the path planning algorithm for large path deviation amplitude, and improving instrument positioning precision for insufficient instrument operation precision); based on the optimization direction, the details parameters in the surgical operation sequence are adjusted, including instrument motion speed (such as adjusting from 5mm / s to 3mm / s) and positioning precision threshold (such as tightening from 0.5mm to 0.3mm), to obtain updated surgical execution scheme data; the recovery index simulation result corresponding to the updated surgical execution scheme data is verified, and if the simulation result meets the preset recovery index requirement (such as core index standard rate ≥90%), the final surgical execution scheme is output.

[0099] The embodiment is based on the construction of a clinical recovery model based on the individual prostate anatomy of a patient, ensures the accuracy of the postoperative recovery simulation, provides objective and quantitative basis for scheme evaluation, avoids the limitations of traditional experience-dependent judgment; Precise extraction of key influencing factors makes the optimization direction more targeted, which can directly locate the weak links in the scheme (such as path deviation, insufficient operation accuracy), improves the optimization efficiency; Fine adjustment of detailed parameters and recovery index verification mechanism ensure that the final scheme not only meets the core demand of precise resection of hyperplastic tissue, but also maximizes the impact of surgery on prostate function, shortens the postoperative recovery period, and improves the recovery quality (such as reducing wound healing time and improving urinary function recovery effect), effectively solves the problem of traditional surgical scheme “emphasizing surgical implementation and ignoring postoperative recovery”, significantly improves the safety, effectiveness and patient prognosis experience of surgery, and provides complete technical support for the precision and individualized treatment of prostate hyperplasia minimally invasive surgery.

[0100] In one of the embodiments, as shown in Figure 2 The application also provides an AI image navigation device for prostate hyperplasia minimally invasive surgery. The device can include:

[0101] An image acquisition enhancement module 601 is configured to acquire a prostate three-dimensional image and a real-time image sequence during surgery, and to process the real-time image sequence after preliminary filtering combined with the prostate three-dimensional image to obtain an optimized enhanced image.

[0102] A data registration correction module 602 is configured to obtain a three-dimensional offset vector based on the enhanced image and the preoperative prostate three-dimensional image by combining a three-dimensional anatomical structure reference multi-modal image registration, and to construct a correction parameter combined with the prostate anatomical structure constraint to generate optimized prostate three-dimensional pose data.

[0103] A risk assessment correction module 603 is configured to determine the boundary region distribution characteristics based on the prostate three-dimensional pose data and the prostate three-dimensional image, calculate the risk assessment value and map the corresponding risk level, and generate a path correction instruction to obtain a surgical instrument navigation path if the risk level exceeds a preset threshold.

[0104] A surgical scheme optimization module 604 is configured to construct a clinical recovery model based on the surgical instrument navigation path combined with the prostate anatomical structure to generate postoperative recovery index simulation data, adjust and verify the recovery index simulation data to meet the preset recovery index requirements, and obtain a final surgical execution scheme.

[0105] The AI image navigation device for the above-mentioned minimally invasive surgery of prostate hyperplasia, the image acquisition enhancement module first acquires the preoperative prostate three-dimensional image (MRI data) and the intraoperative real-time image sequence (such as ultrasound, endoscopic image), after gray threshold screening, redundant frame elimination and Gaussian filter pretreatment on the real-time sequence, the anatomical structure information of the preoperative three-dimensional image is fused, and the optimized enhanced image is output, providing high-quality data basis for subsequent processing; the data registration correction module receives the enhanced image and the preoperative three-dimensional image, calculates the prostate three-dimensional position offset vector through the multi-modal registration algorithm combined with the prostate anatomical structure reference (such as the urethral axis, the capsule contour), and constructs the correction parameter combined with the anatomical structure constraint (such as the physiological motion range limit of the prostate), generates real-time updated prostate three-dimensional posture data, and ensures the accurate positioning of the intraoperative prostate position; the risk assessment correction module uses the prostate three-dimensional posture data to obtain the relative position deviation of the surgical instrument and the prostate through coordinate transformation, fuses the three-dimensional coordinate range of the abnormal hyperplasia area extracted by the boundary segmentation algorithm from the preoperative three-dimensional image, the minimum distance from the key structure (urethra / rectum), the boundary irregularity and the volume proportion, and combines the three-dimensional position offset amount to calculate the risk assessment value and map the risk level through the quantitative feature weighted sum formula, if the preset threshold is exceeded, the path correction instruction is generated, the corrected instrument path data is obtained, the offset amount is recalculated based on the new path and the risk assessment result is updated, forming a closed loop of risk assessment and path correction; the surgical scheme optimization module obtains the surgical operation sequence containing the path adjustment parameter and the instrument action time sequence based on the corrected instrument path data, constructs the clinical recovery model combined with the anatomical structure characteristics of the prostate (such as blood vessel distribution, capsule thickness), simulates the postoperative index data such as wound healing period and urination function recovery score, compares with the preset graded recovery threshold to output the preliminary evaluation result, extracts the path deviation amplitude, instrument operation precision, hyperplasia residue estimation and other key influence factors, determines the optimization direction and adjusts the instrument movement speed, positioning accuracy threshold and other detail parameters, verifies the recovery index simulation result again until the preset requirement is met, and finally outputs the personalized surgical execution scheme.

[0106] The embodiment realizes the precision, personalization and intelligence of the minimally invasive surgery navigation of prostate hyperplasia. The image acquisition enhancement module provides high signal-to-noise ratio and high anatomical detail image data for subsequent steps through multi-modal data fusion and noise suppression; the data registration correction module solves the problem of position deviation of the prostate caused by respiration and body position change during the operation by combining anatomical reference and dynamic constraint, and improves the attitude positioning accuracy; the risk assessment correction module quantizes the risk through multi-dimensional features and dynamically corrects the path, effectively reducing the damage risk of the instrument to the urethra, rectum and other key structures; the surgery scheme optimization module realizes the preoperative prediction and scheme optimization of the surgery effect through clinical recovery simulation and parameter iterative adjustment, shortens the postoperative recovery period of the patient. The whole system significantly improves the safety, accuracy and efficiency of the operation through the close coupling and data flow closed loop between the modules, and provides technical support for the intelligent development of minimally invasive surgery.

[0107] It should be understood that although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0108] In one embodiment, a computer device is provided, comprising a memory and a processor, the memory stores a computer program, and the processor implements the steps of the AI image navigation method for prostate hyperplasia minimally invasive surgery as described above when executing the computer program.

[0109] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps in each method embodiment described above.

[0110] For the device embodiment, since it basically corresponds to the method embodiment, the relevant part can be seen from the part of the method embodiment. The device embodiment described above is only schematic, wherein the components shown as separate components can or can not be physically separate, and the components shown as a unit can or can not be a physical unit, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure. Those skilled in the art can understand and implement it without creative labor.

[0111] The above-described embodiments only express several implementation manners of the present application, which are described in detail, but cannot be understood as a limitation on the patent scope of the application. It should be pointed out that, for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are all within the protection scope of the present application.

Claims

1. An AI-guided image-guided method for minimally invasive surgery for benign prostatic hyperplasia, characterized in that, The method includes: Three-dimensional images of the prostate and real-time image sequences during surgery are acquired. After preliminary filtering of the real-time image sequences, they are combined with the three-dimensional images of the prostate to obtain optimized enhanced images. Based on the enhanced image and the preoperative three-dimensional prostate image, a three-dimensional offset vector is obtained by multimodal image registration with a three-dimensional anatomical structure benchmark. The optimized three-dimensional prostate pose data is generated by constructing correction parameters based on prostate anatomical structure constraints. After determining the boundary region distribution characteristics by combining the prostate three-dimensional pose data and the prostate three-dimensional image, a risk assessment value is calculated and the corresponding risk level is mapped. If the risk level exceeds a preset threshold, a path correction instruction is generated to obtain the surgical instrument navigation path. Based on the surgical instrument navigation path and the prostate anatomical structure, a clinical recovery model is constructed to generate simulated data of postoperative recovery indicators. The simulated data of recovery indicators is adjusted and verified to meet the preset recovery indicator requirements, and the final surgical execution plan is obtained.

2. The method according to claim 1, characterized in that, The process of acquiring a three-dimensional image of the prostate and a real-time image sequence during surgery, followed by preliminary filtering of the real-time image sequence and processing it in conjunction with the three-dimensional image of the prostate to obtain an optimized enhanced image, includes: A three-dimensional image of the prostate and a real-time image sequence during surgery are acquired. The real-time image sequence is then preliminarily filtered to obtain a filtered image sequence. The preliminary filtering includes pixel grayscale thresholding, inter-frame redundant frame removal, and Gaussian low-pass filtering. By combining the anatomical structural benchmarks in the three-dimensional images of the prostate, a dynamic change detection algorithm is used to extract dynamic change features from the real-time image sequence to determine the first image of the change area; the dynamic change features include the displacement vector of the prostate relative to the benchmark, tissue deformation parameters, surgical instrument movement trajectory, and minute displacement features caused by physiological movement; The first image is subjected to noise suppression processing to obtain a denoised second image. By performing time series analysis on the second image, it is determined whether the dynamic changes exceed a preset threshold range; If the dynamic changes exceed a preset threshold range, the second image is further smoothed to filter out high-frequency noise and retain tissue edge information to obtain an optimized enhanced image.

3. The method according to claim 1, characterized in that, The process involves obtaining a 3D offset vector through multimodal image registration based on the enhanced image and the preoperative 3D prostate image, combined with a 3D anatomical structural benchmark. Then, by incorporating prostate anatomical structural constraints, correction parameters are constructed to generate optimized 3D prostate pose data, including: The enhanced image is aligned with the preoperative three-dimensional prostate image by using a multimodal image registration algorithm based on the three-dimensional anatomical structure of the prostate to determine the three-dimensional positional offset of the prostate. The three-dimensional offset vector corresponding to the three-dimensional position offset is extracted and compared with a preset adaptive dynamic threshold to determine whether the three-dimensional offset vector exceeds the threshold range; the dynamic threshold is adjusted based on the surgical stage and the characteristics of the prostate target area. If the three-dimensional offset vector exceeds the dynamic threshold, the prostate posture change trend is predicted in real time based on the deep learning model, and a posture correction instruction containing the three-dimensional spatial posture parameters of the prostate is generated. The surgical robot's robotic arm and imaging device are driven to adjust their posture according to the posture correction command, and the adjusted real-time enhanced image data is acquired simultaneously. The three-dimensional offset vector of the prostate is re-extracted from the adjusted real-time enhanced image data using a multimodal image registration algorithm; If the re-extracted three-dimensional offset vector still exceeds the dynamic threshold, secondary correction parameters are generated based on the change trend of the three-dimensional offset vectors in the previous two iterations and the constraints of the prostate anatomical structure to obtain optimized prostate three-dimensional pose data; the anatomical structure constraints combine historical data trends and anatomical limitations.

4. The method according to claim 1, characterized in that, After determining the boundary region distribution characteristics by combining the prostate three-dimensional pose data and the prostate three-dimensional image, a risk assessment value is calculated and mapped to the corresponding risk level. If the risk level exceeds a preset threshold, a path correction instruction is generated to obtain the surgical instrument navigation path, including: Based on the three-dimensional pose data of the prostate, the relative position of the surgical instruments and the prostate is calculated by coordinate transformation to obtain the relative position deviation value; The relative positional deviation values ​​are combined with the three-dimensional image of the prostate, and a boundary segmentation algorithm is used to divide the abnormal hyperplasia area and normal tissue to determine the distribution characteristics of the boundary area. The distribution characteristics of the boundary area include the three-dimensional spatial coordinate range of the abnormal hyperplasia area, the minimum distance to key anatomical structures, the irregularity parameter of the boundary contour, and the volume ratio of the abnormal area in the prostate. Based on the distribution characteristics of the boundary region and the three-dimensional position offset, the risk assessment value is calculated by weighted summation of quantified features and mapped to the corresponding risk level. If the risk level exceeds a preset threshold, a path correction instruction is generated to obtain the corrected device path data; The new position offset is calculated using the corrected instrument path data, and the risk assessment results for the boundary area are updated. The updated risk assessment results are input into the path optimization model to generate the final surgical instrument navigation path; the path optimization model is constructed based on a reinforcement learning algorithm.

5. The method according to claim 4, characterized in that, The risk assessment value is calculated using the following formula: in, This represents the risk assessment value. This represents the minimum distance between the abnormally proliferating area and key anatomical structures. This indicates the preset safe distance threshold. Indicates the irregularity of the boundary contour. , Indicates the length of the outline. Indicates the measurement scale. Indicates the percentage of volume in the abnormal region. , Represents the three-dimensional volume of the abnormally proliferating region. This represents the total three-dimensional volume of the prostate gland. This represents the ratio of the 3D position offset to a preset safety offset threshold. , This represents the real-time three-dimensional position vector of the prostate gland. This represents the preset reference three-dimensional position vector of the prostate. This represents the preset three-dimensional position offset safety threshold vector. This indicates the preset three-dimensional position offset safety threshold. , , , Indicates the feature weights.

6. The method according to claim 1, characterized in that, The process involves constructing a clinical recovery model based on the surgical instrument navigation path and prostate anatomy to generate simulated postoperative recovery indicators. These simulated recovery indicators are then adjusted and validated to meet preset recovery indicator requirements, resulting in the final surgical execution plan. This includes: Obtain surgical operation sequence data updated based on the surgical instrument navigation path; the surgical operation sequence data includes path adjustment parameters and instrument movement timing information; A clinical recovery model was constructed based on the surgical procedure sequence data and the anatomical characteristics of the prostate, and postoperative recovery index simulation data was generated; the recovery index simulation data included wound healing period and urinary function recovery score. The simulated data of the recovery indicators are compared with the preset graded recovery thresholds, and the preliminary assessment results, including the achievement items, deviation items and risk level, are output. Key influencing factors are extracted from the preliminary assessment results to determine the optimization direction of the final surgical execution plan; the key influencing factors include path deviation magnitude, instrument operation accuracy, and residual prediction of hyperplasia area; Based on the optimization direction, the detailed parameters in the surgical operation sequence are adjusted to obtain updated surgical execution plan data; the detailed parameters include instrument movement speed and positioning accuracy threshold. Verify the simulation results of the recovery indicators corresponding to the updated surgical execution plan data. If the simulation results of the recovery indicators meet the preset recovery indicator requirements, then output the final surgical execution plan.

7. An AI image-guided device for minimally invasive surgery for benign prostatic hyperplasia, characterized in that, The device includes: The image acquisition and enhancement module is used to acquire three-dimensional images of the prostate and real-time image sequences during surgery. After preliminary filtering of the real-time image sequences, the module is combined with the three-dimensional images of the prostate to obtain an optimized enhanced image. The data registration and correction module is used to obtain a three-dimensional offset vector based on the enhanced image and the preoperative three-dimensional prostate image by combining the three-dimensional anatomical structure benchmark, and to generate optimized three-dimensional prostate pose data by constructing correction parameters in combination with prostate anatomical structure constraints. The risk assessment and correction module is used to combine the three-dimensional pose data of the prostate with the three-dimensional image of the prostate to determine the distribution characteristics of the boundary region, calculate the risk assessment value and map the corresponding risk level. If the risk level exceeds the preset threshold, a path correction instruction is generated to obtain the surgical instrument navigation path. The surgical plan optimization module is used to construct a clinical recovery model based on the surgical instrument navigation path and the prostate anatomical structure, generate postoperative recovery indicator simulation data, adjust and verify the recovery indicator simulation data to meet the preset recovery indicator requirements, and obtain the final surgical execution plan.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images