Seminal vesicle endoscope identification and positioning system based on vision enhancement technology

Abstract

The invention discloses a seminal vesicle endoscope identification and positioning system based on a visual enhancement technology, and relates to the technical field of visual detection. A seminal vesicle endoscope is placed into a prostate small vesicle cavity through the urethra, and white light, fluorescent light and near-infrared images are synchronously collected through a three-channel CMOS sensor; extracting a fluorescence characteristic pattern, a scattering gradient characteristic pattern and a white light edge characteristic pattern, and fusing through an RGB thermodynamic diagram synthesis function; optimizing the learning rate of the convolutional neural network by using a genetic algorithm, training a weak region separation model to output a real-time weak region mask, screening a target spot in combination with a centroid, and marking the target spot; in the puncture stage, a tactile feedback needling instrument is integrated, resistance is monitored in real time, threshold response is set, drainage is verified through flushing, and the operation effect is confirmed through a double-black-hole wall breaking channel under a white light microscope; the multi-mode imaging fusion bottleneck is broken through, the weak area in the operation is accurately positioned, the puncture complication risk is reduced, the success rate of a single operation is improved, and a full-chain intelligent positioning navigation scheme is provided for minimally invasive treatment of seminal vesicle diseases.

Description

Technical Field

[0001] The present invention belongs to the field of visual detection technology, and in particular relates to a seminal vesiculoscope identification and positioning system based on vision enhancement technology. Background Art

[0002] The demand for minimally invasive surgery for seminal vesicle disorders (such as hematospermia, inflammation, stones, and outlet obstruction) is increasing year by year. "Windowing" through the prostate's cystic cavity to access the seminal vesicle is a key surgical principle and method for treating these conditions. The weakest area connecting the prostate's cystic cavity to the seminal vesicle provides the most critical and safest path for windowing / breaking the wall to enter the seminal vesicle.

[0003] Existing endoscopic systems mainly use white light imaging, and near-infrared and fluorescence imaging are not fully utilized. Since the prostate cyst cavity is adjacent to the rectum and neurovascular bundle, the incidence of surgical complications (such as rectal fistula and seminal vesicle injury) is as high as 5% to 15%, and intelligent intraoperative navigation and safety assurance technologies are still needed.

[0004] In traditional seminal vesiculoscopic surgery, the structure of the prostate cyst cavity is complex and varies greatly from person to person. It is difficult to clearly distinguish between solid tissue and weak areas (such as potential perforation risk areas) by relying solely on white light endoscopy, which can easily lead to misjudgment during surgery. Although existing equipment can switch between white light / fluorescence / near-infrared modes, the data are displayed independently and rely on the surgeon's subjective integration. There is a lack of automated fusion analysis of multi-spectral features, and it is impossible to generate an intuitive weak area positioning map in real time. The positioning of weak areas relies on experience, and the puncture process lacks real-time tissue mechanical feedback and intelligent early warning mechanisms, which can easily cause rectal injury or seminal vesicle penetration. After surgery, it relies on manual observation of drainage effects and structural changes, and lacks immediate verification methods during surgery, which increases the risk of secondary surgery.

[0005] Existing technologies have problems such as difficulty in identifying anatomical structures, insufficient utilization of multimodal data, low puncture navigation accuracy, and delayed surgical verification. Summary of the Invention

[0006] (1) Technical problems solved

[0007] In response to the problems in the related art, the present invention provides a seminal vesiculoscope identification and positioning method based on vision enhancement technology to overcome the above-mentioned technical problems existing in the existing related art.

[0008] (2) Technical solution

[0009] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0010] S1. Place the seminal vesiculoscope into the prostate cyst cavity through the urethra to collect real-time prostate cyst cavity data;

[0011] S2. Extracting features of real-time prostate cystic cavity image data to obtain a solid tissue fluorescence feature map, a scattering gradient feature map, and a white light edge feature map;

[0012] S3, fusing the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map to generate a real-time RGB thermal map of the prostate cyst;

[0013] S4. Use historical prostate cyst image data combined with optimization algorithms to train and optimize the convolutional neural network to obtain a weak area separation neural network;

[0014] S5, inputting the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map into the weak area separation neural network to obtain a real-time weak area mask;

[0015] The target coordinates are found according to the real-time weak area mask and marked in the real-time prostate cystic cavity RGB thermal map to obtain the real-time prostate cystic cavity RGB thermal map of the marked target;

[0016] S6. Guided puncture is performed based on the RGB thermal image of the prostate cystic cavity where the target is marked;

[0017] S7: After successful puncture, perform seminal vesicle drainage and verify the surgical results;

[0018] The present invention solves the core pain points of difficult identification of weak areas, high puncture risk, and delayed effect verification in seminal vesiculoscopic surgery through a full-chain process of multispectral feature fusion, intelligent target positioning, tactile feedback puncture, and instant verification during surgery. It significantly improves surgical safety and operation accuracy, significantly shortens operation time, and provides a positioning and navigation method for minimally invasive treatment of seminal vesicle diseases.

[0019] Preferably, the S1 comprises the following steps:

[0020] S11, inserting the seminal vesiculoscope into the prostate cyst cavity through the urethra, starting the white light endoscopy mode to observe the anatomical structure, and obtaining white light image data;

[0021] S12: The operator switches to the dual-spectrum excitation mode with a foot control, emitting blue light to stimulate tissue autofluorescence and obtain fluorescence image data;

[0022] Emitting near-infrared light to penetrate the tissue layer and obtain near-infrared image data;

[0023] S13, synchronously capturing white light image data, fluorescence image data, and near-infrared image data through a three-channel CMOS sensor to obtain real-time prostate cyst image data;

[0024] The present invention switches between dual-spectrum modes through foot control. Blue light excites tissue fluorescence to reveal dark spots in weak areas, and near-infrared light penetrates tissue to reveal dark areas in cavities. A three-channel CMOS sensor is used to synchronously capture white light, fluorescence, and near-infrared images, achieving real-time imaging of the anatomy-defect-cavity trinity. This eliminates the image misalignment of time-sharing acquisition and ensures the spatial consistency of multimodal data. Foot-controlled operation saves switching time and improves surgical efficiency. It provides spatiotemporally aligned three-source data for subsequent feature fusion, laying the foundation for precise navigation.

[0025] Preferably, said S2 comprises the following steps:

[0026] S21. Extracting solid tissue fluorescence characteristic values ​​of the fluorescence image data in the real-time prostate cyst image data using a fluorescence intensity distribution function to obtain a solid tissue fluorescence characteristic map; the characteristic map is essentially a two-dimensional matrix formed by arranging the characteristic values ​​of each pixel position according to the original spatial coordinates;

[0027] S22, calculating the scattering gradient amplitude distribution data of the near-infrared image data in the real-time prostate cystic cavity image data according to the scattering gradient value calculation formula;

[0028] Calculating the scattering gradient characteristic value of the near-infrared image data based on the scattering gradient amplitude distribution data in combination with the weak area characteristic mapping formula to obtain a scattering gradient characteristic map;

[0029] S23, using a gradient operator to calculate the edge of the white light image in the real-time prostate cyst image data to obtain a white light edge feature map;

[0030] The present invention quantifies tissue defects through the fluorescence intensity distribution function, identifies low-scattering areas of cavities through scattering gradient amplitude calculation, and strengthens anatomical edges through the white light gradient operator, generating three-modal feature maps respectively. It accurately separates tissue characteristics: the fluorescence map marks weak areas, the scattering map locates cavities, and the white light map preserves structural details. It constructs a multi-dimensional feature space to provide complementary data for fusion thermal maps. It eliminates subjective misjudgment and replaces visual evaluation with mathematical quantification to improve the objectivity of positioning.

[0031] Preferably, the step S3 includes the following steps:

[0032] S31, inputting the solid tissue fluorescence characteristic value distribution, the scattering gradient amplitude distribution data, and the white light edge characteristic distribution into the heat map RGB channel synthesis function to obtain a real-time prostate cyst RGB heat map;

[0033] The present invention automatically fuses fluorescence features, scattering gradient features, and white light edges through an RGB heat map synthesis function to generate a real-time navigation heat map. The weak areas are highlighted in bright yellow patches, improving visual recognition efficiency. Multi-source data is integrated with one click, replacing manual subjective puzzles and improving the accuracy of real-time navigation during surgery.

[0034] Preferably, said S4 comprises the following steps:

[0035] S41. Collect historical solid tissue fluorescence characteristic maps, scattering gradient characteristic maps, white light edge characteristic maps, and weak area masks to obtain historical prostate small cyst image data;

[0036] S42. Construct a convolutional neural network, set a learning rate of the convolutional neural network; set a training accuracy and a training accuracy threshold of the convolutional neural network;

[0037] The convolutional neural network was trained using historical prostate cyst image data. During the training process, a genetic algorithm was used to find the learning rate of the convolutional neural network. The convergence of the convolutional neural network was determined based on the training accuracy and training accuracy threshold to obtain the optimal solution.

[0038] Using the optimal solution as the learning rate of the convolutional neural network to obtain a weak area separation neural network;

[0039] The present invention collects multimodal historical feature maps and mask data, constructs a convolutional neural network, and uses an optimized algorithm for the convolutional neural network to quickly lock the optimal learning rate; outputs a high-precision weak area separation network, which improves the recognition accuracy of the weak area mask; avoids the blindness of manual parameter adjustment, reduces the number of iterations, and provides a lightweight model for real-time segmentation during surgery.

[0040] In S42, finding the learning rate of the convolutional neural network by a genetic algorithm and judging the convergence of the convolutional neural network according to the training accuracy and the training accuracy threshold are performed to obtain the optimal solution, which includes the following steps:

[0041] S421, constructing a chromosome population, setting the size of the chromosome population, using each chromosome in the chromosome population as the learning rate of each different convolutional neural network; setting the maximum number of iterations;

[0042] S422. Setting a fitness function according to the training accuracy and training accuracy threshold of the convolutional neural network;

[0043] S423, start iteration, perform selection, crossover and mutation operations on the chromosome population in each iteration process, and obtain the chromosome population after the operation;

[0044] S424, repeat S423, and when the maximum number of iterations is reached, stop the iteration and obtain the optimal solution;

[0045] The present invention dynamically optimizes the learning rate through a genetic algorithm, encodes the learning rate through chromosomes, and quantifies the training convergence efficiency through a fitness function. Through selection-crossover-mutation iteration, the global optimal solution is locked and the local optimum is avoided. This improves model accuracy, accelerates convergence, and provides a highly robust network for real-time segmentation during surgery.

[0046] Preferably, the S5 comprises the following steps:

[0047] S51, inputting the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map into a weak area separation neural network to obtain a real-time weak area mask;

[0048] S52, binarizing the real-time weak area mask to obtain a binary weak area mask; performing connected region analysis on the binary weak area mask to obtain a connected region set;

[0049] The centroid calculation formula is used to calculate the centroid coordinates of each connected area in the connected area set to obtain a centroid coordinate set;

[0050] The centroid coordinates that meet the axial distribution of the C line are selected from the centroid coordinate set as the target point;

[0051] The target point is superimposed on the real-time RGB thermal map of the prostate cyst cavity to obtain the real-time RGB thermal map of the prostate cyst cavity with the marked target point;

[0052] The present invention uses a weak area separation network to output a mask in real time, locates the target through binarization, connected area analysis, and C-line axial distribution; calculates the target accurately to the pixel level through centroid calculation, and combines C-line axial screening to avoid rectal risk areas; dynamic laser cross marking is superimposed on the RGB heat map, reducing the intraoperative target visualization response delay; thus reducing target positioning error and improving puncture safety.

[0053] Preferably, the S6 steps include:

[0054] S61: The operator inserts the tactile feedback puncture needle along the marked target in the real-time RGB thermal image of the prostate cyst;

[0055] S62, setting a resistance threshold; using the tactile feedback of the needle tip micro-pressure sensor on the puncture needle to collect tissue resistance in real time and obtain a real-time resistance value;

[0056] When the real-time resistance value is within the resistance threshold, a green light will be displayed, indicating normal penetration of the weak area until the puncture is successful;

[0057] When the real-time resistance value exceeds the resistance threshold, the device will automatically stop advancing and sound a high-frequency vibration alarm. The cause of the abnormality will be determined and repaired until the puncture is successful.

[0058] The present invention uses a tactile feedback puncture needle to monitor tissue resistance in real time, sets a threshold value for intelligent response, and immediately prevents rectal damage, thereby reducing the risk of puncture complications and achieving active safety protection.

[0059] Preferably, the S7 includes the following steps:

[0060] S71. After successful puncture, a honeycomb-like seminal vesicle cavity structure can be observed inside the hole. Normal saline is injected through the puncture needle channel to flush and observe the effect of seminal vesicle fluid drainage.

[0061] S72. The operator switches the seminal vesiculoscope to white light mode to confirm whether the characteristic double black hole structure appears in the field of view, and obtains the confirmation result of the characteristic double black hole structure;

[0062] S73. The surgical result is obtained based on the seminal vesicle fluid drainage effect and the characteristic double black hole structure confirmation results.

[0063] The present invention verifies drainage patency through flushing with physiological saline and confirms the double black hole wall-breaking channel under a white light microscope, allowing immediate judgment during the operation whether the operation is successful; the double black hole structure is the key to successful puncture, improving the success rate of a single operation and shortening the recovery period; a closed loop of intraoperative verification is established to eliminate the risk of blind operation.

[0064] A seminal vesiculoscope recognition and positioning system based on visual enhancement technology, used to implement the above-mentioned seminal vesiculoscope recognition and positioning method based on visual enhancement technology, comprising a multimodal image acquisition module, a multidimensional feature extraction module, an RGB heat map generation module, a neural network optimization training module, a target positioning and dynamic marking module, an intelligent puncture feedback control module, and a surgical result verification module;

[0065] The multimodal image acquisition module uses a seminal vesiculoscope to enter the prostate cyst cavity through the urethra, observes the anatomical structure in white light endoscopy mode to obtain white light images, switches to dual-spectrum excitation mode, and separately organizes the fluorescence image of the body itself, penetrates the tissue layer to obtain near-infrared images, and uses an integrated three-channel CMOS sensor to synchronously capture the above white light, fluorescence, and near-infrared image data to form complete real-time prostate cyst cavity image data;

[0066] The multi-dimensional feature extraction module processes the fluorescence image data using the fluorescence intensity distribution function based on the real-time prostate cyst image data to obtain a solid tissue fluorescence feature map reflecting the distribution of solid tissue; calculates the scattering gradient amplitude of the near-infrared image and combines it with the weak area feature mapping formula to generate a scattering gradient feature map reflecting the difference in tissue scattering characteristics; and applies the gradient operator to process the white light image data to extract a white light edge feature map that enhances anatomical contour information;

[0067] The RGB heat map generation module fuses the extracted solid tissue fluorescence feature map, scattering gradient feature map and white light edge feature map to obtain a real-time RGB heat map of the prostate cyst cavity;

[0068] The neural network optimization training module trains a convolutional neural network using historical prostate cyst image data composed of historical solid tissue fluorescence feature maps, scattering gradient feature maps, white light edge feature maps, and weak area masks. A genetic algorithm is used to dynamically optimize the network's learning rate parameters. By constructing a chromosome population, defining a fitness function based on training accuracy, and performing iterative operations such as selection, crossover, and mutation, the optimal learning rate is ultimately found, thereby training a high-precision weak area separation neural network.

[0069] The target location and dynamic marking module uses a weak area separation neural network, combined with solid tissue fluorescence feature maps, scattering gradient feature maps, and white light edge feature maps, to obtain a real-time weak area mask. The real-time weak area mask is binarized and connected area analyzed, and the center of mass coordinates are located. The center of mass coordinates that meet the conditions are selected as target coordinates. The target coordinates are marked on the real-time prostate cyst RGB thermal map, providing intuitive visual guidance for the puncture operation.

[0070] The intelligent puncture feedback control module guides the operator to perform the puncture operation according to the heat map of the marked target; it performs real-time pressure detection during the puncture process and responds to it until the puncture is successful;

[0071] After a successful puncture, the surgical result verification module evaluates the drainage effect by observing the drainage of the seminal vesicle fluid; confirms whether the characteristic "double black hole" wall-breaking channel structure appears; obtains the surgical verification confirmation result, and makes an overall judgment on the success of the operation.

[0072] (3) Beneficial effects

[0073] The present invention has the following beneficial effects:

[0074] The present invention solves the core pain points of difficult identification of weak areas, high puncture risk, and delayed effect verification in seminal vesiculoscopic surgery through a full-chain process of multispectral feature fusion, intelligent target positioning, tactile feedback puncture, and instant verification during surgery. It significantly improves surgical safety and operation accuracy, significantly shortens operation time, and provides a positioning and navigation method for minimally invasive treatment of seminal vesicle diseases.

[0075] The present invention breaks through the bottleneck of multimodal imaging fusion and realizes accurate visualization of weak areas. It uses a three-channel CMOS sensor to synchronously capture white light, fluorescence and near-infrared images, and combines them with the RGB thermal map synthesis function to convert weak areas into bright yellow patches for real-time highlighting, thereby improving the surgeon's recognition efficiency and solving the problem of anatomical structure misjudgment caused by traditional single imaging modes.

[0076] The present invention constructs an intelligent target positioning system to improve surgical operation accuracy; uses a genetic algorithm to optimize the learning rate of a convolutional neural network, dynamically balances training accuracy and convergence threshold through a fitness function, and outputs a high-precision weak area mask; combines the centroid coordinate screening conditions to automatically mark targets, avoid high-risk anatomical areas, reduce target positioning errors, improve positioning accuracy, and significantly reduce the uncertainty of human operation.

[0077] The present invention integrates tactile feedback and an intelligent early warning mechanism to eliminate puncture complications. The tactile feedback puncture needle is equipped with a micro-pressure sensor to monitor tissue resistance in real time, set a threshold response, and effectively prevent rectal injury. It reduces the risk of puncture-related complications and provides active safety protection for the operator.

[0078] The present invention establishes an intraoperative instant verification closed loop to ensure the one-time success of the operation; after puncture, normal saline is injected through the needle channel to verify the patency of the drainage, and the white light endoscope is synchronously switched to observe the characteristic double black hole wall-breaking channel, so as to achieve instant confirmation of the surgical effect during the operation, avoid secondary surgery, improve the success rate of a single operation, and shorten the recovery period.

[0079] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, they can also obtain drawings based on these drawings without paying any creative work.

[0081] Figure 1 This is a flow chart of a method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to the present invention;

[0082] Figure 2 This is a flow chart of a weak area separation neural network obtained in a seminal vesiculoscope identification and positioning method based on visual enhancement technology of the present invention;

[0083] Figure 3 This is a schematic diagram of the process of obtaining a target in a seminal vesiculoscope identification and positioning method based on visual enhancement technology of the present invention;

[0084] Figure 4 This is a module schematic diagram of a seminal vesiculoscope identification and positioning system based on vision enhancement technology of the present invention. DETAILED DESCRIPTION

[0085] The following will clearly and completely describe the technical solutions in the embodiments of the invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0086] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "top", "middle", "inside" and the like indicating orientation or positional relationship are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the components or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the invention.

[0087] Example 1:

[0088] See also Figure 1 、 Figure 2 、 Figure 3 The present invention discloses a method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology, comprising the following steps:

[0089] S1, inserting the seminal vesiculoscope into the prostate cyst cavity through the urethra to collect real-time prostate cyst cavity image data;

[0090] Said S1 comprises the following steps:

[0091] S11, inserting the seminal vesiculoscope into the prostate cyst cavity through the urethra, starting the white light endoscopy mode to observe the anatomical structure, and obtaining white light image data;

[0092] S12: The operator switches to the dual-spectrum excitation mode using a foot control, emitting 405 nm blue light to stimulate tissue autofluorescence (solid tissue appears green, weak areas appear dark spots), and obtain fluorescence image data;

[0093] Emit 850nm near-infrared light to penetrate the tissue layer (the seminal vesicle cavity appears as a low-scattering dark area) to obtain near-infrared image data;

[0094] S13, synchronously capturing white light image data, fluorescence image data, and near-infrared image data through a three-channel CMOS sensor to obtain real-time prostate cyst image data;

[0095] S2. Extracting features of real-time prostate cystic cavity image data to obtain a solid tissue fluorescence feature map, a scattering gradient feature map, and a white light edge feature map;

[0096] The S2 comprises the following steps:

[0097] S21. Using a fluorescence intensity distribution function to extract solid tissue fluorescence characteristic values ​​from the fluorescence image data in the real-time prostate cyst image data, a solid tissue fluorescence characteristic map is obtained. The characteristic map is essentially a two-dimensional matrix formed by arranging the characteristic values ​​of each pixel position according to the original spatial coordinates. The fluorescence intensity distribution function is as follows:

[0098]

[0099] Where (x, y) represents the coordinates of the pixel in the fluorescence image; F(x, y) represents the solid tissue fluorescence characteristic value of the pixel with coordinates (x, y); k1 represents the slope coefficient of the Sigmoid function; G(x, y) represents the intensity value of the pixel with coordinates (x, y) in the green channel; T fluo Indicates the fluorescence intensity threshold (usually 120 / 255);

[0100] S22. Calculate the scattering gradient amplitude distribution data of the near-infrared image data in the real-time prostate cyst image data according to the scattering gradient value calculation formula; the scattering gradient value calculation formula is as follows:

[0101]

[0102] Where (x, y) represents the coordinates of the pixel in the infrared image, M(x, y) represents the scattering gradient amplitude of the pixel with coordinates (x, y), and I NIR (x,y) represents the grayscale value of the infrared image of the pixel with coordinates (x,y). Respectively represent I NIR Partial derivatives of (x,y) in the x and y directions;

[0103] According to the scattering gradient amplitude distribution data combined with the weak area characteristic mapping formula, the scattering gradient characteristic value of the near-infrared image data is calculated to obtain a scattering gradient characteristic map; the calculation formula is as follows:

[0104]

[0105] Where P(x,y) represents the scattering gradient eigenvalue of the pixel with coordinates (x,y) (low gradient areas correspond to high eigenvalues), and k2 represents the adjustment parameter;

[0106] S23, using a gradient operator to calculate the edge of the white light image in the real-time prostate cyst image data to obtain a white light edge feature map; the gradient operator formula is as follows:

[0107]

[0108] Among them, (x, y) represents the coordinates of the pixel point in the white light image, Indicates the white light edge feature intensity value of the pixel with coordinates (x, y), I NIR (x,y) represents the grayscale value of the white light image of the pixel with coordinates (x,y). Respectively represent I NIR Partial derivatives of (x,y) in the x and y directions;

[0109] S3, fusing the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map to generate a real-time RGB thermal map of the prostate cyst;

[0110] The S3 includes the following steps:

[0111] S31, inputting the solid tissue fluorescence characteristic value distribution, the scattering gradient amplitude distribution data and the white light edge characteristic distribution into the heat map RGB channel synthesis function to obtain a real-time prostate small cyst RGB heat map; the heat map RGB channel synthesis function formula is,

[0112]

[0113] Among them, R(x,y) represents the pixel value of the red channel at the coordinate point (x,y), which is an inverse mapping of the fluorescence intensity (the weaker the fluorescence, the darker the red); G(x,y) represents the pixel value of the green channel at the coordinate point (x,y), which is an inverse mapping of the near-infrared scattering gradient (the lower the scattering, the darker the green); B(x,y) represents the pixel value of the blue channel at the coordinate point (x,y), which is used to enhance the edge of the white light structure; the weak area in the RGB heat map of the prostate cyst cavity is displayed as a bright yellow patch superimposed with dark red and dark green;

[0114] S4. Use historical prostate cyst image data combined with optimization algorithms to train and optimize the convolutional neural network to obtain a weak area separation neural network;

[0115] The S4 comprises the following steps:

[0116] S41. Collect historical solid tissue fluorescence characteristic maps, scattering gradient characteristic maps, white light edge characteristic maps, and weak area masks to obtain historical prostate small cyst image data;

[0117] S42. Construct a convolutional neural network, set a learning rate of the convolutional neural network; set a training accuracy and a training accuracy threshold of the convolutional neural network;

[0118] The convolutional neural network was trained using historical prostate cyst image data. During the training process, a genetic algorithm was used to find the learning rate of the convolutional neural network. The convergence of the convolutional neural network was determined based on the training accuracy and training accuracy threshold to obtain the optimal solution.

[0119] Using the optimal solution as the learning rate of the convolutional neural network to obtain a weak area separation neural network;

[0120] In S42, finding the learning rate of the convolutional neural network by a genetic algorithm and judging the convergence of the convolutional neural network according to the training accuracy and the training accuracy threshold are performed to obtain the optimal solution, which includes the following steps:

[0121] S421, construct a chromosome population, set the size of the chromosome population to m, and then the chromosome population is represented by n = {n1, n2, ..., n i ,...,n m}, n i represents the i-th chromosome in the chromosome population; uses each chromosome in the chromosome population as a different normalization coefficient; and sets a maximum number of iterations;

[0122] S422. A fitness function is set according to the training accuracy and training accuracy threshold of the convolutional neural network. The fitness function is used to calculate the fitness of each chromosome in the chromosome population. The fitness function formula is as follows:

[0123] k=|ε1-ε2| -1 +g;

[0124] Among them, k represents the fitness of each chromosome in the chromosome population, ε1 represents the training accuracy of the convolutional neural network, ε2 represents the training accuracy threshold of the convolutional neural network, and g represents the bias constant;

[0125] S423, start iteration, perform selection, crossover and mutation operations on the chromosome population in each iteration process, and obtain the chromosome population after the operation;

[0126] S424, repeat S423, and when the maximum number of iterations is reached, stop the iteration and obtain the optimal solution;

[0127] S5, inputting the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map into the weak area separation neural network to obtain a real-time weak area mask;

[0128] The target coordinates are found according to the real-time weak area mask and marked in the real-time prostate cystic cavity RGB thermal map to obtain the real-time prostate cystic cavity RGB thermal map of the marked target;

[0129] The S5 comprises the following steps:

[0130] S51, inputting the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map into a weak area separation neural network to obtain a real-time weak area mask;

[0131] S52, binarizing the real-time weak area mask to obtain a binary weak area mask; performing connected region analysis on the binary weak area mask to obtain a connected region set;

[0132] The centroid calculation formula is used to calculate the centroid coordinates of each connected area in the connected area set to obtain a centroid coordinate set; the centroid calculation formula is as follows:

[0133]

[0134] in, represents the centroid coordinates of the i-th connected region in the connected region set, R i represents the set of all pixels in the i-th connected region, |R i | represents the total number of pixels in the i-th connected region, Represents R i The sum of the x coordinates of all pixels in , Represents R i Sum the y coordinates of all pixels in ;

[0135] The centroid coordinates that are consistent with the axial distribution of the C line (1-3 mm from the posterior wall of the prostate capsule) were selected from the centroid coordinate set as the target point;

[0136] The target point is superimposed on the real-time RGB thermal map of the prostate cyst cavity to obtain the real-time RGB thermal map of the prostate cyst cavity with the marked target point;

[0137] S6. Guided puncture is performed based on the real-time RGB thermal image of the prostate cystic cavity with the marked target;

[0138] The following steps are described in S6:

[0139] S61: The operator inserts the tactile feedback puncture needle along the marked target in the real-time RGB thermal image of the prostate cyst;

[0140] S62, setting a resistance threshold; using the tactile feedback of the needle tip micro-pressure sensor on the puncture needle to collect tissue resistance in real time and obtain a real-time resistance value;

[0141] When the real-time resistance value is within the resistance threshold, a green light will be displayed, indicating normal penetration of the weak area until the puncture is successful;

[0142] When the real-time resistance value exceeds the resistance threshold, the device automatically stops advancing and issues a high-frequency vibration alarm. The cause of the abnormality is determined and repaired until the puncture is successful. If the resistance value is less than 8kPa, the weak area is penetrated normally (green light prompt). If the resistance value is greater than 8kPa, the device automatically stops advancing and issues a high-frequency vibration alarm (to prevent rectal damage).

[0143] S7: After successful puncture, perform seminal vesicle drainage and verify the surgical results;

[0144] The S7 comprises the following steps:

[0145] S71. After successful puncture, a honeycomb-like seminal vesicle cavity structure can be observed inside the hole. Normal saline is injected through the puncture needle channel to flush and observe the effect of seminal vesicle fluid drainage. The drained seminal vesicle fluid is thick liquid, yellow or milky white.

[0146] S72. The operator switches the seminal vesiculoscope to white light mode to confirm whether a characteristic double black hole structure (wall-breaking channel) appears in the field of view, and obtains a confirmation result of the characteristic double black hole structure;

[0147] S73. The surgical result is obtained based on the seminal vesicle fluid drainage effect and the characteristic double black hole structure confirmation results.

[0148] Example 2:

[0149] See also Figure 4 A seminal vesiculoscope recognition and positioning system based on visual enhancement technology is used to implement the above-mentioned seminal vesiculoscope recognition and positioning method based on visual enhancement technology, including a multimodal image acquisition module, a multi-dimensional feature extraction module, an RGB heat map generation module, a neural network optimization training module, a target positioning and dynamic marking module, an intelligent puncture feedback control module, and a surgical result verification module;

[0150] The multimodal image acquisition module uses a seminal vesiculoscope to enter the prostate cyst cavity through the urethra, observes the anatomical structure in white light endoscopy mode to obtain white light images, switches to dual-spectrum excitation mode, and separately organizes the fluorescence image of the body itself, penetrates the tissue layer to obtain near-infrared images, and uses an integrated three-channel CMOS sensor to synchronously capture the above white light, fluorescence, and near-infrared image data to form complete real-time prostate cyst cavity image data;

[0151] The multi-dimensional feature extraction module processes the fluorescence image data using the fluorescence intensity distribution function based on the real-time prostate cyst image data to obtain a solid tissue fluorescence feature map reflecting the distribution of solid tissue; calculates the scattering gradient amplitude of the near-infrared image and combines it with the weak area feature mapping formula to generate a scattering gradient feature map reflecting the difference in tissue scattering characteristics; and applies the gradient operator to process the white light image data to extract a white light edge feature map that enhances anatomical contour information;

[0152] The RGB heat map generation module fuses the extracted solid tissue fluorescence feature map, scattering gradient feature map and white light edge feature map to obtain a real-time RGB heat map of the prostate cyst cavity;

[0153] The neural network optimization training module trains a convolutional neural network using historical prostate cyst image data composed of historical solid tissue fluorescence feature maps, scattering gradient feature maps, white light edge feature maps, and weak area masks. A genetic algorithm is used to dynamically optimize the network's learning rate parameters. By constructing a chromosome population, defining a fitness function based on training accuracy, and performing iterative operations such as selection, crossover, and mutation, the optimal learning rate is ultimately found, thereby training a high-precision weak area separation neural network.

[0154] The target location and dynamic marking module uses a weak area separation neural network, combined with solid tissue fluorescence feature maps, scattering gradient feature maps, and white light edge feature maps, to obtain a real-time weak area mask. The real-time weak area mask is binarized and connected area analyzed, and the center of mass coordinates are located. The center of mass coordinates that meet the conditions are selected as target coordinates. The target coordinates are marked on the real-time prostate cyst RGB thermal map, providing intuitive visual guidance for the puncture operation.

[0155] The intelligent puncture feedback control module guides the operator to perform the puncture operation according to the heat map of the marked target; it performs real-time pressure detection during the puncture process and responds to it until the puncture is successful;

[0156] After a successful puncture, the surgical result verification module evaluates the drainage effect by observing the drainage of the seminal vesicle fluid; confirms whether the characteristic "double black hole" wall-breaking channel structure appears; obtains the surgical verification confirmation result, and makes an overall judgment on the success of the operation.

[0157] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0158] The preferred embodiments of the invention disclosed above are intended only to help illustrate the invention. These preferred embodiments do not exhaust all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology, characterized in that: The following steps are involved: S1. Place the seminal vesiculoscope into the prostate cyst cavity through the urethra to collect real-time prostate cyst cavity data; S2. Extracting features of real-time prostate cystic cavity image data to obtain a solid tissue fluorescence feature map, a scattering gradient feature map, and a white light edge feature map; S3, fusing the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map to generate a real-time RGB thermal map of the prostate cyst; S4. Use historical prostate cyst image data combined with optimization algorithms to train and optimize the convolutional neural network to obtain a weak area separation neural network; S5, inputting the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map into the weak area separation neural network to obtain a real-time weak area mask; The target coordinates are found according to the real-time weak area mask and marked in the real-time prostate cystic cavity RGB thermal map to obtain the real-time prostate cystic cavity RGB thermal map of the marked target; S6. Guided puncture is performed based on the RGB thermal image of the prostate cystic cavity where the target is marked; S7. After successful puncture, seminal vesicle drainage is performed and the surgical results are verified.

2. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 1, characterized in that: Said S1 comprises the following steps: S11, inserting the seminal vesiculoscope into the prostate cyst cavity through the urethra, starting the white light endoscopy mode to observe the anatomical structure, and obtaining white light image data; S12: The operator switches to the dual-spectrum excitation mode with a foot control, emitting blue light to stimulate tissue autofluorescence and obtain fluorescence image data; Emitting near-infrared light to penetrate the tissue layer and obtain near-infrared image data; S13. Synchronously capture white light image data, fluorescence image data, and near-infrared image data through a three-channel CMOS sensor to obtain real-time prostate cyst image data.

3. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 1, characterized in that: The S2 comprises the following steps: S21, extracting solid tissue fluorescence characteristic values ​​of fluorescence image data in the real-time prostate cystic cavity image data using a fluorescence intensity distribution function to obtain a solid tissue fluorescence characteristic map; S22, calculating the scattering gradient amplitude distribution data of the near-infrared image data in the real-time prostate cystic cavity image data according to the scattering gradient value calculation formula; Calculating the scattering gradient characteristic value of the near-infrared image data based on the scattering gradient amplitude distribution data in combination with the weak area characteristic mapping formula to obtain a scattering gradient characteristic map; S23. Calculate the edge of the white light image in the real-time prostate cyst image data using a gradient operator to obtain a white light edge feature map.

4. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 1, characterized in that: The S3 includes the following steps: S31. Input the solid tissue fluorescence characteristic value distribution, the scattering gradient amplitude distribution data, and the white light edge characteristic distribution into the heat map RGB channel synthesis function to obtain a real-time prostate cyst RGB heat map.

5. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 4, characterized in that: The S4 comprises the following steps: S41. Collect historical solid tissue fluorescence characteristic maps, scattering gradient characteristic maps, white light edge characteristic maps, and weak area masks to obtain historical prostate small cyst image data; S42. Construct a convolutional neural network, set a learning rate of the convolutional neural network; set a training accuracy and a training accuracy threshold of the convolutional neural network; The convolutional neural network was trained using historical prostate cyst image data. During the training process, a genetic algorithm was used to find the learning rate of the convolutional neural network. The convergence of the convolutional neural network was determined based on the training accuracy and training accuracy threshold to obtain the optimal solution. The optimal solution is used as the learning rate of the convolutional neural network to obtain a weak area separation neural network.

6. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 5, characterized in that: In S42, finding the learning rate of the convolutional neural network by a genetic algorithm and judging the convergence of the convolutional neural network according to the training accuracy and the training accuracy threshold are performed to obtain the optimal solution, which includes the following steps: S421, constructing a chromosome population, setting the size of the chromosome population, using each chromosome in the chromosome population as the learning rate of each different convolutional neural network; setting the maximum number of iterations; S422. Setting a fitness function according to the training accuracy and training accuracy threshold of the convolutional neural network; S423, start iteration, perform selection, crossover and mutation operations on the chromosome population in each iteration process, and obtain the chromosome population after the operation; S424. Repeat S423. When the maximum number of iterations is reached, stop the iteration and obtain the optimal solution.

7. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 1, characterized in that: The following steps are described in S5: S51, inputting the solid tissue fluorescence feature map, the scattering gradient feature map, and the white light edge feature map into a weak area separation neural network to obtain a real-time weak area mask; S52, binarizing the real-time weak area mask to obtain a binary weak area mask; performing connected region analysis on the binary weak area mask to obtain a connected region set; The centroid calculation formula is used to calculate the centroid coordinates of each connected area in the connected area set to obtain a centroid coordinate set; The centroid coordinates that meet the axial distribution of the C line are selected from the centroid coordinate set as the target point; The target point is superimposed on the real-time RGB thermal map of the prostate cyst cavity to obtain the real-time RGB thermal map of the prostate cyst cavity with the marked target point.

8. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 1, characterized in that: The S6 comprises the following steps: S61: The operator inserts the tactile feedback puncture needle along the marked target in the real-time RGB thermal image of the prostate cyst; S62, setting a resistance threshold; using the tactile feedback of the needle tip micro-pressure sensor on the puncture needle to collect tissue resistance in real time and obtain a real-time resistance value; When the real-time resistance value is within the resistance threshold, a green light will be displayed, indicating normal penetration of the weak area until the puncture is successful; When the real-time resistance value exceeds the resistance threshold, the device automatically stops advancing and issues a high-frequency vibration alarm. The cause of the abnormality is determined and repaired until the puncture is successful.

9. The method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology according to claim 1, characterized in that: The S7 comprises the following steps: S71. After successful puncture, a honeycomb-like seminal vesicle cavity structure can be observed inside the hole. Normal saline is injected through the puncture needle channel to flush and observe the effect of seminal vesicle fluid drainage. S72. The operator switches the seminal vesiculoscope to white light mode to confirm whether the characteristic double black hole structure appears in the field of view, and obtains the confirmation result of the characteristic double black hole structure; S73. The surgical result is obtained based on the seminal vesicle fluid drainage effect and the characteristic double black hole structure confirmation results.

10. A seminal vesiculoscope identification and positioning system based on visual enhancement technology, characterized in that: A method for identifying and positioning a seminal vesiculoscope based on visual enhancement technology as described in any one of claims 1 to 9 is implemented, wherein the system includes a multimodal image acquisition module, a multidimensional feature extraction module, an RGB heat map generation module, a neural network optimization training module, a target positioning and dynamic marking module, an intelligent puncture feedback control module, and a surgical result verification module.

Images