A system for building a visual model of brain tumors based on network search and machine learning

Through machine learning methods based on network search, a visual model of craniocerebral tumors is constructed, combined with user feedback optimization recognition model, the problems of insufficient recognition accuracy and user interaction in the existing technology are solved, and the recognition accuracy and user satisfaction are significantly improved.

CN118448056BActive Publication Date: 2024-11-22THE FIRST AFFILIATED HOSPITAL OF JINAN UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410624954.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-20
Publication Date
2024-11-22
Estimated Expiration
2044-05-20

AI Technical Summary

Technical Problem

The existing three-dimensional visualization system has shortcomings in image acquisition, model accuracy and user interaction of craniocerebral tumors, and has failed to effectively optimize dynamically with user feedback, resulting in poor recognition accuracy.

Method used

Using machine learning methods based on network search, we can obtain cranial brain tumor-related information data through the network data end, build and train recognition models on the model construction end, evaluate feedback end to collect and analyze user feedback information, dynamically optimize the recognition model, and optimize loss function to improve Model accuracy.

Benefits of technology

Significantly improve identification accuracy and user satisfaction, ensuring that the model can more accurately reflect current medical imaging information and dynamically optimize performance to meet the latest diagnostic needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118448056B_ABST
    Figure CN118448056B_ABST
Patent Text Reader

Abstract

The present invention provides a system for building a visual model of a cranial brain tumor based on network search and machine learning, the system comprising a network data end, a model building end and an evaluation feedback end; the network data end is used to obtain relevant information data of the cranial brain tumor from the outside world; the model building end builds a visual model of the cranial brain tumor based on the relevant information data of the cranial brain tumor; the evaluation feedback end is used to evaluate the constructed visual cranial brain tumor model and provide feedback information to the model building end to optimize the construction of the cranial brain tumor model; the present invention significantly improves recognition accuracy and user satisfaction by combining a dynamically optimized recognition model and a user feedback mechanism.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of medical image three-dimensional reconstruction systems, and in particular to a system for building a visual model of a cranial brain tumor based on network search and machine learning. Background Art

[0002] Brain tumor is one of the major diseases that affect human health. Its early diagnosis and precise treatment are crucial to improving the survival rate of patients. With the rapid development of medical imaging technology, 3D visualization technology plays an increasingly important role in the diagnosis and treatment of brain tumors. Through 3D reconstruction technology, the location, morphology and relationship of the tumor with surrounding tissues can be observed more intuitively, providing doctors with more detailed diagnostic basis. However, the existing 3D visualization system still has certain shortcomings in image acquisition, model accuracy and user interaction, and fails to effectively combine user feedback for dynamic optimization.

[0003] After consulting the relevant disclosed technical solutions, the technology with publication number CN114783591A proposes a brain tumor medical image three-dimensional visualization auxiliary diagnosis system, which specifically relates to the field of medical image three-dimensional reconstruction technology. The system extracts brain tumor images through a brain tumor extraction function, and draws a three-dimensional model of the brain by fusion drawing of the brain tumor image and the brain tissue image, so as to enhance the display effect, and adds a variety of functional modules to assist diagnosis, including: a transfer function module for adjusting the color and opacity of the model, a measurement module for measuring angles and lengths, a cutting function for segmenting the model, and a slice extraction function for observing slices in multiple directions at the same time; however, the scheme only obtains brain tumor images through a static method, with poor recognition accuracy and no feedback mechanism. Summary of the invention

[0004] The purpose of the present invention is to propose a brain tumor visualization model construction system based on network search and machine learning to address the current deficiencies.

[0005] The present invention adopts the following technical solution:

[0006] A brain tumor visualization model construction system based on network search and machine learning, the system comprising a network data end, a model construction end and an evaluation feedback end;

[0007] The network data end is used to obtain relevant information data of craniocerebral tumors from the outside world; the model building end builds a visual model of craniocerebral tumors based on the relevant information data of craniocerebral tumors; the evaluation feedback end is used to evaluate the constructed visual craniocerebral tumor model and provide feedback information to the model building end to optimize the construction of the craniocerebral tumor model;

[0008] The network data terminal includes a data acquisition module, a data processing module, a data annotation module and a data storage module; the data acquisition module is used to connect with an external data source to obtain relevant information data of craniocerebral tumors, the data processing module is used to pre-process the data acquired by the data acquisition module, and the data annotation module is used to annotate the pre-processed data; the data storage module is used to store the pre-processed data information;

[0009] The model building end includes a model building module, a model training module, a model verification module and a three-dimensional visualization rendering module; the model building module is used to select and build a recognition model for craniocerebral tumor recognition; the model training module is used to perform training on the recognition model; the model verification module is used to evaluate the recognition model using an independent verification set to verify its performance on unseen data; the three-dimensional visualization rendering module generates a high-precision three-dimensional visualization model based on the output results of the recognition model, and the output results of the recognition model are various output features describing craniocerebral tumors;

[0010] The evaluation feedback end includes a feedback collection module, a feedback identification module and a model adjustment module; the feedback collection module is used to collect direct feedback information from users, the feedback identification module is used to identify feedback feature information from the direct feedback information, and the model adjustment module optimizes and adjusts the recognition model based on the feedback feature information;

[0011] Furthermore, the content annotated by the data annotation module includes timestamp information of the preprocessed data obtained from the external data source;

[0012] Furthermore, the model training module divides the preprocessed data into a plurality of batches according to the timestamps of the preprocessed data, each timestamp batch includes data uploaded to the external data source within the same time period, wherein the time period corresponding to each timestamp batch is unique;

[0013] Furthermore, the model training module completes the training of the recognition model through the following steps:

[0014] S31: obtaining preprocessed data of multiple consecutive timestamp batches as a training set, assuming that the total number of batches is m, the sequence batch of the timestamp batch farthest from the current time is set to 1, and the sequence batch of the timestamp batch closest to the current time is set to m;

[0015] S32: Process all data in the training set to convert them into input features that can be used to identify model inputs;

[0016] S33: Input the training set into the recognition model in batches from small to large according to the sequence batches to train the recognition model, and calculate the loss function of each batch of recognition model output;

[0017] S34: After each batch of training is completed, the model parameters are updated by back propagation using the loss function to optimize the recognition model until all batches of training are completed;

[0018] Furthermore, during the training process of the model training module, the loss function Loss of each batch is set to:

[0019]

[0020] Among them, w i is the weight of the i-th output feature, which can be pre-set by the user based on prior knowledge; v is the sampling time of the timestamp batch and the weight of the feature richness; y i is the true value of the i-th output feature, is the model prediction value of the i-th output feature, γ1 and γ2 are regularization parameters used to control model complexity and prevent overfitting;

[0021] A method for constructing a visual model of a craniocerebral tumor based on network search and machine learning, the method comprising the following steps:

[0022] S1: Continuously obtain relevant information and data of brain tumors from external data sources;

[0023] S2: Construct a recognition model for brain tumors;

[0024] S3: Train and validate the recognition model for brain tumors;

[0025] S4: Establish a three-dimensional visualization model based on the output results of the recognition model;

[0026] S5: Adjust and optimize the recognition model based on user feedback.

[0027] The beneficial effects achieved by the present invention are:

[0028] The present invention continuously obtains relevant information data of craniocerebral tumors from the outside world through a network data terminal, thereby providing the model with the latest and detailed training data; optimizes the loss function by combining the sampling time and feature richness of the training data through the model building terminal, so that the model can more accurately reflect the current medical imaging information; collects and analyzes user feedback through the evaluation feedback terminal, and dynamically optimizes model performance, thereby significantly improving recognition accuracy and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate corresponding parts.

[0030] Figure 1 It is a schematic diagram of the overall module of the present invention.

[0031] Figure 2 This is a flow chart of the method for constructing a visual model of craniocerebral tumors based on network search and machine learning in the present invention.

[0032] Figure 3 Schematic diagram of the workflow of the model training module of the present invention. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with its 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; for those skilled in the art, other systems, methods and / or features of the present embodiment will become apparent after reviewing the following detailed description; it is intended that all such additional systems, methods, features and advantages are included in this specification; included within the scope of the present invention and protected by the appended claims; additional features of the disclosed embodiments are described in the following detailed description, and these features will be apparent from the following detailed description.

[0034] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right" and the like indicate directions or positional relationships based on the directions or positional relationships shown in the drawings, it is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or component referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limitations on this patent. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0035] Embodiment 1.

[0036] like Figure 1 As shown, this embodiment provides a system for building a visual model of craniocerebral tumors based on network search and machine learning, and the system includes a network data end, a model building end, and an evaluation feedback end;

[0037] The network data end is used to obtain relevant information data of craniocerebral tumors from the outside world; the model building end builds a visual model of craniocerebral tumors based on the relevant information data of craniocerebral tumors; the evaluation feedback end is used to evaluate the constructed visual craniocerebral tumor model and provide feedback information to the model building end to optimize the construction of the craniocerebral tumor model;

[0038] The network data terminal includes a data acquisition module, a data processing module, a data annotation module and a data storage module; the data acquisition module is used to connect with an external data source to obtain relevant information data of craniocerebral tumors, the data processing module is used to pre-process the data acquired by the data acquisition module, and the data annotation module is used to annotate the pre-processed data; the data storage module is used to store the pre-processed data information;

[0039] Furthermore, the external data sources connected to the data acquisition module include but are not limited to information systems of major hospitals and public medical databases; and the brain tumor related information data acquired by the data acquisition module is updated at any time according to the update of the external data source;

[0040] The model building end includes a model building module, a model training module, a model verification module and a three-dimensional visualization rendering module; the model building module is used to select and build a recognition model for craniocerebral tumor recognition; the model training module is used to perform training on the recognition model; the model verification module is used to evaluate the recognition model using an independent verification set to verify its performance on unseen data; the three-dimensional visualization rendering module generates a high-precision three-dimensional visualization model based on the output results of the recognition model, and the output results of the recognition model are various output features describing craniocerebral tumors;

[0041] Furthermore, the content annotated by the data annotation module includes timestamp information of the preprocessed data obtained from the external data source;

[0042] Furthermore, the model training module divides the preprocessed data into multiple batches according to the timestamps of the preprocessed data, each timestamp batch includes data uploaded to the external data source within the same time period, wherein the time period corresponding to each timestamp batch is unique, thereby ensuring the time seriality and organization of the preprocessed data;

[0043] Further, such as Figure 3 As shown, the model training module completes the training of the recognition model through the following steps:

[0044] S31: obtaining preprocessed data of multiple consecutive timestamp batches as a training set, assuming that the total number of batches is m, the sequence batch of the timestamp batch farthest from the current time is set to 1, and the sequence batch of the timestamp batch closest to the current time is set to m;

[0045] S32: Process all data in the training set to convert them into input features that can be used to identify model inputs;

[0046] S33: Input the training set into the recognition model in batches from small to large according to the sequence batches to train the recognition model, and calculate the loss function of each batch of recognition model output;

[0047] S34: After each batch of training is completed, the model parameters are updated by back propagation using the loss function to optimize the recognition model until all batches of training are completed;

[0048] Furthermore, during the training process of the model training module, the loss function Loss of each batch is set to:

[0049]

[0050] Among them, w i is the weight of the i-th output feature, which can be pre-set by the user based on prior knowledge; v is the sampling time of the timestamp batch and the weight of the feature richness; y i is the true value of the i-th output feature, is the model prediction value of the i-th output feature, γ1 and γ2 are regularization parameters used to control model complexity and prevent overfitting;

[0051] Furthermore, for v,

[0052]

[0053] Among them, β is the influence adjustment coefficient, which is used to adjust the relative importance of time factors and feature richness in the loss function; k is the sequence batch to which the timestamp batch belongs, l is the number of valid feature types in the preprocessed data of the timestamp batch, and n is the total number of feature types in all preprocessed data;

[0054] Furthermore, for l,

[0055]

[0056] Among them, l j is the number of valid selection items for the jth feature in the preprocessed data of the timestamp batch, and p is the total number of feature types of the timestamp batch; satisfying:

[0057]

[0058] Among them, k j is the total number of data in the jth feature in the preprocessed data of the timestamp batch, and ε is the preset effective selection ratio parameter;

[0059] Due to the complexity of brain tumor-related data and the continuous advancement of diagnosis and imaging technology, samples with more recent time and detailed data are usually more reliable. This solution inputs training samples in batches according to different timestamps during model training, and introduces sampling time and feature richness to influence weights during the model training of each batch. First, the recognition model obtains higher weights for newer data during training, ensuring that the recognition model can preferentially learn and adapt to the latest medical technologies and diagnostic schemes. Secondly, the introduced input feature richness makes feature-rich data samples contribute more to model training, thereby improving the recognition model's ability to learn detailed and complete data. By introducing a regularization coefficient, overfitting is effectively prevented and the performance of the recognition model on unseen data is enhanced. This can provide reliable and efficient diagnostic support in practical applications.

[0060] Embodiment 2:

[0061] This embodiment should be understood to include at least all the features of any of the above embodiments, and further improve upon them;

[0062] This embodiment provides a system for building a visual model of craniocerebral tumors based on network search and machine learning, the system comprising a network data end, a model building end, and an evaluation feedback end;

[0063] The network data end is used to obtain relevant information data of craniocerebral tumors from the outside world; the model building end builds a visual model of craniocerebral tumors based on the relevant information data of craniocerebral tumors; the evaluation feedback end is used to evaluate the constructed visual craniocerebral tumor model and provide feedback information to the model building end to optimize the construction of the craniocerebral tumor model;

[0064] The model building end includes a model building module, a model training module, a model verification module and a three-dimensional visualization rendering module; the model building module is used to select and build a recognition model for craniocerebral tumor recognition; the model training module is used to perform training on the recognition model; the model verification module is used to evaluate the recognition model using an independent verification set to verify its performance on unseen data; the three-dimensional visualization rendering module generates a high-precision three-dimensional visualization model based on the output results of the recognition model, and the output results of the recognition model are various output features describing craniocerebral tumors;

[0065] Furthermore, the evaluation feedback end includes a feedback collection module, a feedback identification module and a model adjustment module; the feedback collection module is used to collect direct feedback information from users, the feedback identification module is used to identify feedback feature information from the direct feedback information, and the model adjustment module optimizes and adjusts the recognition model based on the feedback feature information;

[0066] Furthermore, the three-dimensional visualization rendering module generates a high-precision three-dimensional visualization model based on the output results of the recognition model after training and verification; the generation process of the three-dimensional visualization model includes:

[0067] S41: receiving output features of the recognition model for the craniocerebral tumor, wherein the output features include but are not limited to features related to the location, size, shape, and color of the tumor in three-dimensional space;

[0068] S42: Building a three-dimensional visualization model of the tumor based on the output features, and performing texture mapping, coloring, and rendering. This process can use existing technologies such as point cloud reconstruction and ray tracing, which will not be described in detail here.

[0069] S43: exporting the established 3D visualization model to an interactive software interface, and providing interactive tools so that the user can rotate, zoom, and translate the 3D visualization model to better observe the details and overall structure of the tumor;

[0070] Furthermore, after the user obtains the three-dimensional visualization model, the feedback collection module provides an interactive interface to collect the user's direct feedback information on the three-dimensional visualization model, and the interactive interface has voice receiving and text writing functions, and converts the voice information input by the user into text information through voice recognition technology;

[0071] The feedback identification module uses natural language processing technology to analyze and identify feedback feature information, wherein the feedback feature information includes output features and sentiment tendencies related to the output features;

[0072] Furthermore, the model adjustment module first analyzes and quantifies the sentiment tendency types related to the output features through a pre-trained sentiment classifier, and the sentiment tendency types include positive, neutral and negative; and normalizes the output feature quantization data corresponding to the negative sentiment, and further optimizes and adjusts the loss function of the recognition model. The adjusted loss function Loss' satisfies:

[0073]

[0074] Among them, f i is the negative sentiment normalized data of the i-th output feature, satisfying 0≤f i ≤1, f i The larger the value, the more serious the negative emotion;

[0075] By combining user feedback information, while adjusting the recognition model according to the input information, combined with the output feedback information, the recognition model performance is further optimized, and user satisfaction is improved; specifically, this solution not only considers the time and feature richness of the input data, but also dynamically responds to user feedback on the model output, thereby ensuring that the direction and focus of the recognition model improvement are more in line with actual needs; this process effectively improves the accuracy and practicality of the recognition model, and at the same time improves the effect of the recognition model in practical applications;

[0076] like Figure 2 As shown, this embodiment also provides a method for constructing a visual model of craniocerebral tumors based on network search and machine learning, and the method comprises the following steps:

[0077] S1: Continuously obtain relevant information and data of brain tumors from external data sources;

[0078] S2: Construct a recognition model for brain tumors;

[0079] S3: Train and validate the recognition model for brain tumors;

[0080] S4: Establish a three-dimensional visualization model based on the output results of the recognition model;

[0081] S5: Adjust and optimize the recognition model based on user feedback.

[0082] The contents disclosed above are only preferred feasible embodiments of the present invention, and do not limit the protection scope of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated as technology develops.

Claims

1. A brain tumor visualization model construction system based on network search and machine learning, It is characterized in that The system includes a network data terminal, a model building terminal and an evaluation feedback terminal; The network data end is used to obtain relevant information data of craniocerebral tumors from the outside world; the model building end builds a visual model of craniocerebral tumors based on the relevant information data of craniocerebral tumors; the evaluation feedback end is used to evaluate the constructed visual craniocerebral tumor model and provide feedback information to the model building end to optimize the construction of the craniocerebral tumor model; The network data terminal includes a data acquisition module, a data processing module, a data annotation module and a data storage module; the data acquisition module is used to connect with an external data source to obtain relevant information data of craniocerebral tumors, the data processing module is used to pre-process the data acquired by the data acquisition module, and the data annotation module is used to annotate the pre-processed data; the data storage module is used to store the pre-processed data information; The model building end includes a model building module, a model training module, a model verification module and a three-dimensional visualization rendering module; the model building module is used to select and build a recognition model for craniocerebral tumor recognition; the model training module is used to perform training on the recognition model; the model verification module is used to evaluate the recognition model using an independent verification set to verify its performance on unseen data; the three-dimensional visualization rendering module generates a high-precision three-dimensional visualization model based on the output results of the recognition model, and the output results of the recognition model are various output features describing craniocerebral tumors; The evaluation feedback end includes a feedback collection module, a feedback identification module and a model adjustment module; the feedback collection module is used to collect direct feedback information from users, the feedback identification module is used to identify feedback feature information from the direct feedback information, and the model adjustment module optimizes and adjusts the recognition model based on the feedback feature information; The content annotated by the data annotation module includes the timestamp information of the pre-processed data obtained from the external data source; The model training module divides the preprocessed data into multiple batches according to the timestamps of the preprocessed data, each timestamp batch includes data uploaded to the external data source within the same time period, wherein the time period corresponding to each timestamp batch is unique; During the training process of the model training module, the loss function of each batch is Set to: ; in, For the The weight of each output feature can be pre-set by the user based on prior knowledge; The sampling time of the timestamp batch and the influence weight of feature richness; For the The true value of the output feature, For the The model predicts the value of the output feature, and is a regularization parameter used to control model complexity and prevent overfitting.

2. According to claim 1, a brain tumor visualization model construction system based on network search and machine learning, It is characterized in that The model training module completes the training of the recognition model through the following steps: S31: Get multiple consecutive timestamp batches of preprocessed data as training sets. Let the total number of batches be , the sequence batch of the timestamp batch farthest from the current time is set to 1, and the sequence batch of the timestamp batch closest to the current time is set to ; S32: Process all data in the training set to convert them into input features that can be used to identify model inputs; S33: Input the training set into the recognition model in batches from small to large according to the sequence batches to train the recognition model, and calculate the loss function of each batch of recognition model output; S34: After each batch of training is completed, the loss function is used to update the model parameters through back propagation to optimize the recognition model until all batches of training are completed.

3. A method for constructing a visual model of a cranial brain tumor based on network search and machine learning, applied to the system for constructing a visual model of a cranial brain tumor based on network search and machine learning as claimed in claim 2, It is characterized in that The method comprises the following steps: S1: Continuously obtain relevant information and data of brain tumors from external data sources; S2: Construct a recognition model for brain tumors; S3: Train and validate the recognition model for brain tumors; S4: Establish a three-dimensional visualization model based on the output results of the recognition model; S5: Adjust and optimize the recognition model based on user feedback.