Fusion targeted prostate puncture model, matched medical instrument and use method

Abstract

The invention relates to a fusion targeted prostate puncture model, a matched medical instrument and a use method, and aims to solve the technical problems that the actual position occupied by the prostate of a patient cannot be simulated before puncture, and a doctor cannot simulate puncture before an operation, the fusion targeted prostate puncture model comprises a prostate simulation main body, the middle of the prostate simulation main body is hollowed, and the top of the prostate simulation main body is a sealing cover; the bottom is a prostate model nest; a groove is formed in the rear part of the prostate model nest and is used for taking out the focus simulator and the storage liquid pool; the prostate model nest is used for placing a prostate model for simulating the focus of a patient; the prostate model is manufactured by simulating and calculating according to a nidus examination result of a patient; according to the technical scheme, the sealing cover is tightly attached to the prostate model after the prostate model is put in, the prostate model is pressed tightly, the effect of simulating the pubis union structure of a patient is achieved, the problem is well solved, and the prostate model simulation device can be used for prostate treatment.

Description

Technical Field

[0001] The present invention relates to the field of prostate medical devices, and particularly relates to a fusion targeted prostate biopsy model, supporting medical devices and a usage method thereof. Background Art

[0002] The incidence of prostate cancer is increasing and has become an important type of cancer endangering male health. Prostate cancer has become the most common cancer in the urinary system. Prostate biopsy has always been the gold standard for the diagnosis of prostate cancer. How to improve the accuracy of prostate biopsy is a problem worthy of attention. Fusion biopsy has important value in prostate biopsy. This technique combines multiple imaging techniques, such as MRI and TRUS, to improve the accuracy and efficiency of biopsy. Fusion biopsy can help doctors more accurately locate suspicious areas in the prostate, thus avoiding damage to normal tissues and increasing the chance of obtaining tissue samples of lesions. There are various types of fusion biopsy, including MRI-TRUS image fusion biopsy, AI fusion imaging-guided biopsy, cognitive fusion-guided biopsy, etc. These techniques all utilize the advantages of different imaging techniques to achieve more accurate positioning and biopsy. Different types of fusion biopsy have their own advantages and disadvantages. Taking MRI-TRUS image fusion biopsy as an example, its advantage lies in combining the high resolution of MRI and the real-time nature of TRUS, and can more accurately locate suspicious areas in the prostate. However, this technique may be affected by factors such as the maturity of the real-time ultrasound and MRI image fusion algorithm, the establishment of relevant standards, and the proficiency of operation. AI fusion imaging-guided biopsy uses artificial intelligence technology for image analysis and processing, improving the accuracy and efficiency of biopsy. This technique is applicable to a variety of patients, including those with a high suspicion of prostate cancer by clinical and MRI examinations but negative by ultrasound examination, and those who need to reduce the number of biopsy needles or avoid complications. However, the application of AI technology may involve issues such as data privacy and algorithm accuracy, which need to be noted during use. Both of these fusion biopsies require certain equipment support and there are also deviations in actual use. Cognitive fusion biopsy conforms to the habits of most clinicians and has clinical promotion value. Some studies have shown that there is no statistical difference in tumor detection rate between targeted biopsies using cognitive fusion targeted biopsy (COG-TB) or software fusion targeted biopsy (FUS-TB). However, most experts still have difficulty agreeing to directly start cognitive fusion biopsy only by doctors reading MRI images.

[0003] The existing technology cannot realize a fusion targeted prostate biopsy model, supporting medical devices and a usage method that can simulate the actual position of the prostate mass in the patient before biopsy, allowing doctors to perform simulated biopsy before surgery and improve the accuracy of the surgery. The present invention solves the above-mentioned technical problems. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that the existing technology cannot simulate the actual position of the patient's prostate before puncture, allow doctors to simulate puncture before surgery, and improve the accuracy of surgery. Provided is a new fusion targeted prostate puncture model, which can simulate the actual position of the patient's prostate before puncture, allow doctors to simulate puncture before surgery, and may improve the accuracy of surgery. It can also meet the needs of medical student training and doctor-patient communication, which also require similar simulated patients and real situations.

[0005] In order to solve the above technical problems, the technical solutions adopted are as follows:

[0006] A fusion targeted prostate puncture model, the model includes a prostate simulation body, the prostate simulation body is made of elastic material, its shape matches the shape and size of the real prostate, the middle of the prostate simulation body is hollowed out, the top is a sealing cover, and the bottom is a prostate model nest; the rear of the prostate model nest is provided with a groove, the groove is used to take out the lesion simulation body and the storage liquid pool;

[0007] The prostate model nest is used to place a prostate model simulating the patient's lesion;

[0008] The prostate model is produced by simulation and calculation based on the patient's lesion examination results, including: obtaining the patient's prostate examination image, processing and analyzing the prostate examination image, outputting a three-dimensional model of the prostate according to the processing and analysis results, and printing and producing a physical model based on the three-dimensional model;

[0009] After the prostate model is placed in, the sealing cover is tightly attached to the prostate model to press it tightly and play a role in simulating the pubic symphysis structure of the patient.

[0010] The working principle of the present invention is as follows: First, a variety of prostate examination images of patients are collected, such as MRI, CT, PET-CT, etc. These images of different modalities are proofread and fused through specific algorithms, and the advantage information of each image is accurately aligned and integrated to generate a fused image. Then, based on the fused image, a three-dimensional reconstruction algorithm is used to output the three-dimensional model data of the prostate and suspected space-occupying lesions, and then a puncture model is made based on this. The model can accurately simulate the anatomical structure and lesion conditions of the prostate, and the material properties are adapted to ultrasound, MRI and other imaging. When in use, the puncture needle can be guided by ultrasound and MRI imaging equipment to insert into the predetermined lesion position through the puncture needle guiding device of the model, or the operator can simulate the puncture by touching the model and reading the patient's images. At the same time, the supporting data recording and analysis system will record the puncture parameters in real time, and provide a reference for the real operation after analysis and evaluation, thereby realizing the effective simulation and planning of prostate puncture.

[0011] In the above solution, for optimization, further, the simulation calculation and production of the prostate model include:

[0012] Step 1, including obtaining the prostate examination images of the patient: The prostate examination images include MRI images, CT images, and PET-CT images;

[0013] Step 2, processing and analyzing the prostate examination images to complete image calibration and fusion:

[0014] Step 3, using a two-dimensional image segmentation method to roughly segment the fused image, and determining the normal prostate tissue and suspected lesion tissue; at the same time, using a three-dimensional reconstruction algorithm to process the normal prostate tissue and suspected lesion tissue to obtain a three-dimensional model;

[0015] Step 4, using a three-dimensional space segmentation method to perform three-dimensional space segmentation on the suspected lesion tissue, segmenting out the lesion tissue, and thus obtaining a precise three-dimensional prostate model;

[0016] Step 5, based on the three-dimensional model of the prostate and lesion segmentation completed in Step 4, performing the printing and production of the physical model.

[0017] In the preferred solution: First, obtain prostate examination images containing multiple modalities such as MRI, CT, and PET-CT. Then, perform processing and analysis to accurately align and integrate different image information through image calibration and fusion. Subsequently, use two-dimensional image segmentation to roughly segment the fused image, initially distinguish the normal prostate tissue and suspected lesion tissue, and at the same time use a three-dimensional reconstruction algorithm to convert these tissues into a three-dimensional model. Then, use a three-dimensional space segmentation method to further subdivide the suspected lesion tissue and accurately segment out the lesion tissue, thereby obtaining a precise three-dimensional prostate model. Finally, perform the printing and production of the physical model based on this precise model. The acquisition of multi-modal images provides more comprehensive and rich prostate information, improving the diagnostic accuracy. Image calibration and fusion can integrate the advantages of each image and reduce the limitations of a single image. The combination of two-dimensional rough segmentation and three-dimensional reconstruction can quickly outline the general structure and form a three-dimensional presentation, facilitating an overall understanding. The three-dimensional space segmentation realizes the precise positioning and separation of the lesion, providing a reliable basis for subsequent treatment planning, etc. The finally printed physical model can be used for preoperative simulation, medical student training, doctor-patient communication, etc., which is intuitive and practical, and effectively assists in the diagnosis and treatment of prostate-related diseases.

[0018] Further, Step 2 includes:

[0019] Step 1, for MRI images, CT images, and PET-CT images, respectively use specific image feature extraction algorithms for processing, and extract the key information that can effectively characterize the prostate structure and potential lesion features in each image, including edge features, texture features, and gray-scale features;

[0020] Step 2, select a reference image and define the images to be matched: Among the MRI image, CT image, and PET-CT image, randomly select one of the images as the reference image T1, and define the remaining two images as the images to be matched T2 and T3 respectively;

[0021] Step 3, precisely copy the reference image T1 to obtain the second reference image T11. This copying operation needs to ensure that T11 is exactly the same as T1 in all characteristic attributes of the image, including but not limited to pixel values, resolution, image size, etc., so as to provide the same reference template for subsequent parallel calibration operations with different images to be matched;

[0022] Step 4, perform an image matching operation to complete the calibration of the MRI image, CT image, and PET-CT image, ensuring that the three images are accurately aligned in spatial position, that is, the coordinate positions of the same anatomical structure correspond in different images: At the same time, complete the image matching between the reference image T1 and the image to be matched T2, and between the second reference image T11 and the image to be matched T3;

[0023] Step 5, divide each image into a three-dimensional voxel grid form. Each voxel represents a small three-dimensional spatial unit and has corresponding attribute values (such as gray values, etc.), which are respectively denoted as voxel data V T1 、V T2 、V T3 ;

[0024] Step 6, perform feature analysis on the calibrated T1, T2, and T3 images respectively, and extract key feature information related to the target area; for each image, calculate its feature descriptor, including calculating the gray histogram as the density feature descriptor, extracting the texture feature descriptor using the gray-level co-occurrence matrix, and obtaining the morphological feature descriptor by analyzing the shape boundary, etc.;

[0025] Step 7, initially determine the weights, including initially determining the weight of each image according to the performance of different images in specific features and their advantages in reflecting target area information;

[0026] Let the initial weights of T1, T2, and T3 be w T1 、w T2 、w T3 , and w T1 +w T2 +w T3 =1

[0027] If the T1 image performs excellently in showing the overall morphology of the prostate, a relatively high weight can be assigned to T1 in the feature dimension of reflecting the overall morphology; if the T2 image has more advantages in detecting subtle texture changes of lesions, a relatively high weight can be assigned to T2 in the dimension of reflecting texture features; similarly, determine the weight of the T3 image in the corresponding feature dimension according to its advantages;

[0028] Step 8, weight optimization, including extracting sample data with accurate annotation results that are known, where the sample data is case data that has been pathologically verified before, and performing fusion processing on these sample data according to the methods of proofreading and initial weight determination to obtain the fused result, including;

[0029] Step 8.1, calculate the mean squared error MSE. For each voxel position (x, y, z) in each sample data, assume that the voxel value of the target region with accurate annotation is V True (x, y, z), and the fused voxel value is V Fusion (x, y, z). The calculation formula for the mean squared error MSE is:

[0030]

[0031] where N is the total number of voxels in the sample data. By summing the squared differences of all voxel positions and taking the average, the mean squared error is obtained; it measures the average squared error between the fused result and the true value. The smaller the MSE value, the better the fusion effect;

[0032] Step 8.2, calculate the structural similarity index SSIM. The structural similarity index SSIM is used to measure the similarity degree of the fused image and the true image with accurate annotation in terms of structural information;

[0033] For each voxel position (x, y, z), calculate the following several parameters:

[0034] Calculate the mean μ Fusion : Assume that the fused voxel value is V Fusion (x, y, z), and the calculation formula is:

[0035]

[0036] Calculate the standard deviation σ Fusion : Assume that the fused voxel value is V Fusion (x, y, z), and the calculation formula is:

[0037]

[0038] For the voxel value V True (x, y, z) of the target region with accurate annotation, calculate its mean μTrue and standard deviation

[0039] Calculate the covariance C Fusion , True is as follows:

[0040]

[0041] Calculate the structural similarity index SSIM as follows:

[0042]

[0043] where C 1 and C 2 are constants set to avoid a zero denominator. Usually, C 1 =(0.01×L) 2 , C 2 =(0.03×L) 2 , L is the dynamic range of voxel values (for example, for an 8-bit grayscale image, L = 255). The closer the SSIM value is to 1, the better the fusion effect;

[0044] Step 8.3, use the bee colony optimization algorithm to adjust and optimize the weights, complete the weight optimization, and obtain the optimized weights as w′ T1 , w′ T2 , w′ T3 :

[0045] Step 8.3.1, initialize the bee colony: Set the size and number of the bee colony. Each bee represents a set of possible weight values w T1 , w T2 , w T3 . Randomly initialize the position and speed of each M bee; Let the size of the bee colony be M, and the position vector of the i-th bee be The speed vector is where i = 1, 2,..., M;

[0046] Step 8.3.2, define the fitness function:

[0047] F(X i )=α×SSIM + β×MSE -1

[0048] where α and β are weight coefficients used to adjust the relative importance of the structural similarity index SSIM and the mean square error MSE in the fitness function;

[0049] Step 8.3.3, Iteratively update the positions and velocities of the bees: Calculate the individual best positions pbest and the global best position gbest for each bee; the individual best position refers to the best position reached by each bee during the iteration, i.e., the position with the optimal fitness function value; the global best position refers to the best position reached by the entire bee swarm during the iteration;

[0050] Update the velocities and positions of the bees according to the following formulas. The velocity update formula is:

[0051] V i (t + 1) = w × V i (t) + c 1 × r 1 × (pbest i - X i (t)) + c 2 × r 2 × (gbest - X i (t))

[0052] where t represents the iteration number, w is the inertia weight used to balance the historical velocity and the current search direction of the bees, and generally takes values between 0.5 and 0.9; c 1 and c 2 are learning factors, usually taking values of c 1 = c 2 = 2; r 1 and r 2 are numbers randomly generated between 0 and 1;

[0053] The position update formula is: X i (t + 1) = X i (t) + V i (t + 1)

[0054] Repeat until the preset iteration number is reached or the stopping condition is satisfied, and complete the iterative update optimization; the stopping condition includes that the fitness function value reaches a certain threshold; the optimized weights are w′ T1 、w′ T2 、w′ T3 ;

[0055] Step 9, Complete the fusion calculation, including traversing voxels and weighted average calculation:

[0056] Among them, traversing voxels includes:

[0057] For the coordinates of each voxel position (x, y, z) in the three-dimensional space, simultaneously traverse the corresponding voxels in the voxel data V T1 、V T2 、V T3 ;

[0058] The weighted average calculation includes: performing a weighted average calculation on the voxel values of the three images at each voxel position according to the optimized weights;

[0059] Let V T1 (x, y, z), V T2 (x, y, z), V T3 (x, y, z) be the voxel values of images T1, T2, and T3 at the coordinate position (x, y, z), respectively;

[0060] The voxel value V Fusion (x, y, z) of the fused image is calculated as follows:

[0061] V Fusion (x,y,z) = w′ T1 V T1 (x,y,z) + w′ T2 V T2 (x,y,z) + w′ T3 V T3 (x,y,z)

[0062] Step 10, generating the fused image: By performing the above weighted average calculation on all voxel positions, the complete fused voxel data V Fusion is obtained, and then it is converted into the representation of the fused image to complete the fusion operation of the three images;

[0063] Step 11, evaluating and adjusting the fusion result: Using the method of subjective evaluation or objective evaluation, the fusion result is adjusted according to the evaluation result to obtain the image fusion result.

[0064] In the preferred embodiment, multi-modal information integration: By calibrating and fusing MRI, CT, and PET-CT images, the advantages of different imaging modalities can be combined. For example, MRI has high resolution for soft tissues, CT clearly shows bone structures, and PET-CT can reflect metabolic conditions. After fusion, it can comprehensively present various characteristics of the prostate and its lesions, avoiding the one-sidedness of a single image and laying a foundation for subsequent accurate analysis. Improving diagnostic accuracy: The calibration algorithm ensures accurate alignment of different images in spatial position, enabling the fused image to accurately reflect the true state of the prostate. Doctors can thus more accurately judge the location, size of the lesion, and its relationship with surrounding tissues, which helps in formulating a more reasonable diagnosis and treatment plan. Rapid preliminary positioning: The two-dimensional image segmentation method can roughly segment the fused image, quickly determine the normal tissue of the prostate and the suspected lesion tissue, efficiently screen out the regions of interest in the preprocessing, and reduce the data volume and complexity of subsequent analysis. Three-dimensional visualization: Combining with a three-dimensional reconstruction algorithm, the roughly segmented result is transformed into a three-dimensional model, achieving a leap from two dimensions to three dimensions. This enables doctors to visually and stereoscopically observe the morphological structure of the prostate and suspected lesions, better understand their spatial relationships, and assist in operations such as surgical planning. Precise lesion segmentation: Specifically performing three-dimensional space segmentation on the suspected lesion tissue, fully considering the three-dimensional space characteristics, can more finely separate the lesion tissue, obtain an accurate three-dimensional model of the prostate, provide accurate target area positioning for subsequent treatments such as targeted puncture and radiotherapy, and improve the pertinence and effectiveness of treatment.

[0065] Furthermore, the prostate model is pre-made of a material that can be penetrated by a puncture needle and can take specimens. The materials include TPU, TUE, clay, and flour. During production, a predetermined prostate tumor simulator is embedded; or it is made according to the real imaging examination results of the prostate tumor of the patient who is about to undergo puncture.

[0066] Furthermore, the sealing cover is reinforced by a sealing cover fixing rope.

[0067] The present invention also provides a supporting medical device, including the aforementioned fusion targeted prostate puncture model. The fusion targeted prostate puncture model is used to simulate the anatomical structure of the patient's prostate and the actual position of the occupying lesion therein, and this model has the following characteristics:

[0068] It is made from a three-dimensional model output after calibrating, fusing, and processing and analyzing the patient's MRI images, CT images, and PET-CT images, and can accurately present the morphology, size, position of the prostate and its lesions in three-dimensional space, as well as their relationship with surrounding tissues;

[0069] The model material has no interference with ultrasonic imaging and MRI imaging, and can provide ultrasonic propagation characteristics similar to human tissues under ultrasonic examination. Under MRI scanning, it can be clearly identified and positioned through an internal special marker;

[0070] There is a puncture needle guiding device, which is made of transparent material and can accurately guide the puncture needle to insert into the predetermined position to simulate prostate puncture operation, and can provide the resistance and feedback of simulating real puncture operation when the puncture needle is inserted;

[0071] The supporting medical device also includes a data recording and analysis system used in conjunction with the fusion-targeted prostate puncture model. This system can record parameters such as the position, depth, and angle of the puncture needle each time when simulating puncture operation on the model, and analyze and process the recorded data to generate a puncture path map and a lesion localization map, which are used to assist medical staff in preoperative planning and operation evaluation of prostate puncture surgery.

[0072] The present invention also provides a usage method of the fusion-targeted prostate puncture model. Based on the fusion-targeted prostate puncture model described above, the method includes:

[0073] Determine the size and position of the prostate and suspicious tumors through imaging examinations such as MRI, CT, and PETCT. Reconstruct the three-dimensional structure diagram of the prostate and the occupied area with imaging images of 1 mm per layer. Make a prostate model by 3D printing or artificial stacking method and put it into the prostate fossa of the model main body. Inject normal saline into the prostate fossa to discharge air bubbles to facilitate ultrasonic imaging.

[0074] Furthermore, the usage method can perform various puncture simulation operations, including puncturing the prostate through the tunnel under the model main body to simulate transrectal prostate biopsy, or through the perineum simulation area between the tunnel under the model main body and the simulated testis to simulate perineal prostate biopsy. It can also perform cognitive fusion puncture simulation operation by palpating the prostate model in the rectum with fingers and reconstructing in the brain in combination with imaging images without the assistance of imaging.

[0075] Furthermore, in the usage method, the operator can guide the puncture needle to penetrate into the predetermined prostate area through imaging methods such as color Doppler ultrasound, image fusion color Doppler ultrasound, nuclear magnetic resonance, and CT, or can also use the prostate puncture positioning plate to complete prostate puncture. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] The present invention will be further described below with reference to the drawings and embodiments.

[0077] Figure 1 , Schematic diagram of the fusion-targeted prostate puncture system in Embodiment 1.

[0078] Figure 2 , Schematic diagram of the fusion puncture in Embodiment 1.

[0079] Figure 3 , Schematic diagram of the prostate model in Embodiment 1. Figure 4, Schematic diagram of biopsy tissue disc puncture in Example 1. Detailed implementation manners

[0080] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0081] Example 1

[0082] This embodiment provides a fusion targeted prostate puncture model, such as Figure 1 and Figure 2 , the model includes a prostate simulation main body, the prostate simulation main body is made of an elastic material, its shape matches the shape and size of a real prostate, the middle of the prostate simulation main body is hollowed out, the top is a sealing cover, and the bottom is a prostate model socket; a groove is provided at the rear of the prostate model socket, and the groove is used to take out the lesion simulation body and store the liquid pool;

[0083] The prostate model socket is used to place a prostate model simulating the lesion of a patient;

[0084] The prostate model is simulated and calculated and made according to the lesion examination results of a patient, including: obtaining the prostate examination images of the patient, processing and analyzing the prostate examination images, outputting a three-dimensional model of the prostate according to the processing and analysis results, and printing and making a physical model based on the three-dimensional model;

[0085] After the prostate model is placed, the sealing cover closely adheres to the prostate model, presses it tightly and plays the role of simulating the pubic symphysis structure of the patient.

[0086] In this example, first, a variety of prostate examination images of the patient are collected, such as MRI, CT, PET-CT, etc. These different modality images are calibrated and fused through a specific algorithm, accurately aligned and the advantageous information of each image is integrated to generate a fused image. Then, based on the fused image, a three-dimensional model data of the prostate and suspicious occupying lesions is output by using a three-dimensional reconstruction algorithm, and a puncture model is made accordingly. The model can accurately simulate the prostate anatomical structure and lesion conditions, and the material characteristics are adapted to imaging such as ultrasound and MRI. When in use, the puncture needle can be guided by an ultrasound or MRI imaging device to insert into a predetermined lesion position through the puncture needle guiding device of the model, or the puncture can be simulated by the operator's touch perception of the model and the study of the patient's images. At the same time, the supporting data recording and analysis system will record the puncture parameters in real time, and provide a reference for the real operation through analysis and evaluation, so as to realize the effective simulation and planning of prostate puncture.

[0087] Among them, transrectal ultrasound is not necessary, and it can also be positioned by the doctor's hand touch.

[0088] Preferably, the simulation calculation and production of the prostate model include:

[0089] Step 1, including obtaining the prostate examination images of the patient: The prostate examination images include MRI images, CT images, and PET-CT images;

[0090] Step 2, processing and analyzing the prostate examination images to complete image calibration and fusion:

[0091] Step 3, using a two-dimensional image segmentation method to roughly segment the fused image, and determining the normal prostate tissue and suspected lesion tissue; at the same time, using a three-dimensional reconstruction algorithm to process the normal prostate tissue and suspected lesion tissue to obtain a three-dimensional model;

[0092] Step 4, using a three-dimensional space segmentation method to perform three-dimensional space segmentation on the suspected lesion tissue, segmenting out the lesion tissue, and further obtaining a precise three-dimensional prostate model;

[0093] Step 5, based on the three-dimensional model in which the prostate and the lesion have been segmented in Step 4, performing the printing and production of the physical model.

[0094] In the preferred solution: First, obtain prostate examination images containing multiple modalities such as MRI, CT, and PET-CT. Then, perform processing and analysis to accurately align and integrate different image information through image calibration and fusion. Subsequently, use two-dimensional image segmentation to roughly segment the fused image, initially distinguish the normal prostate tissue from the suspected lesion tissue, and at the same time use a three-dimensional reconstruction algorithm to convert these tissues into a three-dimensional model. Then, use a three-dimensional space segmentation method to further subdivide the suspected lesion tissue and accurately segment out the lesion tissue, thereby obtaining a precise three-dimensional prostate model. Finally, perform the printing and production of the physical model based on this precise model. The acquisition of multi-modal images provides more comprehensive and rich prostate information, improving the diagnostic accuracy. Image calibration and fusion can integrate the advantages of each image and reduce the limitations of a single image. The combination of two-dimensional rough segmentation and three-dimensional reconstruction can quickly outline the general structure and form a three-dimensional presentation, facilitating an overall understanding. The three-dimensional space segmentation realizes the precise positioning and separation of the lesion, providing a reliable basis for subsequent treatment planning, etc. The finally printed physical model can be used for preoperative simulation, medical student training, doctor-patient communication, etc., which is intuitive and practical, and effectively assists in the diagnosis and treatment of prostate-related diseases.

[0095] Preferably, Step 2 includes:

[0096] Step 1, for MRI images, CT images, and PET-CT images, respectively use specific image feature extraction algorithms for processing, and extract the key information that can effectively characterize the prostate structure and potential lesion features in each image, including edge features, texture features, and gray-scale features;

[0097] Step 2, select the reference image and define the images to be matched: Among the MRI image, CT image, and PET-CT image, randomly select one of the images as the reference image T1, and define the remaining two images as the images to be matched T2 and the image to be matched T3 respectively;

[0098] Step 3, precisely copy the reference image T1 to obtain the second reference image T11. This copying operation needs to ensure that T11 is exactly the same as T1 in all characteristic attributes of the image, including but not limited to pixel values, resolution, image size, etc., so as to provide the same reference template for subsequent parallel proofreading operations with different images to be matched;

[0099] Step 4, perform the image matching operation to complete the proofreading of the MRI image, CT image, and PET-CT image, ensuring that the three images are accurately aligned in spatial position, that is, the coordinate positions of the same anatomical structure correspond in different images: At the same time, complete the image matching between the reference image T1 and the image to be matched T2, and between the second reference image T11 and the image to be matched T3;

[0100] Step 5, divide each image into a three-dimensional voxel grid form. Each voxel represents a small three-dimensional spatial unit and has corresponding attribute values (such as gray values, etc.), which are respectively denoted as voxel data V T1 、V T2 、V T3 ;

[0101] Step 6, perform feature analysis on the proofread T1, T2, and T3 images respectively, and extract the key feature information related to the target area; for each image, calculate its feature descriptor, including calculating the gray histogram as the density feature descriptor, extracting the texture feature descriptor using the gray-level co-occurrence matrix, and obtaining the morphological feature descriptor by analyzing the shape boundary, etc.;

[0102] Step 7, initially determine the weights, including initially determining the weight of each image according to the performance of different images in specific features and their advantages in reflecting the information of the target area;

[0103] Let the initial weights of T1, T2, and T3 be w T1 、w T2 、w T3 , and w T1 +w T2 +w T3 =1

[0104] If the T1 image performs well in showing the overall morphology of the prostate, then a relatively high weight can be assigned to T1 in the feature dimension of reflecting the overall morphology; if the T2 image has more advantages in detecting subtle texture changes of lesions, then a relatively high weight can be assigned to T2 in the dimension of reflecting texture features; similarly, determine the weight of the T3 image in the corresponding feature dimension according to its advantages.

[0105] Step 8, weight optimization, including extracting sample data with accurate annotation results that are known. The sample data is case data that has been pathologically verified before. These sample data are processed by fusing them according to the methods of proofreading and initial weight determination to obtain the fused result, including:

[0106] Step 8.1, calculate the mean square error MSE. For each voxel position (x, y, z) in each sample data, let the voxel value of the target area with known accurate annotation be V True (x, y, z), and the fused voxel value be V Fusion (x, y, z). The calculation formula for the mean square error MSE is:

[0107]

[0108] where N is the total number of voxels in the sample data. By summing the squares of the differences at all voxel positions and taking the average, the mean square error is obtained. It measures the average squared error between the fused result and the true value. The smaller the MSE value, the better the fusion effect.

[0109] Step 8.2, calculate the structural similarity index SSIM. The structural similarity index SSIM is used to measure the similarity degree of the fused image and the true image with known accurate annotation in terms of structural information.

[0110] For each voxel position (x, y, z), calculate the following several parameters:

[0111] Calculate the mean μ Fusion : Let the fused voxel value be V Fusion (x, y, z), and the calculation formula is:

[0112]

[0113] Calculate the standard deviation σ Fusion : Let the fused voxel value be V Fusion (x, y, z), and the calculation formula is:

[0114]

[0115] For the voxel value V True (x, y, z) of the target area with known accurate annotation, calculate its mean μTrue and the standard deviation σ True ;

[0116] Calculate the covariance C Fusion，True as:

[0117]

[0118] Calculate the structural similarity index SSIM as:

[0119]

[0120] where C 1 and C 2 are constants set to avoid a zero denominator, and usually C 1 =(0.01×L) 2 , C 2 =(0.03×L) 2 , L is the dynamic range of voxel values (for example, for an 8-bit grayscale image, L = 255), and the closer the SSIM value is to 1, the better the fusion effect;

[0121] Step 8.3, use the bee colony optimization algorithm to adjust and optimize the weights, complete the weight optimization, and obtain the optimized weights as w′ T1 , w′ T2 , w′ T3 :

[0122] Step 8.3.1, initialize the bee colony: Set the scale and number of the bee colony, and each bee represents a set of possible weight values w T1 , w T2 , w T3 , randomly initialize the position and speed of each M bee; Let the scale of the bee colony be M, and the position vector of the i-th bee be The speed vector is where

[0123] i = 1, 2,..., M;

[0124] Step 8.3.2, define the fitness function:

[0125] F(X i ) = α×SSIM + β×MSE - 1

[0126] where α and β are weight coefficients used to adjust the relative importance of the structural similarity index SSIM and the mean square error MSE in the fitness function;

[0127] Step 8.3.3, iteratively update the positions and velocities of the bees: Calculate the individual best position pbest and the global best position gbest for each bee; the individual best position refers to the best position reached by each bee during the iteration, that is, the position with the optimal fitness function value; the global best position refers to the best position reached by the entire bee swarm during the iteration;

[0128] Update the velocities and positions of the bees according to the following formulas. The velocity update formula is:

[0129] V i (t + 1) = w × V i (t) + c 1 × r 1 × (pbest i - X i (t)) + c 2 × r 2 × (gbest - X i (t))

[0130] where t represents the number of iterations, w is the inertia weight used to balance the historical velocity and the current search direction of the bees, and generally takes values between 0.5 and 0.9; c 1 and c 2 are learning factors, usually taking the value c 1 = c 2 = 2; r 1 and r 2 are numbers randomly generated between 0 and 1;

[0131] The position update formula is: X i (t + 1) = X i (t) + V i (t + 1)

[0132] Repeat until the preset number of iterations is reached or the stopping condition is satisfied to complete the iterative update optimization; the stopping condition includes that the fitness function value reaches a certain threshold; the optimized weights are w′ T1 、w′ T2 、w′ T3 ;

[0133] Step 9, complete the fusion calculation, including traversing voxels and weighted average calculation:

[0134] Among them, traversing voxels includes:

[0135] For the coordinates of each voxel position (x, y, z) in the three-dimensional space, simultaneously traverse the corresponding voxels in the voxel data V T1 、V T2 、V T3 ;

[0136] The weighted average calculation includes: performing a weighted average calculation on the voxel values of the three images at each voxel position according to the optimized weights;

[0137] Let V T1 (x, y, z), V T2 (x, y, z), V T3 (x, y, z) be the voxel values of images T1, T2, and T3 at the coordinate position (x, y, z), respectively;

[0138] The voxel value V Fusion (x, y, z) of the fused image is calculated as follows:

[0139] V Fusion (x,y,z) = w′ T1 V T1 (x,y,z) + w′ T2 V T2 (x,y,z) + w′ T3 V T3 (x,y,z)

[0140] Step 10, generating a fused image: By performing the above-mentioned weighted average calculation on all voxel positions, the complete fused voxel data V Fusion is obtained, and then it is converted into a representation of the fused image to complete the fusion operation of the three images;

[0141] Step 11, evaluating and adjusting the fusion result: Using subjective evaluation or objective evaluation methods, the fusion result is adjusted according to the evaluation result to obtain the image fusion result.

[0142] In the preferred solution, multi-modal information integration: By calibrating and fusing MRI, CT, and PET-CT images, the advantages of different imaging modalities can be combined. For example, MRI has high resolution for soft tissues, CT clearly shows bone structures, and PET-CT can reflect metabolic conditions. After fusion, it can comprehensively present various characteristics of the prostate and its lesions, avoiding the one-sidedness of a single image and laying a foundation for subsequent accurate analysis. Improving diagnostic accuracy: The calibration algorithm ensures that different images are accurately aligned in spatial position, enabling the fused image to accurately reflect the true state of the prostate. Doctors can thus more accurately judge the location, size of the lesion, and its relationship with surrounding tissues, which helps in formulating a more reasonable diagnosis and treatment plan. Rapid preliminary localization: The two-dimensional image segmentation method can roughly segment the fused image, quickly determine the normal prostate tissue and suspected lesion tissue, efficiently screen out the areas of concern in the preprocessing, and reduce the data volume and complexity of subsequent analysis. Three-dimensional visualization: Combining with the three-dimensional reconstruction algorithm, the roughly segmented result is transformed into a three-dimensional model, achieving a leap from two-dimensional to three-dimensional, enabling doctors to visually and stereoscopically observe the morphological structure of the prostate and suspected lesions, better understand their spatial relationships, and assist in operations such as surgical planning. Precise lesion segmentation: Specifically performing three-dimensional space segmentation on the suspected lesion tissue, fully considering the three-dimensional space characteristics, can more precisely separate the lesion tissue, obtain an accurate three-dimensional model of the prostate, provide accurate target area localization for subsequent treatments such as targeted puncture and radiotherapy, and improve the pertinence and effectiveness of treatment.

[0143] Specifically, the prostate model is pre-made of a material that can be penetrated by a puncture needle and can take specimens. The materials include thermoplastic polyurethane elastomer rubber TPU, textured polyurethane elastomer TUE, clay, and flour. When making it, a predetermined prostate tumor simulator is embedded; or it is made according to the real imaging examination results of the prostate tumor of the patient who is about to undergo puncture.

[0144] Preferably, the sealing cover is reinforced by a sealing cover fixing rope. It can further reinforce the sealing cover.

[0145] Such as Figure 3 , when making the prostate model, the suspicious occupancy site can be filled with different color materials for construction.

[0146] This embodiment also provides a supporting medical device, including the aforementioned fusion targeted prostate puncture model. The fusion targeted prostate puncture model is used to simulate the anatomical structure of the patient's prostate and the actual position of the occupancy lesion therein, and this model has the following characteristics:

[0147] It is made from a three-dimensional model output after calibrating, fusing, and processing and analyzing the patient's MRI images, CT images, and PET-CT images, and can accurately present the morphology, size, position of the prostate and its lesions in three-dimensional space, as well as their relationship with surrounding tissues;

[0148] The model material does not interfere with ultrasonic imaging and MRI imaging, and can provide ultrasonic propagation characteristics similar to those of human tissues under ultrasonic examination. It can be clearly identified and located by the built-in special markers under MRI scanning;

[0149] It is provided with a puncture needle guiding device, which is made of transparent material and can accurately guide the puncture needle to insert into the predetermined position to simulate the prostate puncture operation, and can provide the resistance and feedback of the simulated real puncture operation when the puncture needle is inserted;

[0150] The supporting medical device also includes a data recording and analysis system used in conjunction with the fusion-targeted prostate puncture model. This system can record parameters such as the position, depth, and angle of the puncture needle each time when the simulated puncture operation is performed on the model, and analyze and process the recorded data to generate a puncture path map and a lesion localization map, which are used to assist medical staff in preoperative planning and operation evaluation of prostate puncture surgery.

[0151] This embodiment also provides a method for using a fusion-targeted prostate puncture model. Based on the fusion-targeted prostate puncture model described above, the method includes:

[0152] Determine the size and position of the prostate and suspicious tumors through imaging examinations such as MRI, CT, and PETCT. Reconstruct the three-dimensional structure diagram of the prostate and the occupied area with imaging images of 1 mm per layer. Use 3D printing or artificial stacking method to make a prostate model and place it in the prostate fossa of the model main body. Inject normal saline into the prostate fossa to discharge air bubbles to facilitate ultrasonic imaging.

[0153] Specifically, as Figure 2 , the method can perform various puncture simulation operations, including puncturing the prostate through the tunnel under the model main body to simulate transrectal prostate biopsy, or through the perineum simulation area between the tunnel under the model main body and the simulated testis to simulate perineal prostate biopsy. It can also perform cognitive fusion puncture simulation operations by palpating the prostate model in the rectum with fingers and reconstructing in the brain in combination with imaging images without the assistance of imaging.

[0154] Specifically, in the method, the operator can guide the puncture needle to penetrate into the predetermined prostate area through imaging methods such as color Doppler ultrasound, image fusion color Doppler ultrasound, nuclear magnetic resonance, and CT, or can also use a prostate puncture positioning plate to complete prostate puncture.

[0155] Such as Figure 4, the biopsy tissue tray includes a conventional tray with 12 needles for conventional puncture and a 4-needle tray for enhanced puncture of the suspicious area shown in imaging. This not only ensures balanced puncture but also guarantees the puncture density in high-risk areas. The specimen cells of the specimen tray are rectangular, facilitating the placement of tissue strips. The double-tray combination is particularly suitable for fusion puncture. When used as a medical device, a sealed lid is added.

[0156] Although the above illustrative specific embodiments of the present invention have been described to enable those skilled in the art to understand the present invention, the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, all inventions made using the concept of the present invention are within the scope of protection.

Claims

1. A fusion-targeted prostate biopsy model, characterized in that: The model includes a prostate simulation body, which is made of elastic material and has an appearance that matches the shape and size of a real prostate. The middle of the prostate simulation body is hollowed out, the top is a sealing cover, and the bottom is a prostate model nest; a groove is provided at the rear of the prostate model nest, and the groove is used to take out the lesion simulation body and the storage liquid pool; The prostate model nest is used to place a prostate model simulating the patient's lesion; The prostate model is produced by simulation and calculation based on the patient's lesion examination results, including: obtaining the patient's prostate examination image, processing and analyzing the prostate examination image, outputting a three-dimensional model of the prostate according to the processing and analysis results, and printing and producing a physical model based on the three-dimensional model; After the prostate model is placed in, the sealing cover is tightly attached to the prostate model to press it tightly and play a role in simulating the pubic symphysis structure of the patient.

2. The fusion targeted prostate puncture model according to claim 1, characterized in that: The simulation calculation and production of the prostate model includes: Step 1, including obtaining a prostate examination image of the patient: the prostate examination image includes an MRI image, a CT image, and a PETCT image; Step 2: Process and analyze the prostate examination image to complete image correction and melting: Step 3: Use a two-dimensional image segmentation method to roughly segment the image after image fusion to determine normal prostate tissue and suspected lesion tissue; and use a three-dimensional reconstruction algorithm to process the normal prostate tissue and suspected lesion tissue to obtain a three-dimensional model; Step 4: Use a three-dimensional space segmentation method to perform three-dimensional space segmentation on the suspected lesion tissue, segment the lesion tissue, and then obtain an accurate three-dimensional model of the prostate; Step 5: Based on the three-dimensional model of the prostate and lesion segmentation completed in step 4, the solid model is printed.

3. The fusion targeted prostate puncture model according to claim 1, characterized in that: Step 2 includes: Step 1: MRI images, CT images, and PET-CT images are processed using specific image feature extraction algorithms to extract key information from each image that can effectively characterize the prostate structure and potential lesion characteristics, including edge features, texture features, and grayscale features; Step 2, select the reference image and define the image to be matched: randomly select one of the MRI images, CT images, and PET-CT images as the reference image T1, and define the remaining two images as the image to be matched T2 and the image to be matched T3; Step 3: Accurately copy the reference image T1 to obtain the second reference image T11. This copying operation must ensure that T11 and T1 are completely consistent in all image feature attributes, including but not limited to pixel value, resolution, image size, etc., so as to provide the same reference template for subsequent parallel proofreading operations with different images to be matched; Step 4, perform image matching operation to complete the calibration of MRI images, CT images, and PET-CT images, and ensure that the three images are accurately aligned in spatial position, that is, the coordinate positions of the same anatomical structure in different images correspond to each other: at the same time, complete image matching of the reference image T1 and the image to be matched T2, and the second reference image T11 and the image to be matched T3; Step 5: Divide each image into a three-dimensional voxel grid. Each voxel represents a small three-dimensional space unit and has a corresponding attribute value (such as grayscale value, etc.), which is recorded as voxel data V T1 、V T2 、V T3 ; Step 6, feature analysis is performed on the corrected T1, T2, and T3 images respectively to extract key feature information related to the target area; for each image, its feature descriptor is calculated, including calculating the grayscale histogram as the density feature descriptor, using the grayscale co-occurrence matrix to extract the texture feature descriptor, and obtaining the morphological feature descriptor by analyzing the shape boundary, etc.; Step 7: Initially determine the weights. Set the initial weights of T1, T2, and T3 to be w T1 、w T2 、w T3 , and w T1 +w T2 +w T3 =1 Step 8, weight optimization, includes extracting known sample data with accurate labeling results, fusing these sample data according to the method of proofreading and initial weight determination, and obtaining the fused results, including; Step 8.1, calculate the mean square error MSE. For each voxel position (x, y, z) in each sample data, let the voxel value of the target area with known accurate annotation be V True (x, y, z), the fused voxel value is V Fusion (x, y, z), the calculation formula for calculating the mean square error MSE is: Where N is the total number of voxels in the sample data. The mean square error is obtained by summing and averaging the squared differences of all voxel positions. It measures the average square error between the fusion result and the true value. The smaller the MSE value, the better the fusion effect. Step 8.2, calculate the structural similarity index SSIM, which is used to measure the similarity between the fused image and the real image with known accurate annotations in terms of structural information; For each voxel position (x, y, z), the following parameters are calculated: Calculate the mean μ Fusion :Let the fused voxel value be V Fusion (x, y, z), the calculation formula is: Calculate the standard deviation σ Fusion :Let the fused voxel value be V Fusion (x, y, z), the calculation formula is: For the voxel value V of the target area with known accurate annotation True (x, y, z), calculate its mean μ True and standard deviation σ True ; Calculate the covariance C Fusion,True for: The structural similarity index SSIM is calculated as: Among them, C1 and C2 are constants set to avoid the denominator being zero, and L is the dynamic range of the voxel value; Step 8.3, use the bee swarm optimization algorithm to adjust and optimize the weights, complete the weight optimization, and obtain the optimized weight w′ T1 , w′ T2 , w′ T3 : Step 8.3.1, initialize the bee colony: set the size and number of the bee colony, each bee represents a set of possible weight values ​​w T1 、w T2 、w T3 , randomly initialize the position and velocity of each M bee; let the size of the bee colony be M, and the position vector of the i-th bee be The velocity vector is Where i = 1, 2, ..., M. Step 8.3.2, define the fitness function: F(X i )=α×SSIM+β×MSE -1 Among them, α and β are weight coefficients, which are used to adjust the relative importance of the structural similarity index SSIM and the mean square error MSE in the fitness function; Step 8.3.3, iteratively update the bee position and speed: calculate the individual optimal position pbest and the global optimal position gbest of each bee; the individual optimal position refers to the best position reached by each bee during the iteration process, that is, the position with the best fitness function value; the global optimal position refers to the best position reached by the entire bee colony during the iteration process; Update the bee's speed and position according to the following formula. The speed update formula is: V i (t+1)=w×V i (t)+c1×r1×(pbest i -X i (t))+c2×r2×(gbest-X i (t)) Where t represents the number of iterations, w is the inertia weight used to balance the bee's historical speed and current search direction; c1 and c2 are learning factors; r1 and r2 are randomly generated numbers between 0 and 1; The position update formula is: X i (t+1)=X i (t)+V i (t+1) Repeat until the preset number of iterations is reached or the stopping condition is met to complete the iterative update optimization; the stopping condition includes the fitness function value reaching a certain threshold; the optimized weight is w′ T1 , w′ T2 , w′ T3 ; Step 9, complete the fusion calculation, including traversing voxels and weighted average calculation: Among them, traversing voxels includes: For each voxel position (x, y, z) in three-dimensional space, traverse the voxel data V at the same time T1 、V T2 、V T3 The corresponding voxel in The weighted average calculation includes: performing weighted average calculation on the voxel values ​​of the three images at each voxel position according to the optimized weights; Let V T1 (x, y, z), V T2 (x, y, z), V T3 (x, y, z) are the voxel values ​​of images T1, T2, and T3 at the coordinate position (x, y, z), respectively; The fused voxel value V Fusion The calculation formula for (x, y, z) is as follows: V Fusion (x,y,z)=w′ T1 V T1 (x,y,z)+w′ T2 V T2 (x,y,z)+w′ T3 V T3 (x,y,z) Step 10: Generate fused image: By performing the above weighted average calculation on all voxel positions, the complete fused voxel data V is obtained. Fusion , and then convert it into a fused image representation to complete the fusion operation of the three images; Step 11, fusion result evaluation and adjustment: Use subjective evaluation or objective evaluation method to adjust the fusion result according to the evaluation result to obtain the image fusion result.

4. The fusion-targeted prostate biopsy model according to any one of claims 1 to 3, characterized in that: The sealing cover is reinforced by a sealing cover fixing rope.

5. A matching medical device, characterized in that: The fusion-targeted prostate biopsy model comprises any one of claims 1 to 4, wherein the fusion-targeted prostate biopsy model is used to simulate the anatomical structure of the patient's prostate and the actual location of the space-occupying lesions therein, and the model has the following characteristics: The three-dimensional model is produced by collating, fusing, processing and analyzing the patient's MRI images, CT images, and PET-CT images. It can accurately present the shape, size, position of the prostate and its lesions in three-dimensional space, as well as the relationship with surrounding tissues. The model material has no interference with ultrasound imaging and MRI imaging, and can provide ultrasound propagation characteristics similar to those of human tissue under ultrasound examination. It can be clearly identified and located through built-in special markers under MRI scanning. A puncture needle guiding device is provided. The guiding device is made of transparent material and can accurately guide the puncture needle to be inserted into a predetermined position to simulate a prostate puncture operation. When the puncture needle is inserted, it can provide resistance and feedback simulating a real puncture operation; The supporting medical device also includes a data recording and analysis system used in conjunction with the fusion targeted prostate puncture model. The system can record parameters such as the position, depth and angle of the puncture needle each time a simulated puncture operation is performed on the model, and analyze and process the recorded data to generate a puncture path map and a lesion location map to assist medical staff in preoperative planning and operation evaluation of prostate puncture surgery.

6. A method for using a fusion targeted prostate puncture model, characterized in that: The method is based on the fusion-targeted prostate puncture model described in any one of claims 1 to 4, and the method comprises: The size and location of the prostate and suspected tumors are determined through imaging examinations such as MRI, CT, and PET-CT. The three-dimensional structure of the prostate and the space-occupying tumor is reconstructed with imaging images at a layer of 1 mm. A prostate model is made by 3D printing or artificial stacking and placed in the prostate fossa of the model body. Physiological saline is injected into the prostate fossa to expel bubbles to facilitate ultrasound imaging.

7. The method for using the fusion targeted prostate biopsy model according to claim 6, characterized in that: The method of use can perform a variety of puncture simulation operations, including puncturing the prostate from the tunnel below the model body to simulate transrectal prostate puncture biopsy, or puncturing the perineum simulation area between the tunnel below the model body and the simulated testicle to simulate transperineal prostate puncture biopsy. The method can also perform cognitive fusion puncture simulation operations by palpating the prostate model in the rectum with fingers without the assistance of imaging and reconstructing it in the brain in combination with imaging images.

8. The method for using the fusion targeted prostate puncture model according to claim 6, characterized in that: In the method of use, the operator can guide the puncture needle into the predetermined prostate area through imaging methods such as color ultrasound, image fusion color ultrasound, nuclear magnetic resonance, CT, etc., and can also use a prostate puncture positioning plate to complete prostate puncture.

Images