System for optimizing puncture angle of biopsy needle

Abstract

The invention relates to a biopsy needle puncture angle optimization system, and relates to the technical field of medical puncture, and the system comprises a physical sign morphology analysis module which is used for collecting the body characteristics of a user, analyzing the morphological change of a puncture part, and obtaining the morphological change rate and amplitude; the image adjustment and analysis module is used for collecting a puncture part image of the user, adjusting a standard part image according to the form change rate and amplitude, and obtaining an adjusted part image sequence and a similarity sequence; the puncture adaptation analysis module is used for randomly generating a puncture angle and analyzing puncture adaptation in combination with the adjustment part image sequence and the similarity sequence to obtain puncture adaptation parameters; and the angle optimization display module is used for optimizing the puncture angle, obtaining the optimal puncture angle and displaying the optimal puncture angle. The problems that in traditional biopsy puncture, body differences of different users are not considered, and the same puncture angle is adopted, so that errors are large, precision is low, and the puncture effect is affected are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical puncture, in particular to a biopsy needle puncture angle optimization system. BACKGROUND

[0002] With the continuous development of medical puncture technology, the precision of biopsy puncture has an increasingly significant impact on the diagnosis effect, and optimizing the puncture angle for different user body differences has become a key technical problem to improve puncture quality. At present, the traditional biopsy puncture angle selection mostly adopts a unified standard, which is difficult to adapt to the body feature differences of different users and the shape changes of the puncture site, and is prone to cause large puncture angle errors and insufficient precision.

[0003] The existing puncture angle determination method does not fully consider the individual body feature differences of users, resulting in poor adaptability of the puncture angle, which not only reduces the success rate of biopsy puncture, but also increases the pain of patients and the complexity and risk of subsequent diagnosis and treatment. SUMMARY

[0004] To solve the above technical problems, the present application provides a biopsy needle puncture angle optimization system, which improves the status quo of large errors, low precision and insufficient adaptability caused by not considering the body differences of different users and adopting a unified puncture angle in traditional biopsy puncture.

[0005] The embodiments of the present application disclose the following technical solutions: The embodiments of the present application provide a biopsy needle puncture angle optimization system, which comprises: A physical feature analysis module is configured to collect the body features of a user, analyze the shape changes of the puncture site, obtain a shape change rate and a shape change amplitude, and analyze the shape changes of the puncture site. An image adjustment analysis module is configured to collect a site image of the puncture site of a user, adjust a standard site image according to the shape change rate and the shape change amplitude, obtain an adjusted site image, perform similarity analysis on the adjusted site image and the site image, and obtain an adjusted site image sequence and a similarity sequence, wherein the number of adjustments is configured according to the shape change rate. A puncture adaptability analysis module is configured to randomly generate a biopsy puncture angle, combine the adjusted site image sequence and the similarity sequence, and perform puncture adaptability analysis to obtain puncture adaptability parameters. An angle optimization display module is configured to perform puncture angle optimization, obtain an optimal puncture angle, and display the optimal puncture angle as an optimization result.

[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages: The application provides a biopsy needle puncture angle optimization system, which realizes accurate optimization of the biopsy needle puncture angle through the following steps: collecting user body features, analyzing morphological changes of a puncture site to obtain a morphological change rate and amplitude, adjusting a standard image in combination with a site image, analyzing similarity to form an adjusted site image sequence and a similarity sequence, randomly generating a puncture angle and combining the sequence to obtain adaptive parameters through adaptive analysis, continuously optimizing the puncture angle until convergence and showing the optimal result. First, user body features are collected, input into a morphological change predictor trained by historical data, and the morphological change rate and amplitude of the puncture site are obtained. Then, the puncture site image is collected, the standard image is adjusted according to the morphological change rate and amplitude, the similarity between the adjusted image and the actual image is analyzed through a similarity identifier, and an adjusted site image sequence and a similarity sequence are formed. Then, a puncture angle is randomly generated, the adjusted site image sequence is input into a puncture adaptive analysis device to obtain a single adaptive parameter sequence, and the puncture adaptive parameter is obtained through weighted calculation of the similarity sequence. Finally, new angles are continuously generated and adaptive parameters are calculated until optimization convergence, and the optimal puncture angle is determined and displayed.

[0007] The technical scheme of the application solves the problems of low puncture precision and high risk caused by individual morphological differences and blind angle selection in traditional biopsy puncture, realizes personalized optimization of the puncture angle, and provides technical support for improving the safety and accuracy of biopsy puncture. BRIEF DESCRIPTION OF DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0009] Figure 1 A structure diagram of a biopsy needle puncture angle optimization system provided by the embodiment of the application is provided. Figure 2 A flowchart of puncture site image adjustment and similarity analysis provided by the embodiment of the application is provided.

[0010] In the drawings, the components represented by the numbers are described as follows: Signs 01, image adjustment analysis module 02, puncture adaptive analysis module 03, angle optimization display module 04. DETAILED DESCRIPTION

[0011] The application provides a biopsy needle puncture angle optimization system, which is used to solve the technical problems of large error, low precision and poor puncture adaptability caused by not considering the differences in physical characteristics of different users and adopting a unified puncture angle, and not dynamically adjusting in combination with the shape change of the puncture site.

[0012] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person skilled in the art without creative work fall within the scope of protection of the application.

[0013] In the description of the application, the terms "first", "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can explicitly or implicitly include one or more of the features. In the description of the application, the meaning of "multiple" is two or more, unless otherwise explicitly and specifically limited.

[0014] In the description of the application, the term "for example" is used to indicate "as an example, illustration or explanation". Any embodiment described as "for example" in the application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the application. In the following description, details are listed for the purpose of explanation. It should be understood that a person skilled in the art can realize the application without using these specific details. In other examples, well-known structures and processes will not be described in detail to avoid unnecessary details making the description of the application obscure. Therefore, the application is not intended to be limited to the shown embodiments, but is consistent with the broadest scope of principles and features disclosed in the application.

[0015] Embodiments, as shown in the accompanying drawings Figure 1 The application provides a biopsy needle puncture angle optimization system, which comprises the following steps: The physical sign shape analysis module 01 is used to collect the physical characteristics of the user, analyze the shape change of the puncture site, and obtain the shape change rate and the shape change amplitude; In the embodiment of the application, in the scene of biopsy puncture angle optimization, in order to accurately obtain the shape change of the puncture site, a shape change predictor is constructed to realize efficient analysis of the shape change rate and the change amplitude of the puncture site.

[0016] Specifically, first, based on the historical user puncture record data, a comprehensive sample body feature set is collected, which covers the body information of users of different ages, body types, health conditions, etc., and also contains the relevant features of typical puncture sites such as the posterior superior iliac spine.

[0017] At the same time, the probability and average amplitude of the morphological change of the puncture site corresponding to each sample body feature are accurately labeled, thereby constructing a complete sample morphological change rate set and a sample morphological change amplitude set.

[0018] Further, the sample body feature set is taken as input data, and the sample morphological change rate set and the sample morphological change amplitude set are taken as supervision signals, which are introduced into a machine learning model for training. By continuously adjusting the model parameters until the model converges, the obtained morphological change predictor has accurate analysis capability for the morphological change rate and the morphological change amplitude of the puncture site.

[0019] Finally, the collected user body features are input into the morphological change predictor, and the predictor can quickly and accurately analyze the morphological change rate and the morphological change amplitude of the user puncture site and output.

[0020] The technical scheme of the model training in this step provides basic data support for subsequent standard site image adjustment and puncture adaptability analysis, and guarantees the accuracy of the puncture angle optimization.

[0021] The sign morphology analysis module 01 in the system provided by the embodiment of the application comprises: Collect the body features of the user; Input the body features into the morphological change predictor, and output the morphological change rate and the morphological change amplitude.

[0022] In the embodiment of the application, in order to realize accurate analysis of the morphological change of the user puncture site, a machine learning model needs to be trained based on historical puncture record data to construct a morphological change predictor, and the correlation between the body features and the morphological change of the puncture site is learned by the model to adapt to the individual difference analysis needs of different users.

[0023] Specifically, first, based on the puncture record data of other users in a historical period, a comprehensive sample body feature set is collected, which covers the user information related to the possible puncture site morphology, such as different ages, genders, heights, weights, body mass indexes (BMI), and previous medical history (such as whether there is a puncture site related disease), and also contains the basic physiological feature data of typical puncture sites such as the posterior superior iliac spine.

[0024] Exemplarily, the sample body features of a certain 25-year-old male user can include: age 25, male, height 178 cm, weight 70 kg, BMI 21.8, no history of puncture site related diseases, and the basic physiological features of the posterior superior iliac spine are bone density 1.2 g / cm 3 , local subcutaneous fat thickness 0.8 cm. These diversified sample body features provide rich basic data for subsequent model learning of the morphological change rules of different individuals.

[0025] On this basis, the obtained body feature information is input into the morphological change predictor, and the corresponding morphological change rate and morphological change amplitude are output after analysis by the predictor.

[0026] The training step of the "morphological change predictor" in the system provided by the embodiments of the present application includes: Using body features as input data, using morphological change rate and morphological change amplitude as output data, and based on machine learning, a morphological change predictor is constructed; According to the puncture record data of other users in the historical time, a sample body feature set is collected, and the probability and average amplitude of morphological change of the user puncture site under different sample body features are collected, and a sample morphological change rate set and a sample morphological change amplitude set are labeled; Based on the sample body feature set, the sample morphological change rate set and the sample morphological change amplitude set, the morphological change predictor is iteratively supervised trained, and the training is completed after convergence.

[0027] In the embodiments of the present application, in order to enable the morphological change predictor to accurately capture the correlation between body features and morphological change of the puncture site, so as to accurately output the morphological change rate and the morphological change amplitude, the predictor needs to be constructed and trained based on a large amount of historical data, so as to realize personalized analysis of the morphological change of the puncture site of different users.

[0028] Firstly, the core elements of model construction need to be clarified, that is, the body features of the user are taken as input data, the morphological change rate and the morphological change amplitude of the puncture site are taken as output data, and the basic framework of the morphological change predictor is built relying on the machine learning algorithm, so as to ensure that the model has the ability to learn the mapping relationship between the input and the output.

[0029] Specifically, first, the collection and labeling of sample data are carried out. That is, from the puncture record data of other users in the historical time, a sample body feature set is systematically collected, and these features cover various key information that can reflect the body condition of the user.

[0030] Meanwhile, the probability (i.e., the morphological change rate) and the average amplitude (i.e., the morphological change amplitude) of the morphological change of the corresponding puncture site are collected in detail according to the physical characteristics of different samples, and these data are accurately labeled, and then a complete sample morphological change rate set and a sample morphological change amplitude set are formed to provide high-quality supervised data for model training.

[0031] For example, the sample physical characteristics of a 30-year-old male user collected from historical puncture records are: age 30, male, height 180 cm, weight 75 kg, BMI 23.1, no puncture site related diseases, and the puncture site is the posterior superior iliac spine. Correspondingly, the probability of morphological change (morphological change rate) of the puncture site of the user during the puncture process is 12%, and the average amplitude of morphological change is 1.5 mm. These data are labeled into the sample morphological change rate set and the sample morphological change amplitude set, respectively.

[0032] Further, the sample physical characteristic set is used as the input data, and the sample morphological change rate set and the sample morphological change amplitude set are used as the output data, and a morphological change predictor is constructed based on machine learning and trained.

[0033] Specifically, a multilayer perceptron (MLP) is used as the basic architecture to construct the morphological change predictor. The architecture includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is determined according to the dimension of the sample physical characteristics, for example, if the sample physical characteristics contain 8 dimensions of information, 8 nodes are set in the input layer; the hidden layer can be set to 2-3 layers, and the number of nodes in each layer is determined through experimental debugging, for example, 64 nodes are set in the first layer and 32 nodes are set in the second layer, which are used to extract the nonlinear association features between the physical characteristics and the morphological change; the output layer is set to 2 nodes, which correspond to the predicted values of the morphological change rate and the morphological change amplitude, respectively.

[0034] Further, the sample physical characteristic set, the sample morphological change rate set, and the sample morphological change amplitude set are divided in a ratio of 8:1:1, for example, 1600 groups are selected from 2000 groups of samples as the training set, 200 groups as the validation set, and 200 groups as the test set, to ensure that the samples cover different population characteristics and guarantee the comprehensiveness and representativeness of the data distribution.

[0035] Further, in the model training stage, the sample physical characteristics of the training set are input into the model, and the network parameters are optimized through the back propagation algorithm, taking the loss function (such as mean square error) between the predicted morphological change rate and amplitude and the sample labeled value as the optimization objective, so that the loss value gradually decreases. The initial learning rate is set to 0.005, and the learning rate is attenuated to half of the previous value every 100 iterations to improve the convergence speed and accuracy of the model.

[0036] Meanwhile, the model performance is monitored in real time using the validation set. If the loss function of the validation set decreases by less than 0.0001 for 20 consecutive rounds, the model is determined to have converged, and the training is stopped.

[0037] Further, in the test phase, the test set is input into the trained model. If the average deviation between the predicted morphological change rate and the actual labeled rate is less than 3%, and the average deviation between the predicted morphological change amplitude and the actual labeled amplitude is less than 0.3 mm, the morphological change predictor is determined to be qualified and can be put into use.

[0038] Finally, the trained morphological change predictor can accurately output the corresponding morphological change rate and morphological change amplitude according to the input user body features.

[0039] For example, when the body features of a certain user (such as a 35-year-old female, 162 cm tall, 55 kg, BMI 20.9, no history of puncture site-related diseases) are input into the morphological change predictor, the model outputs the morphological change rate of the puncture site (such as the posterior superior iliac spine) of the user as 18%, and the morphological change amplitude as 2.7 mm. Compared with the actual morphological change record of the puncture site of the user, the actual change rate is 17%, and the change amplitude is 2.6 mm, both of which are within the allowed range and meet the accuracy requirements.

[0040] The morphological change predictor constructed through the above steps quickly operates through the built-in machine learning model to output the morphological change rate and morphological change amplitude of the puncture site of the user, and stores these two data in association with the user identification information, providing key basic data support for subsequent standard site image adjustment and puncture adaptability analysis.

[0041] The image adjustment analysis module 02 is used to collect the site image of the puncture site of the user, adjust the standard site image according to the morphological change rate and morphological change amplitude, obtain the adjusted site image, and perform similarity analysis on the adjusted site image and the site image to obtain the adjusted site image sequence and the similarity sequence, wherein the adjustment times are configured according to the morphological change rate. In the biopsy puncture angle optimization scenario in the embodiments of the present application, in order to make the adjusted site image more consistent with the actual puncture site of the user and provide accurate image basis for subsequent puncture adaptability analysis, the adjusted site image sequence and the similarity sequence that are adapted need to be obtained through dynamic adjustment and similarity analysis of the standard site image.

[0042] Specifically, first, the site image of the puncture site of the user is collected and used as a comparison reference. Meanwhile, the preset adjustment times are obtained, and the preset adjustment times are adjusted according to the ratio of the morphological change rate of the current user to the average morphological change rate of users with different body features to obtain the final adjustment times, so as to ensure that the adjustment times match the morphological change probability of the puncture site of the user.

[0043] Meanwhile, a historical morphological change image set corresponding to the current user morphological change amplitude is obtained, and the image set contains image features of the puncture site under a similar morphological change amplitude in the past; a first historical morphological change image is randomly selected from the historical morphological change image set, and the features of the first historical morphological change image are applied to the adjustment and replacement of the standard site image to generate a first adjusted site image.

[0044] Further, a similarity analysis is performed on the first adjusted site image and the site image of the user to obtain a first similarity; in the same way, the historical morphological change image set is continuously selected to adjust the standard site image, and the similarity between the image after each adjustment and the site image of the user is calculated until the determined adjustment number is reached, and finally an adjusted site image sequence and a corresponding similarity sequence are formed.

[0045] The technical scheme of dynamically adjusting the image by combining the individual morphological change characteristics of the user and analyzing the similarity provides a multi-dimensional image reference for subsequent adaptive analysis of the puncture angle, and guarantees the pertinence and reliability of the puncture angle optimization.

[0046] As shown in the accompanying Figure 2 The image adjustment and analysis module 02 in the system provided by the embodiments of the present application includes: Collecting a site image of a puncture site of a user; Obtaining a preset adjustment number, adjusting and calculating the preset adjustment number according to the ratio of the morphological change rate and the average morphological change rate of different body feature users to obtain an adjustment number; Obtaining a historical morphological change image set under the morphological change amplitude; Randomly selecting a first historical morphological change image from the historical morphological change feature set, adjusting and replacing the standard site image to obtain a first adjusted site image; Analyzing the similarity of the first adjusted site image and the site image to obtain a first similarity; Continuing the standard site image adjustment and similarity analysis until the preset adjustment number is reached to obtain an adjusted site image sequence and a similarity sequence.

[0047] In the embodiments of the present application, in order to dynamically adjust the standard site image to match the actual puncture site characteristics of the user and quantify the fit degree of different adjustment results and the actual site image, a scientific image adjustment and similarity analysis model needs to be constructed to realize the accurate adaptive analysis of the puncture site image and provide a multi-dimensional image reference for subsequent puncture angle optimization.

[0048] Specifically, first, the collection of user puncture site images is carried out, that is, the high-definition part image of the user puncture site is obtained through a medical imaging device (such as an ultrasound device, a CT device, etc.), which is used as a reference template for subsequent image adjustment and similarity analysis.

[0049] Illustratively, for the posterior superior iliac spine puncture site, an ultrasound image containing the details of the bone contour, surrounding soft tissue distribution, etc. is collected.

[0050] Further, the number of adjustments is calculated. First, the initial adjustment number preset by the system (such as 20 times) is obtained, and then the preset adjustment number is dynamically corrected according to the ratio of the current user's morphological change rate to the average morphological change rate of users with different body characteristics. The specific calculation formula can be expressed as "adjustment number = preset adjustment number × (current user morphological change rate / average morphological change rate)".

[0051] Illustratively, if the preset adjustment number is 20 times, the current user's morphological change rate is 25%, and the average morphological change rate is 15%, then the adjustment number = 20 × (25% / 15%) ≈ 33 times, which ensures that users with a higher morphological change probability can obtain more adjustment samples and improve the comprehensiveness of image adaptation.

[0052] At the same time, the historical morphological change image set matching the current user's morphological change amplitude is obtained. This historical morphological change image set comes from the historical puncture records of the puncture site images with the same puncture site as the current user and similar morphological change amplitude (such as a deviation within ±0.5 mm), covering different morphological change types such as local bulging, depression, bone density change, etc.

[0053] Illustratively, when the morphological change amplitude of the current user's posterior superior iliac spine is 3 mm, the images with a morphological change amplitude in the range of 2.5-3.5 mm of this site are selected from the historical puncture records to form a targeted historical morphological change image set.

[0054] On this basis, the adjustment of the standard part image is carried out. The standard part image refers to the standard image template of the puncture site in the ideal state. For example, the standard ultrasound image of the posterior superior iliac spine without obvious morphological change.

[0055] Further, a first historical morphological change image is randomly selected from the historical morphological change image set, the morphological change features (such as local tissue thickness change, bone edge offset, etc.) are extracted, and these features are applied to the corresponding area of the standard part image to realize the adjustment and replacement of the standard image, thereby generating a first adjusted part image.

[0056] Exemplarily, a feature of a 2mm edge offset of the posterior superior iliac spine in a certain historical morphological change image is selected, the corresponding edge in the standard site image is adjusted by the same offset, and a first adjusted site image containing the change feature is obtained.

[0057] Further, the similarity of the adjusted image and the actual image of the user is analyzed to obtain a first similarity.

[0058] In the system provided by the embodiments of the present application, the step of "analyzing the similarity of the first adjusted site image and the site image to obtain a first similarity" comprises: obtaining a sample site image set, randomly combining the sample site images to obtain a plurality of sample site image combinations; annotating the morphological similarity of each sample site image combination to obtain a sample similarity set; constructing a site image similarity identifier based on a twin neural network; training the site image similarity identifier using the plurality of sample site image combinations and the sample similarity set, and completing the training after testing convergence; inputting the first adjusted site image and the site image combination into the site image similarity identifier to output the first similarity.

[0059] In the embodiments of the present application, in order to accurately quantify the morphological fit degree of the first adjusted site image and the user's puncture site image, a site image similarity identifier based on a twin neural network needs to be constructed, and the morphological correlation features between images are learned through model learning, so as to realize efficient and accurate analysis of the similarity of different site image combinations.

[0060] Specifically, first, sample data collection and combination work is carried out. That is, a sample site image set is comprehensively obtained from an image library of historical puncture records, which includes puncture site images of different puncture sites (such as the posterior superior iliac spine, breast, liver, etc.), different imaging devices (such as ultrasound, CT, MRI, etc.) and different morphological features (such as normal morphology, different degrees of lesion morphology, postoperative morphology, etc.).

[0061] Exemplarily, 10,000 ultrasound images of the posterior superior iliac spine are collected, including image data of the site of users of different ages and genders in different physiological states.

[0062] On this basis, the sample site images are randomly combined two by two to generate a plurality of sample site image combinations.

[0063] The combination mode includes both the combination of images of the same part in different states, such as the combination of the normal form image of the posterior superior iliac spine of a certain user and the slightly deformed image of the part, and the combination of images of the same part of different individuals, such as the combination of the posterior superior iliac spine image of user A and the posterior superior iliac spine image of user B, so as to ensure the diversity and comprehensiveness of the sample combination.

[0064] Meanwhile, the morphological similarity of each sample part image combination is labeled to form a sample similarity set.

[0065] Specifically, the morphological similarity labeling work is completed by experienced medical image experts, and a quantitative scoring system of 0-100 points is adopted based on indicators such as the contour coincidence degree of the puncture part in the image, the internal structure distribution consistency, and the key feature point offset, wherein 0 points represent complete dissimilarity and 100 points represent complete consistency.

[0066] For example, in a group of posterior superior iliac spine image combinations, the bone contour coincidence degree of the two is 95%, the internal soft tissue distribution difference is less than 3%, and the expert labels the similarity as 92 points; in another group of images, the bone edge is obviously offset, and the internal structure difference is large, and the labeled similarity is 35 points. Through such a way, the labeling of 50,000 sample image combinations is completed to form a complete sample similarity set.

[0067] Further, a part image similarity recognizer is constructed based on a twin neural network. The twin neural network is composed of two sub-networks with the same structure and shared parameters, and each sub-network includes an input layer, a convolution layer, a pooling layer, a full connection layer, etc.

[0068] Specifically, the input layer receives a single part image (such as an ultrasound image with a size normalized to 256x256 pixels); the convolution layer uses multiple convolution kernels of different sizes (such as 3x3 and 5x5) to extract local features (such as edges, textures, and gray scale distribution) of the image. For example, the first layer of convolution layer uses 32 3x3 convolution kernels, and the second layer uses 64 5x5 convolution kernels.

[0069] Meanwhile, the pooling layer compresses the feature dimension through the maximum pooling operation to reduce the amount of calculation while retaining the key features; the full connection layer maps the high-dimensional features to a fixed-dimensional feature vector (such as 128 dimensions). After the two sub-networks extract features from the input two images respectively, the similarity of the two feature vectors is calculated through the distance calculation layer, and then the similarity is converted to a score result of 0-100 points through the output layer.

[0070] Further, the part image similarity recognizer is trained using the sample part image combination and the sample similarity set.

[0071] Specifically, first, the sample data is divided into a training set, a validation set, and a test set in a ratio of 7:2:1. For example, 35000 groups are selected from 50000 groups of samples as the training set, 10000 groups as the validation set, and 5000 groups as the test set.

[0072] At the same time, in the training process, the mean square error between the predicted similarity and the expert-labeled similarity is used as the loss function, and the network parameters are optimized through the back propagation algorithm. The initial learning rate is set to 0.001, and the learning rate is attenuated to 1 / 10 of the previous one every 60 iterations to balance the model convergence speed and training accuracy.

[0073] Further, in the training phase, the model performance is monitored in real time using the validation set. If the loss function of the validation set decreases by less than 0.001 for 15 consecutive rounds, the model is determined to have converged and training is stopped.

[0074] Further, the performance of the trained model is evaluated on the test set. If the average absolute error between the predicted similarity and the expert-labeled similarity is less than 2 points, the part image similarity recognizer is determined to be qualified and can be put into actual use.

[0075] Finally, the first adjusted part image and the user's part image are combined into an image pair and input into the qualified part image similarity recognizer. The model extracts the feature vectors of the two images through two sub-networks, calculates the similarity between the vectors and converts it into a score, and the output is the first similarity.

[0076] For example, a first adjusted part image (posterior superior iliac spine adjustment image) and a user's part image (actual image of the same part) are input into the part image similarity recognizer, and the model outputs a similarity of 88 points. After expert review, the two images have a high degree of agreement in key features such as bone contour and soft tissue distribution, and the score is consistent with the actual situation, verifying the accuracy of the model.

[0077] The first similarity obtained through the above steps can objectively reflect the matching degree of the adjusted image and the user's actual puncture site, ensuring the scientificity and accuracy of the puncture angle optimization.

[0078] Further, after obtaining the first adjusted part image and the first similarity, the standard part image is adjusted and analyzed according to the same process.

[0079] Specifically, a second historical morphological change image is randomly selected from the historical morphological change image set again, its morphological change features are extracted, the standard part image is adjusted and replaced for the second time, and a second adjusted part image is generated. Then, the second adjusted part image and the user's part image are combined and input into the qualified part image similarity recognizer to output a second similarity.

[0080] Similarly, each round of adjustment randomly selects new historical morphological change images from the historical morphological change image set, differentiates the standard site images, generates new adjusted site images, and synchronously calculates the similarity between the images and the user site images. This process is continuously repeated until the adjustment and analysis operations equal the preset number of adjustments are completed.

[0081] For example, if the preset number of adjustments is 30 times, after 30 rounds of adjustment, an adjusted site image sequence containing 30 adjusted images will be generated, and a similarity sequence composed of 30 similarity values will be generated.

[0082] The adjusted site image sequence covers image samples generated based on different historical morphological change characteristics, which may have different degrees of fit with the user's puncture site, and the similarity sequence quantifies the matching degree of each sample with the user's actual site image (for example, the sequence may contain different scores such as 90 points, 85 points, and 72 points).

[0083] This step can fully cover the potential state of the user's puncture site under different morphological change scenarios, taking into account the difference in the number of adjustments caused by the morphological change rate, and quantifying the fit degree of different adjustment results with the actual situation through similarity, providing comprehensive and accurate image data support for subsequent adaptive analysis of the puncture angle, to ensure that the subsequent optimization process can filter out the most suitable puncture angle for the user's individual characteristics based on diversified image samples.

[0084] The puncture adaptation analysis module 03 is used for randomly generating a biopsy puncture angle, combining the adjusted site image sequence and the similarity sequence, and performing puncture adaptability analysis to obtain a puncture adaptation parameter. In the embodiment of the present application, in the scenario of optimizing the biopsy puncture angle, in order to evaluate the adaptability of different puncture angles to the user's actual puncture site and provide a quantitative basis for subsequent angle optimization, the puncture adaptation parameter reflecting the angle adaptability is obtained by randomly generating a puncture angle and combining the adjusted site image sequence and the similarity sequence for adaptability analysis.

[0085] Specifically, first, based on the conventional angle range of the puncture operation, a number of initial biopsy puncture angles are generated by random generation to ensure the comprehensiveness of the angle coverage range and avoid the bias in adaptability evaluation caused by limited angle selection.

[0086] Meanwhile, each generated biopsy puncture angle is combined with each adjusted site image in the adjusted site image sequence and input into the puncture adaptation analyzer.

[0087] The puncture adaptation analyzer analyzes the feasibility, safety and accuracy of the puncture according to the biopsy puncture angle and the anatomical structure characteristics of the puncture site in the image, outputs corresponding single puncture adaptation parameters, and finally forms a single puncture adaptation parameter sequence corresponding to the adjustment site image sequence.

[0088] Finally, the single puncture adaptation parameter sequence is weighted calculated according to the similarity sequence to obtain a puncture adaptation parameter that can comprehensively reflect the puncture angle adaptation.

[0089] Through this weighting method, the final puncture adaptation parameter can more focus on the adaptation analysis result corresponding to the image of the actual situation of the user, and comprehensively reflect the overall adaptability of the puncture angle under different potential morphologies, providing more targeted and reliable quantitative basis for subsequent puncture angle optimization.

[0090] The puncture adaptation analysis module 03 in the system provided by the embodiment of the application comprises: randomly generating a biopsy puncture angle; inputting the biopsy puncture angle combined with the influence of each adjustment site into the puncture adaptation analyzer, and outputting a single puncture adaptation parameter sequence; weighting calculating the single puncture adaptation parameter sequence according to the similarity sequence to obtain a puncture adaptation parameter.

[0091] In the embodiment of the application, in order to comprehensively evaluate the adaptation degree of different puncture angles to the actual puncture site of the user, the process of randomly generating angles, analyzing single adaptation combined with image sequences and weighting calculating comprehensive parameters is needed to construct a scientific puncture adaptation evaluation mechanism to provide quantitative basis for subsequent angle optimization.

[0092] First, a biopsy puncture angle is randomly generated. Specifically, based on the anatomical characteristics of the puncture site and the clinical operation specification, a reasonable angle range is set, and an initial puncture angle is generated in the range by using a random sampling algorithm in the prior art.

[0093] For example, a uniform distribution random function is used to generate a plurality of angle values between 10° and 80° to ensure the comprehensiveness of the puncture angle coverage and avoid the deviation of the adaptation evaluation caused by the limitation of the initial angle selection.

[0094] At the same time, a puncture adaptation analyzer is constructed to realize the output of single puncture adaptation parameters.

[0095] The training steps of the "puncture adaptation analyzer" in the system provided by the embodiment of the application comprise: constructing a puncture adaptation analyzer based on machine learning; Collect a sample site image set and a sample biopsy puncture angle set, label puncture adaptation parameters of different sample site images and sample biopsy puncture angles, and obtain a sample puncture adaptation parameter set; The sample site image set, the sample biopsy puncture angle set, and the sample puncture adaptation parameter set are used to supervise training of the puncture adaptation analyzer, and the training is completed after convergence.

[0096] In the embodiments of the present application, in order to enable the puncture adaptation analyzer to accurately evaluate the adaptation degree of different puncture angles and puncture site images and output reliable puncture adaptation parameters, the analyzer needs to be constructed and trained based on a large amount of historical sample data to learn the correlation between the puncture angle and the site image features, thereby meeting the demand for individualized puncture angle evaluation.

[0097] Firstly, the core function of the puncture adaptation analyzer needs to be determined, that is, according to the input puncture site image and puncture angle, a quantitative puncture adaptation parameter is output, and the parameter needs to comprehensively reflect the safety, accuracy and feasibility of puncture.

[0098] Based on this, a deep neural network (DNN) is selected as the basic architecture to construct the puncture adaptation analyzer.

[0099] The deep neural network architecture includes an input layer, a hidden layer and an output layer. The input layer receives a preprocessed sample site image feature vector and a sample biopsy puncture angle value. The hidden layer extracts nonlinear correlation features between the two through multiple neurons. The output layer outputs the corresponding puncture adaptation parameter.

[0100] Specifically, first, a sample site image set is systematically collected from a historical puncture record database. These images cover different puncture sites such as posterior superior iliac spine, breast, thyroid, different imaging modalities such as ultrasound, CT, MRI, and different pathological states such as normal tissue, tumor tissue, and inflammatory tissue, to ensure the diversity and representativeness of the samples.

[0101] For example, 5000 ultrasound images of the posterior superior iliac spine are collected, including images of this site of users of different ages, genders and body types. At the same time, a sample biopsy puncture angle set is collected. For each sample site image, multiple puncture angles commonly used in clinical practice (such as 30°, 45°, 60°, etc.) are selected as sample angles to form an angle combination corresponding to the sample site image.

[0102] On this basis, the combination of different sample site images and sample biopsy puncture angles is labeled for puncture adaptation parameters to form a sample puncture adaptation parameter set.

[0103] Specifically, the puncture adaptation parameter labeling work is also completed by experienced clinicians based on the anatomical structure of the puncture site in the image (such as bone position, blood vessel distribution, tissue density) and the path simulation of the puncture angle, and is scored from three dimensions of safety, accuracy and feasibility, using a quantitative scoring system of 0-100 points, and the higher the score, the better the adaptability.

[0104] For example, for a combination of a certain posterior superior iliac spine ultrasound image and a 45° puncture angle, if the puncture path avoids major blood vessels and nerves with a probability of 95%, the estimated error of reaching the target position is 1 mm, and the operation difficulty is low, then the puncture adaptation parameter is labeled as 90 points.

[0105] In addition, for the same image and a 70° puncture angle combination, if the puncture path may damage the adjacent blood vessels, the estimated error is 3 mm, and the operation difficulty is high, then the parameter is labeled as 60 points. In this way, the labeling of all sample combinations is completed to form a complete sample puncture adaptation parameter set.

[0106] Further, the sample site image set, the sample biopsy puncture angle set and the sample puncture adaptation parameter set are used to supervise the training of the puncture adaptation analyzer.

[0107] Specifically, first, the sample data is preprocessed, that is, the sample site image is converted to a fixed size of 256x256 pixels and the feature vector is extracted through a convolutional neural network, and the sample biopsy puncture angle value is converted to the range of 0-1 for normalization processing.

[0108] Further, the preprocessed sample site image feature vector and sample biopsy puncture angle value are used as input, and the sample puncture adaptation parameter is used as a supervision signal, which is divided into a training set, a validation set and a test set in a ratio of 7:2:1. For example, 7000 groups are selected from 10000 sample combinations as the training set, 2000 groups as the validation set, and 1000 groups as the test set.

[0109] In the training phase, the training set is input into the puncture adaptation analyzer, and the network parameters are optimized through the back propagation algorithm, and the mean square error between the predicted puncture adaptation parameter and the sample labeled value is used as the loss function, and the loss value is continuously reduced. The initial learning rate is set to 0.001, and the learning rate is attenuated to 1 / 10 of the previous one every 50 iterations to balance the speed and accuracy of training.

[0110] At the same time, the performance of the model is monitored in real time using the validation set, and if the loss function of the validation set decreases by less than 0.001 for 15 consecutive rounds, the model is determined to be converged, and the training is stopped immediately.

[0111] After the training is completed, the performance of the puncture adaptation analyzer is evaluated on the test set, and if the average absolute error between the predicted puncture adaptation parameters and the sample labeled values is less than 2 points, the analyzer is determined to be qualified and can be put into actual use.

[0112] Finally, the qualified puncture adaptation analyzer can quickly and accurately output the corresponding puncture adaptation parameters according to the input puncture site image and puncture angle, providing reliable quantitative basis for subsequent puncture angle adaptation analysis and ensuring the scientificity and effectiveness of puncture angle optimization.

[0113] For example, when a certain ultrasound image of the posterior superior iliac spine and a 40° puncture angle are input into the qualified puncture adaptation analyzer, the model outputs a puncture adaptation parameter of 86 points by analyzing the matching relationship between the bone edges, blood vessel directions in the image and the puncture path under the angle.

[0114] Meanwhile, according to the review of clinicians, the probability of puncture avoiding major blood vessels under this angle reaches 92%, and the estimated error to the target position is only 0.8 mm, which is consistent with the parameter score output by the analyzer, verifying the accuracy of the analyzer.

[0115] Further, randomly generated biopsy puncture angles are combined with each image in the adjusted site image sequence in turn and input into the qualified puncture adaptation analyzer. The analyzer will output corresponding single puncture adaptation parameters for each "biopsy puncture angle-adjusted site image" combination based on the associated rules learned by it, and finally form a single puncture adaptation parameter sequence consistent with the length of the adjusted site image sequence.

[0116] Further, the single puncture adaptation parameter sequence is weighted and calculated according to the obtained similarity sequence to obtain a comprehensive puncture adaptation parameter. The specific calculation formula can be expressed as "puncture adaptation parameter = Σ(single puncture adaptation parameter i × similarity i / Σ similarity i )", where i represents the ith element in the sequence.

[0117] Among them, since each value in the similarity sequence represents the degree of fit between the corresponding adjusted site image and the user's actual puncture site image, the higher the fit degree, i.e. the larger the similarity value, the more the adjusted site image can reflect the user's true situation, so it needs to give higher weight to the corresponding single puncture adaptation parameter in the calculation. The calculation method of the weight is "the weight corresponding to a certain adjusted site image = the similarity value of the image / the sum of all similarity values in the similarity sequence." Exemplarily, if the adjustment site image sequence contains 3 images, the corresponding similarity sequence is [90, 80, 70], and the single body puncture adaptation parameter sequence is [85, 72, 68]. First, the similarity sum is calculated as 90+80+70=240; then the weight of each element is calculated as 90 / 240=0.375, 80 / 240≈0.333, and 70 / 240≈0.292, respectively; and finally, the weighted calculation is performed: 85x0.375+72x0.333+68x0.292≈75.707, that is, the puncture adaptation parameter of the puncture angle is about 75.7 points.

[0118] This step enables the final puncture adaptation parameter to focus on the adaptability analysis results corresponding to the adjustment site images with high fitting degree with the actual puncture site image of the user by performing weighted calculation on the single body puncture adaptation parameter sequence based on the similarity sequence, so as to more objectively and accurately reflect the overall adaptability of the puncture angle under different morphologies.

[0119] The angle optimization display module 04 is configured to perform puncture angle optimization, obtain an optimal puncture angle, and display the optimization result.

[0120] In the embodiments of the present application, in order to select the most suitable angle for the actual puncture site of the user from a large number of possible puncture angles, new angles are continuously generated and their adaptability is analyzed until the optimization convergence condition is reached, and finally the optimal puncture angle is determined and displayed to assist clinical puncture operation.

[0121] Specifically, on the basis of having obtained the initial puncture angle and the corresponding puncture adaptation parameter, new biopsy puncture angles are continuously generated in the same random generation manner. For each newly generated angle, the steps of “randomly generating an angle-puncture adaptation analysis-weighted calculation-obtaining a puncture adaptation parameter” are repeated to obtain the corresponding new puncture adaptation parameter.

[0122] Further, by continuously generating new puncture angles and calculating their puncture adaptation parameters, a mapping relationship between a series of puncture angles and corresponding adaptation puncture adaptation parameters is formed.

[0123] During this process, the change trend of the puncture adaptation parameter is continuously monitored, and if the improvement amplitude of the puncture adaptation parameter corresponding to the newly generated puncture angle for a plurality of times is less than a preset threshold, it is determined that the optimization process meets the convergence condition.

[0124] At this time, the angle corresponding to the highest puncture adaptation parameter is selected from all generated puncture angles, which is the optimal puncture angle.

[0125] Subsequently, the optimal puncture angle is displayed in an intuitive manner, and puncture path simulation, adaptability score and other auxiliary information under the angle are provided, so that medical staff can obtain clear and reliable puncture angle reference, and the accuracy and safety of biopsy puncture operation are improved.

[0126] The angle optimization display module 04 in the system provided by the embodiment of the application comprises: The biopsy puncture angle is continuously randomly generated, and new puncture adaptability parameters are obtained; Until the optimization meets the convergence, the optimal puncture angle is obtained, and the optimization result is displayed.

[0127] In the embodiment of the application, in order to screen the angle most suitable for the individual puncture site characteristics of the user from a large number of possible puncture angles, new angles are continuously generated and their adaptability is analyzed until the optimization convergence condition is reached, and finally the optimal puncture angle is determined and presented in an intuitive manner, thereby providing accurate reference for clinical puncture operation.

[0128] Specifically, on the basis of the obtained initial puncture angle and corresponding puncture adaptability parameters, new biopsy puncture angles are continuously generated in the same random sampling manner as the initial angle generation.

[0129] Further, for each newly generated biopsy puncture angle, the processing flow of “randomly generating an angle-puncture adaptability analysis-weighted calculation-obtaining puncture adaptability parameters” is strictly repeated, that is, the biopsy angle is combined with each image in the adjustment site image sequence in turn, the single puncture adaptability parameter sequence is obtained by inputting the puncture adaptability analyzer, and the single puncture adaptability parameter sequence is weighted calculated according to the similarity sequence, and finally the puncture adaptability parameters corresponding to the new puncture angle are obtained.

[0130] On this basis, the mapping data of the puncture angle and the corresponding puncture adaptability parameters are continuously accumulated to form a data set containing multiple groups of “puncture angle-puncture adaptability parameters”.

[0131] At the same time, the dynamic change trend of the puncture adaptability parameters needs to be monitored in real time to determine whether the optimization meets the convergence condition. The specific convergence condition can be set as: when the difference between the puncture adaptability parameters corresponding to the new puncture angles generated for a continuous preset number of times (such as 10 times) and the current optimal puncture adaptability parameter is less than a set threshold (such as 1 minute), it is considered that the puncture adaptability parameter improvement amplitude tends to be stable.

[0132] For example, if the current optimal puncture adaptability parameter is 89 minutes, and the puncture adaptability parameters of the new puncture angles for 10 consecutive times are 88.5 minutes, 89 minutes, 88.8 minutes, etc., all of which fluctuate within 89 ± 1 minutes, it is determined that the optimization meets the convergence condition.

[0133] Further, when the optimization converges, the angle corresponding to the highest puncture adaptation parameter among all generated puncture angles is selected as the optimal puncture angle. This angle comprehensively considers the morphological change characteristics of the user's puncture site, image adaptability, and puncture adaptability under different potential morphologies, and is the puncture angle that theoretically best meets the individual characteristics of the user.

[0134] Finally, the optimal puncture angle is displayed in various forms. On the one hand, the specific angle value is displayed; on the other hand, the puncture path under the puncture angle is dynamically marked in combination with the three-dimensional image model of the user's puncture site, and the trajectory of the needle tip reaching the target position is intuitively presented; at the same time, the puncture adaptation parameter score corresponding to the optimal puncture angle is also attached as a key evaluation index.

[0135] For example, a user obtains an optimal puncture angle of 45° after optimization, and the corresponding puncture adaptation parameter is 92 points. In the specific display, not only is the "optimal puncture angle: 45°" displayed, but also the 45° puncture path is marked in the three-dimensional ultrasound model of the user's posterior superior iliac spine, and it is noted that "the puncture adaptability score under this puncture angle is 90 points", to help medical personnel quickly understand the adaptation advantages of the puncture angle, thereby improving the efficiency and safety of the biopsy puncture operation.

[0136] Through the specific embodiments described above, the following technical effects are achieved: The application provides a biopsy needle puncture angle optimization system. First, the body characteristics of a user are collected and input into a morphological change predictor trained by historical data to obtain the morphological change rate and morphological change amplitude of the puncture site, providing a data basis for subsequent precise adjustment. Then, the site image of the user's puncture site is collected, the standard site image is adjusted multiple times according to the morphological change rate and morphological change amplitude, and the adjusted site image sequence and similarity sequence are obtained by combining similarity analysis to comprehensively cover the possible morphological states of the puncture site. Subsequently, a biopsy puncture angle is randomly generated, and a single puncture adaptation parameter sequence is obtained by combining the adjusted site image sequence and similarity sequence through a puncture adaptation analyzer, and a puncture adaptation parameter is obtained by weighted calculation. Finally, new puncture angles are continuously generated and their corresponding puncture adaptation parameters are calculated until the optimization converges, and the angle with the highest puncture adaptation parameter is determined as the optimal puncture angle and displayed, providing a precise reference for clinical operation.

[0137] The system provided by the embodiments of the application solves the problems of low puncture precision and high risk caused by insufficient consideration of individual differences and fixed angle selection in traditional biopsy puncture, realizes personalized and precise optimization of the puncture angle, and provides reliable technical support for improving the safety and effectiveness of biopsy puncture operation.

[0138] It should be noted that the above-mentioned embodiment sequences of the present application are merely for description only, but not for representing the advantages and disadvantages of the embodiments. And the above-mentioned embodiments of the present specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0139] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0140] The specification and drawings are merely exemplary of the present application, and any and all modifications, variations, combinations or equivalents that are within the scope of the present application should be considered covered by the present application. Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A biopsy needle puncture angle optimization system, characterized by, The system comprises: a physical feature analysis module for collecting physical features of a user, performing puncture site morphological change analysis, and obtaining a morphological change rate and a morphological change amplitude; an image adjustment analysis module for collecting a site image of a puncture site of a user, adjusting a standard site image according to the morphological change rate and the morphological change amplitude, obtaining an adjusted site image, and performing similarity analysis on the adjusted site image and the site image to obtain an adjusted site image sequence and a similarity sequence, wherein the number of adjustments is configured according to the morphological change rate; a puncture adaptability analysis module for randomly generating a biopsy puncture angle, combining the adjusted site image sequence and the similarity sequence, and performing puncture adaptability analysis to obtain puncture adaptability parameters; an angle optimization display module for performing puncture angle optimization to obtain an optimal puncture angle, and displaying the optimal puncture angle as an optimization result.

2. The biopsy needle puncture angle optimization system of claim 1, wherein, Collecting physical features of a user, performing puncture site morphological change analysis, and obtaining a morphological change rate and a morphological change amplitude comprise: collecting physical features of a user; inputting the physical features into a morphological change predictor to output a morphological change rate and a morphological change amplitude.

3. The biopsy needle puncture angle optimization system of claim 2, wherein, The morphological change predictor is trained by the following steps: using physical features as input data and using a morphological change rate and a morphological change amplitude as output data, constructing a morphological change predictor based on machine learning; collecting a sample physical feature set from puncture record data of other users in a historical period, and collecting the probability and average amplitude of morphological change of a user's puncture site under different sample physical features to label a sample morphological change rate set and a sample morphological change amplitude set; iteratively supervising and training the morphological change predictor based on the sample physical feature set, the sample morphological change rate set, and the sample morphological change amplitude set, and completing the training after convergence.

4. The biopsy needle puncture angle optimization system of claim 1, wherein, Collecting a site image of a puncture site of a user, adjusting a standard site image according to the morphological change rate and the morphological change amplitude, obtaining an adjusted site image, and performing similarity analysis on the adjusted site image and the site image to obtain an adjusted site image sequence and a similarity sequence comprise: collecting a site image of a puncture site of a user; adjusting a preset number of adjustments according to the ratio of the morphological change rate to the average morphological change rate of users with different physical features to obtain the number of adjustments; obtaining a historical morphological change image set under the morphological change amplitude; randomly selecting a first historical morphological change image from the historical morphological change feature set to replace and adjust the standard site image to obtain a first adjusted site image; analyzing the similarity between the first adjusted site image and the site image to obtain a first similarity; continuing to adjust the standard site image and perform similarity analysis until the preset number of adjustments is reached to obtain an adjusted site image sequence and a similarity sequence.

5. The biopsy needle puncture angle optimization system of claim 4, wherein, Analyzing the similarity between the first adjusted site image and the site image to obtain a first similarity comprises: obtaining a sample site image set, randomly combining sample site images to obtain a plurality of sample site image combinations; annotating the morphological similarity of each sample site image combination to obtain a sample similarity set; Based on the twin neural network, a part image similarity recognizer is constructed; The plurality of sample part image combinations and sample similarity sets are used to train the part image similarity recognizer, and the training is completed after convergence is tested; The first adjusted part image and the part image combination are input into the part image similarity recognizer, and a first similarity is output.

6. The biopsy needle puncture angle optimization system of claim 1, wherein, Randomly generate a biopsy puncture angle, combine the adjusted part image sequence and the similarity sequence, and perform puncture adaptability analysis to obtain puncture adaptation parameters, including: Randomly generating a biopsy puncture angle; The biopsy puncture angle is input into the puncture adaptability analyzer combined with the influence of each adjusted part, and a single puncture adaptability parameter sequence is output. According to the similarity sequence, the single puncture adaptability parameter sequence is weighted calculated to obtain the puncture adaptability parameter.

7. The biopsy needle puncture angle optimization system of claim 1, wherein, The training steps of the puncture adaptability analyzer include: Based on machine learning, a puncture adaptability analyzer is constructed; A sample part image set is collected, and a sample biopsy puncture angle set is collected. The puncture adaptability parameters of different sample part images and sample biopsy puncture angles are labeled to obtain a sample puncture adaptability parameter set; The sample part image set, the sample biopsy puncture angle set, and the sample puncture adaptability parameter set are used to supervise the training of the puncture adaptability analyzer, and the training is completed after convergence.

8. The biopsy needle puncture angle optimization system of claim 1, wherein, Optimize the puncture angle to obtain the optimal puncture angle as the optimization result for display, including: Continue to randomly generate a biopsy puncture angle to obtain a new puncture adaptability parameter; Until the optimization meets the convergence, the optimal puncture angle is obtained as the optimization result for display.

Images