Segmented prompt type boron neutron capture therapy target area intelligent sketching method

By adopting a segmented prompt-type target area intelligent outline method in boron neutron capture treatment, using medical imaging data and neural network technology, the accuracy and efficiency of traditional target area segmentation methods are solved, and more efficient target area is achieved Segmentation and therapeutic effects.

CN119028539AInactive Publication Date: 2024-11-26HUAPENG NEUTRON TECH (HANGZHOU) CO LTD
View PDF 10 Cites 0 Cited by

Patent Information

Application Number
CN202411507501.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-01-16
Filing Date
2024-10-28
Publication Date
2024-11-26
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In boron neutron capture treatment, the traditional target segmentation method relies on manual operations and is easily affected by subjective factors, resulting in inaccurate and stable segmentation results. In actual application, the traditional automated methods have a large deviation in the calculation results, which affects the treatment. accuracy and efficiency.

Method used

Using segmented tip-type target area intelligent outline method, a voxel model is established through medical image data, a neural network and machine learning algorithm are used to extract target area features, a target area intelligent outline neural network is built, and a target area automatic segmentation and classification are carried out. Improve the accuracy and efficiency of target area outlines.

Benefits of technology

It realizes rapid and precise segmentation of target areas, reduces artificial errors and subjective interference, improves the accuracy and safety of treatment, and provides patients with more effective treatment plans.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119028539A_ABST
    Figure CN119028539A_ABST
Patent Text Reader

Abstract

The invention discloses a segmented prompt type boron neutron capture therapy target region intelligent sketching method, which comprises the following steps of: reading medical image data, converting pixel information into three-dimensional voxel model information, and extracting and processing image features in a medical image; a target region intelligent delineation neural network is constructed, the target region intelligent delineation neural network takes a segmented prompt type method as a basic framework, and the structural information is utilized to guide the intelligent delineation process of the boron neutron capture therapy target region; extracting features of a target region from image features in the medical image, and training and optimizing the target region intelligent sketching neural network; and performing target region sketching on the new medical image by using the trained target region intelligent sketching neural network, and separating a target region from a non-target region. According to the method, the target tumor and the endangered organ region in the boron neutron capture therapy process can be quickly and accurately segmented, and cancer tissues needing to be emphatically killed and endangered organ tissues needing to be emphatically protected can be better marked.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a target area intelligent delineation method, in particular to a segmented prompting type boron neutron capture therapy target area intelligent delineation method. Background Art

[0002] The main treatments for cancer include: surgery, chemotherapy and radiotherapy. For patients with malignant tumors, about 70% of them need radiotherapy, and about 40% of malignant tumors can be cured by radiotherapy. After nearly a century of development of radiotherapy technology, scientists have continuously explored and developed more effective and less toxic innovative therapies to improve local tumor control rates, patient survival rates and quality of life. Boron Neutron Capture Therapy (BNCT), as a new generation of innovative radiotherapy methods, is a binary targeted tumor radiotherapy method that combines boron-containing targeted drugs and neutron irradiation. It can selectively kill cancer cells with complex shapes without damaging normal tissues. It is one of the most advanced cancer treatments in the world. In theory, it may be an ideal treatment for many types of cancer and is effective in treating advanced and recurrent cancers.

[0003] Boron neutron capture therapy is a type of radiotherapy used to treat certain types of cancer. In BNCT, doctors inject compounds containing boron into the patient's body and then expose the patient to neutron beams. The neutron beams react with boron atoms in the body to produce high-energy particles to kill cancer cells. BNCT is a targeted therapy that can selectively destroy cancer cells without affecting healthy tissue. In BNCT, it is very important to determine the target area. Because BNCT is a targeted therapy, only after the exact location and range of the cancer can the boron-containing compound be accurately injected and the neutron beam be accurately irradiated to the location of the cancer cells. If the target area is not accurately determined, the treatment effect may be reduced or even fail to achieve the expected therapeutic effect. In addition, determining the target area can also help doctors better assess the risks and potential side effects of treatment and develop more effective treatment plans. Therefore, in BNCT, it is very important to accurately determine the target area to provide patients with safer and more effective treatment.

[0004] In BNCT, the importance of target segmentation is also very high. Target segmentation refers to the further subdivision of the cancer site into smaller areas or parts in order to more accurately determine the injection location of the boron-containing compound and the irradiation location of the neutron beam. Through target segmentation, the accuracy of treatment can be improved to a higher level, which helps to maximize the destruction of cancer cells while maximizing the protection of surrounding healthy tissues. In addition, target segmentation can also help doctors better assess the risks and potential side effects of treatment and develop more effective treatment plans. Therefore, in BNCT, target segmentation is a very important step that can improve the accuracy and effectiveness of treatment and provide patients with safer and more effective treatment.

[0005] Intelligent target segmentation refers to the use of computer algorithms and artificial intelligence technology to analyze and process medical imaging data to achieve more accurate and automated target segmentation. In BNCT, the necessity of intelligent target segmentation is also very high. Traditional manual segmentation of the target area usually requires doctors to spend a lot of time and energy, and is easily affected by subjective factors, resulting in inaccurate and unstable segmentation results. Intelligent target segmentation can help doctors segment the target area more quickly and accurately, and can reduce human errors and interference from subjective factors, thereby improving the accuracy and effectiveness of treatment. In addition, intelligent target segmentation can also improve the standardization of target segmentation, making treatment results easier to compare and evaluate. Therefore, the necessity of intelligent target segmentation in BNCT is very high, which can improve the effectiveness and safety of treatment and provide patients with a better treatment experience.

[0006] Traditional radiation simulation human models are mostly composed of simple mathematical geometric models, such as isotropic cubes, cylinders and spheres. With the continuous optimization of anatomical models, the internal tissue structure of organisms has begun to be expressed by simple mathematical models through plane, cylinder, ellipse and sphere equations. However, in radiation simulation calculations, these simple structures often lead to deviations in calculation results, and the credibility is reduced. Traditional automated target delineation methods usually require preprocessing and manual marking, which limits their accuracy and efficiency in practical applications. Summary of the invention

[0007] In order to overcome the shortcomings of the prior art, the present invention proposes a segmented prompt type boron neutron capture therapy target area intelligent delineation method. After a voxel model of the human body is established based on medical imaging data, a series of characteristic information existing in the voxel model is used to segment the organ area or other areas of interest in the model. The segmentation information is used to obtain a target area information file such as an RT-Struct file (a file that stores medical image information) in a dicom file (Digital Imaging and Communications in Medicine, a type of medical digital imaging and communication standard) to provide the Monte Carlo program with the required geometric model, so as to complete the relevant calculation of radiation dose in the simulated human body model.

[0008] The technical solution of the present invention is achieved in this way:

[0009] A segmented prompt type boron neutron capture therapy target area intelligent delineation method, comprising the steps of:

[0010] S1, reads in medical image data, converts pixel information into three-dimensional voxel model information, and extracts and processes image features in medical images;

[0011] S2, constructing a target area intelligent delineation neural network, wherein the target area intelligent delineation neural network is based on a segmented prompting method and includes a branch structure of a neural differential equation, wherein the branch structure guides the intelligent delineation process of the target area of ​​the boron neutron capture therapy by using the structural information by learning the mapping relationship of the structural information;

[0012] S3, extracting the features of the target area from the image features in the medical image as the input of the target area intelligent delineation neural network, using a machine learning algorithm to classify or regress the target area, and training and tuning the target area intelligent delineation neural network;

[0013] S4, use the trained target area intelligent delineation neural network to delineate the target area of ​​the new medical image and separate the target area from the non-target area.

[0014] Furthermore, S1 includes the step of: preprocessing the medical image, wherein the preprocessing includes but is not limited to noise removal, filtering, correction, standardization, segmentation, enhancement and smoothing.

[0015] Furthermore, the medical imaging data in S1 includes but is not limited to CT images, MRI images and PET images.

[0016] Furthermore, the method of extracting image features from medical images in S1 includes but is not limited to edge detection and morphological operations.

[0017] Furthermore, the intelligent delineation process of the boron neutron capture therapy target area in S2 includes using a fast edge detection algorithm, an adaptive edge detection algorithm or a fast region growing algorithm to perform target delineation to improve the speed and accuracy of target delineation; and using parallel computing to accelerate the calculation process of the target delineation algorithm.

[0018] Furthermore, the characteristics of the target area in S3 include but are not limited to shape information, texture information and grayscale information.

[0019] Furthermore, the method of classifying or regressing the target area by the machine learning algorithm in S3 includes but is not limited to support vector machine, random forest and neural network.

[0020] Furthermore, the S4 also includes the step of post-processing the delineation result, including but not limited to removing noise and smoothing boundaries.

[0021] Furthermore, the algorithms for delineating the target area in S4 include but are not limited to a fast edge detection algorithm, an adaptive edge detection algorithm and a fast region growing algorithm.

[0022] Furthermore, S4 also includes the steps of: establishing a corresponding model library according to different types of medical images, using existing annotated data to train a target area intelligent delineation neural network, and realizing automated target area delineation.

[0023] The beneficial effect of the present invention is that, compared with the prior art, the present invention can quickly and accurately segment the target tumor and its endangered organ area during boron neutron capture therapy, and better mark the cancer tissue that needs to be killed and the endangered organ tissue that needs to be protected. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of an embodiment of a segmented prompt type boron neutron capture therapy target area intelligent delineation method of the present invention;

[0025] Figure 2 The present invention is a schematic diagram of a neural network structure for intelligent target area delineation in a segmented prompting type boron neutron capture therapy target area intelligent delineation method. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] See also Figure 1The object of the present invention is to provide a segmented prompt type boron neutron capture therapy target area intelligent delineation method, comprising the steps of:

[0028] S1, reads in medical image data, converts pixel information into three-dimensional voxel model information, and extracts and processes image features in medical images;

[0029] S2, constructing a target area intelligent delineation neural network, wherein the target area intelligent delineation neural network is based on a segmented prompting method and includes a branch structure of a neural differential equation, wherein the branch structure guides the intelligent delineation process of the target area of ​​the boron neutron capture therapy by using the structural information by learning the mapping relationship of the structural information;

[0030] S3, extracting the features of the target area from the image features in the medical image as the input of the target area intelligent delineation neural network, using a machine learning algorithm to classify or regress the target area, and training and tuning the target area intelligent delineation neural network;

[0031] S4, use the trained target area intelligent delineation neural network to delineate the target area of ​​the new medical image and separate the target area from the non-target area.

[0032] In step S1, the medical image data, including medical information such as Hounsfield unit (HU), is input into a computer, and a specific algorithm and model are used to convert pixel information into three-dimensional voxel model information. In order to facilitate later computer processing, it is generally necessary to pre-process the medical image, and the pre-processing includes but is not limited to noise removal, filtering, enhancement and smoothing.

[0033] Noise and distortion often exist in medical image data, which may have a negative impact on the accuracy of automatic target delineation. Therefore, before automatic delineation, some preprocessing steps are required for medical image data to reduce the impact of noise and distortion, thereby improving the accuracy of automatic delineation. For example: filtering, use appropriate filters to filter medical images to remove noise and blur; enhancement, use image enhancement algorithms to enhance medical images to enhance the edge and texture information of the target area; correction, for distorted medical images, image correction processing can be performed, such as distortion correction and artifact removal; standardization, before automatic delineation, medical images need to be standardized to the same grayscale value range to eliminate differences between different images; segmentation, before automatic delineation, image segmentation algorithms can be used to segment medical images into different areas to reduce the impact of noise and background, so as to better identify and delineate the target area.

[0034] In step S2, an efficient target delineation algorithm can be selected, such as a fast edge detection algorithm, an adaptive edge detection algorithm, a fast region growing algorithm, etc., which can significantly improve the speed and accuracy of target delineation. Adjust algorithm parameters: According to specific application scenarios, adjust the parameters of the target delineation algorithm, such as the threshold of the edge detection algorithm, the growth condition of the region growing algorithm, etc., to further optimize the speed and accuracy of the algorithm. Use parallel computing: Using parallel computing technology, such as GPU (Graphic Processing Unit) acceleration, distributed computing, etc., can accelerate the calculation process of the target delineation algorithm, thereby improving speed and accuracy.

[0035] The target area intelligent delineation neural network structure of the present invention is as follows Figure 2 As shown, the target intelligent delineation neural network takes the segmented prompt method as the basic framework of BNCT target delineation, designs a branch structure based on neural differential equations in the segmented prompt method, learns the mapping relationship of structural information in the production process through branches, uses structural information to guide the process of BNCT target intelligent delineation, and improves the network residual link structure based on neural differential equations to enhance the adaptability of the feature network to diversified features, and ultimately improves the accuracy of BNCT target intelligent delineation, thereby avoiding a few tumor areas being forgotten without adequate dose distribution.

[0036] In the process of target area delineation, machine learning and image segmentation technology can be used in combination to achieve automatic target area delineation. Specifically, the following steps can be adopted: data preprocessing: first, the medical image needs to be preprocessed, such as filtering, denoising and enhancement, to improve image quality and reduce interference; feature extraction: next, the target area features, such as shape, texture and grayscale information, need to be extracted from the medical image as input to the machine learning model; model training: then, a machine learning algorithm (such as support vector machine, random forest or neural network) is used to classify or regress the target area, and the model is trained and tuned; image segmentation: the trained model is used to segment the new medical image to separate the target area from the non-target area; post-processing: finally, the segmentation results can be post-processed, such as removing noise and smoothing boundaries, to obtain more accurate and stable target area delineation results.

[0037] In order to improve the application of automated target delineation technology in different types of medical images, more powerful and rich model training is needed. At the same time, in order to speed up image segmentation and adapt to different image types, it is necessary to first understand the type of information data contained in different medical image types, and then select appropriate target delineation algorithms and parameters for different types of medical images (such as CT (Computed Tomography, electronic computed tomography) images, MRI (magnetic resonance imaging) images, PET (Positron Emission Tomography, positron emission tomography) images, etc.) to adapt to the characteristics of different image types.

[0038] Considering the differences between different medical images and image quality, such as the differences in the quality of different medical images, such as noise, blur, etc., corresponding preprocessing methods are needed to improve the accuracy and stability of target delineation. These information features can be used to establish a corresponding model library, such as the more typical head-neck model, chest-body model, etc. For different types of medical images, a corresponding model library can be established, and the model can be trained using existing annotated data to achieve automated target delineation.

[0039] Since some Monte Carlo post-processing programs that require automatic target area delineation systems need to have a certain delineation sequence, the present invention is used in combination with the corresponding model library to establish it, and at the same time, a dicom file sequence that can meet the geometric processing of the MC program can be generated to meet the requirements of the post-processing program for generating corresponding geometry and corresponding voxel grid models.

[0040] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A segmented prompt type boron neutron capture therapy target area intelligent delineation method, characterized in that: Includes steps: S1, reads in medical image data, converts pixel information into three-dimensional voxel model information, and extracts and processes image features in medical images; S2, constructing a target area intelligent delineation neural network, wherein the target area intelligent delineation neural network is based on a segmented prompting method and includes a branch structure of a neural differential equation, wherein the branch structure guides the intelligent delineation process of the target area of ​​the boron neutron capture therapy by using the structural information by learning the mapping relationship of the structural information; S3, extracting the features of the target area from the image features in the medical image as the input of the target area intelligent delineation neural network, using a machine learning algorithm to classify or regress the target area, and training and tuning the target area intelligent delineation neural network; S4, use the trained target area intelligent delineation neural network to delineate the target area of ​​the new medical image and separate the target area from the non-target area.

2. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The S1 comprises the steps of: preprocessing the medical image, wherein the preprocessing includes but is not limited to noise removal, filtering, correction, standardization, segmentation, enhancement and smoothing.

3. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The medical imaging data in S1 include but are not limited to CT images, MRI images and PET images.

4. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The method for extracting image features in medical images in S1 includes but is not limited to edge detection and morphological operation.

5. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The intelligent delineation process of the target area of ​​the boron neutron capture therapy described in S2 includes using a fast edge detection algorithm, an adaptive edge detection algorithm or a fast region growing algorithm to perform target delineation to improve the speed and accuracy of target delineation; and using parallel computing to accelerate the calculation process of the target delineation algorithm.

6. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The characteristics of the target area in S3 include but are not limited to shape information, texture information and grayscale information.

7. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The method for classifying or regressing the target area using the machine learning algorithm in S3 includes but is not limited to support vector machine, random forest and neural network.

8. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The S4 also includes the step of post-processing the delineation result, including but not limited to removing noise and smoothing boundaries.

9. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The algorithms for target area delineation in S4 include but are not limited to a fast edge detection algorithm, an adaptive edge detection algorithm and a fast region growing algorithm.

10. The segmented prompt type boron neutron capture therapy target area intelligent delineation method according to claim 1, characterized in that: The S4 also includes the steps of: establishing a corresponding model library according to different types of medical images, using existing annotated data to train a target area intelligent delineation neural network, and realizing automated target area delineation.

Citation Information

Patent Citations

  • Intelligent and automatic delineation method for gross tumor volume and organs at risk

    CN107403201A

  • Tumor target area delineation method and device based on variable human body model and storage medium

    CN110322426A

  • Organ sketching method and device

    CN111738989A

  • Image detection method, system and related device for medical image

    CN116863246A

  • Identification method and identification equipment for endoscopic submucosal dissection operation stage

    CN117037043A