Method, equipment and program product for preventing brain radiation injury

Abstract

The invention relates to the field of intelligent medical treatment, in particular to a method, equipment and a program product for preventing brain radiation injury. Comprising the following steps: S1, acquiring brain image data of a to-be-tested person; s2, performing region sketching on the brain image data to obtain a lateral ventricular wall region; and S3, giving an exposure dose lower than a preset exposure dose to the lateral ventricular wall area. The clinical work intensity of doctors and physicians in the radiotherapy department is greatly reduced while the life quality of the patient is remarkably improved, and the clinical value is very good.

Description

Technical Field

[0001] The present application relates to the field of intelligent medical treatment, and specifically to a method, device, program product and computer-readable storage medium for preventing brain radiation damage. Background Art

[0002] Radiation brain injury refers to a complication of brain tissue damage or dysfunction caused by radiotherapy, including radiation necrosis and pseudoprogression. Among them, brain radiation necrosis (RN) is a common clinical problem encountered by glioma surgeons and radiotherapists. There are two difficulties: first, RN is difficult to distinguish from glioma recurrence; second, once RN occurs, it will continue to deteriorate and there is no effective treatment. Therefore, RN focuses on prevention. At present, the identification of RN and tumor recurrence has been studied in the field of neuroimaging for many years, but there are few reports on systematic studies on the characteristics of RN occurrence and preventive intervention at home and abroad. There are nearly 28,000 new patients with long-term survival glioma in my country every year. Because long-term survival gliomas are in the brain and are more common in young and middle-aged people, the complications such as hemiplegia and aphasia caused by them bring heavy burdens to families and society. Long-term survival gliomas have the best prognosis among gliomas. This type of tumor is sensitive to radiotherapy, and postoperative radiotherapy can significantly prolong the patient's survival. The median survival of patients with grade 3 postoperative radiotherapy can reach 14.7 years. Long-term survival requires a higher quality of life.

[0003] There are two main factors that lead to a decline in the quality of life of patients: one is surgical damage to the functional area; the other is radiation necrosis (RN) caused by radiotherapy involving the functional area. In recent years, auxiliary technologies such as intraoperative magnetic resonance imaging, fluorescence tracking imaging, and intraoperative ultrasound have been widely used in neurosurgery, which have maximized the removal of tumors under safe conditions. Surgical resection of gliomas has reached a bottleneck period. Under the existing technology, it is difficult to achieve significant improvements in both the degree of resection and the protective function. RN caused by radiotherapy is another important reason for the decline in the quality of life of patients. Although long-term gliomas are sensitive to radiotherapy, 20-30% of patients will develop RN after radiotherapy, especially RN involving the functional area, which leads to complications such as hemiplegia, hemianopsia, aphasia, and urinary incontinence, which seriously affects the quality of life of patients. Once RN occurs, there is currently no effective clinical treatment. Mannitol, hormones, hyperbaric oxygen, and bevacizumab can only temporarily reduce the edema caused by RN, and necrosis and symptoms will continue to progress repeatedly. Therefore, RN focuses on prevention, but there are few reports at home and abroad on the occurrence, development characteristics, intervention and prevention strategies of RN. Summary of the invention

[0004] In view of the above problems, the present invention discovers the susceptible area of ​​radiation brain damage in the lateral ventricle wall through patient radiotherapy and follow-up MRI data, and therefore innovatively proposes to develop a lateral ventricle wall radiotherapy protection technology system by combining radiotherapy technology and artificial intelligence, thereby reducing the risk of lateral ventricle wall damage and improving the quality of life of patients in the long survival period. Specifically, a method for preventing brain radiation damage is provided, including: S1. Obtaining brain imaging data of the subject; S2, performing regional delineation on the brain image data to obtain a lateral ventricle wall region; S3. Give the lateral ventricular wall area a dose lower than the preset dose.

[0005] The S2 is replaced by S2' to delineate the brain imaging area to obtain the tumor target area and the lateral ventricle wall area, and the S3 is replaced by S3'. When the lateral ventricle wall area intersects with the tumor target area, the lateral ventricle wall area within the intersection area is given a first irradiation dose lower than a preset dose; and the lateral ventricle wall area outside the intersection area is given a preset second irradiation dose.

[0006] Optionally, the first irradiation dose is a dose hotspot of a tumor target area, and the second irradiation dose is calculated in the same manner as a preset irradiation dose.

[0007] Optionally, S3' further includes reducing the irradiated volume of the lateral ventricle wall area outside the intersection area.

[0008] The preset exposure dose or the preset second exposure dose is an exposure dose obtained by calculating the relationship between brain radiation necrosis occurring in the lateral ventricle wall area and different exposure doses.

[0009] Optionally, the calculation method of the preset exposure dose or the preset second exposure dose includes: Obtain brain images of patients with brain radiation necrosis; Delineating the brain image to obtain the lateral ventricle wall area; Calculate the maximum dose or average dose in the area of ​​brain radiation necrosis in the lateral ventricular wall area; The relationship between brain radiation necrosis and irradiation dose can be obtained by fitting the mathematical relationship based on the maximum dose or average dose; A preset exposure dose or a preset second exposure dose is obtained based on the relationship between brain radiation necrosis and the exposure dose.

[0010] Optionally, the brain image in S1 is a fusion image of CT or MRI.

[0011] The delineation of the lateral ventricle wall area is obtained by automatically delineating the target area on the brain image through a network model.

[0012] Optionally, the network model includes one or more of the following: CNN, mask RNN, Faster R-CNN, FCN, SegNet, PSPNet, DeepLab v1 / v2 / v3.

[0013] Optionally, the training process of the network model is: Obtain patient brain image sets and labels; The brain image set is input into the network model to be trained for model training to obtain the network model.

[0014] The preset exposure dose is replaced by a preset total exposure dose.

[0015] Optionally, the preset second exposure dose is replaced by a preset total exposure dose.

[0016] Optionally, the calculation process of the total irradiated dose is: Obtain the brain image of the subject and the dose of each minimum beam to a single voxel; Performing regional delineation on the brain image data to obtain a lateral ventricle wall region sequence; The dose of each minimum beam to a single voxel in the lateral ventricle wall region sequence was input into the topological network structure for calculation, and the total dose of the lateral ventricle wall region was obtained.

[0017] Optionally, the dose of each minimum beam to a single voxel is calculated through a topological network structure to obtain the total dose of a single voxel point in the lateral ventricle wall region sequence, and the total dose of the single voxel point in the lateral ventricle wall region is summed to obtain the total dose of the lateral ventricle wall region.

[0018] Optionally, the total irradiation dose of the lateral ventricle wall region is obtained by weighted averaging the total irradiation doses of a single voxel point.

[0019] The preset exposure dose is replaced by a preset exposure dose distribution map.

[0020] Optionally, the preset second exposure dose is replaced by a preset exposure dose distribution map.

[0021] Optionally, the calculation process of the irradiated dose distribution map is: Obtaining brain images of the subject; Performing regional delineation on the brain image data to obtain a tumor target area and a lateral ventricle wall area; Calculate the radiation intensity of the lateral ventricle wall region sequence and the target region sequence to obtain the radiation intensity map of the lateral ventricle wall region and the radiation intensity map of the target region; The lateral ventricle wall region sequence, the target region sequence, the lateral ventricle wall region exposure intensity map, and the target region exposure intensity map are input into the neural network to obtain the lateral ventricle wall region sequence exposure dose distribution map and the target region sequence exposure dose distribution map.

[0022] Optionally, the Shenjiang network is a neural network obtained by training by acquiring a lateral ventricle wall region sequence, a target region sequence, a lateral ventricle wall region illumination intensity map, and a target region illumination intensity map for feature learning.

[0023] The object of the present invention is to provide a computer program product, comprising a computer program or instructions, comprising: The computer program or instructions are executed by a processor to implement the above-mentioned method for preventing brain radiation damage.

[0024] The object of the present invention is to provide a computer device, comprising a memory, a processor and a computer program or instructions stored in the memory, including: the computer program or instructions are executed by the processor to implement the above-mentioned method for preventing brain radiation damage.

[0025] The object of the present invention is to provide a computer-readable storage medium having a computer program or instruction stored thereon, including: the computer program or instruction is executed by a processor to implement the above-mentioned method for preventing brain radiation damage.

[0026] The object of the present invention is to provide a brain intelligent radiotherapy method, comprising: Obtain the brain image of the subject and the dose of each minimum beam to a single voxel; The dose distribution map of the target area sequence, lateral ventricle wall area sequence or other involved organ area sequence is obtained according to the calculation of the irradiated dose distribution; The dose distribution maps of the target area sequence, lateral ventricle wall area sequence or other involved organ area sequence are input into the target model to obtain the predicted results of the irradiated dose of the target area, lateral ventricle wall area or other involved organ area.

[0027] Optionally, the target model is a topological network structure model, and the construction process of the topological network structure is: obtain the dose of each minimum beam to a single voxel, and based on the topological structure, construct a topological network of the dose distribution map of each minimum beam to a single voxel, the target area sequence, the lateral ventricle wall area sequence or other involved organ area sequence to obtain a topological network structure model.

[0028] Optionally, the target model also includes parameter optimization, and the optimization parameters are obtained through the dose distribution map. The optimization parameters are compared with the prediction results to obtain the loss function, and the parameters of the target model are adjusted based on the loss function. The training is repeated until the loss function remains unchanged to obtain the target model.

[0029] Optionally, the parameter optimization further includes introducing a non-differentiable regularization term, and removing interference information of the sequence or distribution graph through the regularization term.

[0030] The object of the present invention is to provide a computer program product, comprising a computer program or instructions, comprising: The computer program or instructions are executed by the processor to implement the above-mentioned intelligent brain radiotherapy method.

[0031] The object of the present invention is to provide a computer device, including a memory, a processor and a computer program or instructions stored in the memory, including: the computer program or instructions are executed by the processor to implement the above-mentioned intelligent brain radiotherapy method.

[0032] The object of the present invention is to provide a computer-readable storage medium having a computer program or instruction stored thereon, including: the computer program or instruction is executed by a processor to implement the above-mentioned intelligent brain radiotherapy method.

[0033] Advantages of the present invention: 1. In response to the problem of brain radiation damage, the present invention found that the lateral ventricular wall (LVW) is a high-incidence area of ​​radiation damage. The present invention innovatively delineates and limits the protection of LVW as an organ at risk (OAR) during radiotherapy. Under the premise of not reducing the dose of the tumor target area, the LVW dose is reduced, the occurrence of RN is reduced, and the long-term quality of life is ultimately improved. The specific LVW protection principle is: LVW within the target area avoids high-dose hot spots to the greatest extent; LVW outside the target area minimizes the irradiated volume and / or dose.

[0034] 2. The present invention obtains the optimal limiting dose for RN occurrence by fitting the relationship between LVW occurrence RN and dose received (dose-response curve).

[0035] 3. The present invention proposes a dose calculation model based on deep learning, which calculates the dose intensity map of each part according to the LVW and target area information, and then calculates the dose distribution map based on the dose intensity map. It can intuitively observe the area and amount of radiotherapy dose, providing a basis for subsequent radiotherapy protection.

[0036] 4. The present invention proposes a network topology structure to calculate the radiation dose of a single voxel in an image, and can predict the radiation dose based on the structure, and can predict the radiation dose of the radiotherapy area, providing basic conditions for subsequent automatic radiotherapy protection.

[0037] 5. The present invention proposes an intelligent automatic radiotherapy protection method, which calculates the dose distribution through artificial intelligence, and then performs target modeling and prediction based on the dose distribution results, thereby realizing the correlation coupling and feedback adjustment between prediction and optimization, and finally building artificial intelligence automated radiotherapy. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0039] Figure 1 A schematic flow chart of a method for preventing brain radiation damage provided by an embodiment of the present invention; Figure 2 A schematic diagram of a system for preventing brain radiation damage provided by an embodiment of the present invention; Figure 3 A schematic diagram of a device for preventing brain radiation damage provided by an embodiment of the present invention; Figure 4 A patient profile for a RN study on glioma-related long-term survival in the radiotherapy department of Tiantan Hospital provided by an embodiment of the present invention; Figure 5 LVW target area delineation provided by an embodiment of the present invention; Figure 6 The embodiment of the present invention provides a retrospective assessment of the relationship between LVW radiation necrosis and dose; Figure 7 The related indicators and relationships of brain area radiation provided by the embodiment of the present invention are: A is a nomogram of each characteristic indicator; B is the relationship between radiation damage and age; C is the relationship between radiation loss and radiation dose; D is the relationship between radiation loss and irradiated volume; E is the location where radiation damage occurs; Figure 8 The embodiment of the present invention provides the changes of patients after receiving brain radiation; A is the number of patients with radiation damage in ALL, A2, O2, and O3 after the radiation time increases; B is the cumulative percentage of ALL, A2, O2, and O3 after the radiation time increases; C is the ratio of ALL, A2, O2, and O3 in radionecrosis and pseudoprogression; D is the ratio of ALL, A2, O2, and O3 in different areas, ALL refers to all patients, A2 is a secondary astrocytoma, O2 is a secondary long-survival glioma, and O3 is a tertiary long-survival glioma; Fig. 9 Comparison between the LVW protection radiotherapy provided by the embodiment of the present invention and conventional radiotherapy shows that the LVW protection radiotherapy plan (B) significantly reduces the LVW dose compared with the conventional radiotherapy plan (A) (red line: irradiated target area; color heat map: high-dose area >62Gy; dotted box: LVW high-dose area distribution; Fig.10The relative radiation doses of the conventional LVM and the protective LVM and the relative radiation doses of the protective PTV and the conventional PTV provided in the embodiments of the present invention, the red line is the dose received by the tumor, it can be seen that the tumor dose of the protection plan is not affected (the two curves are inseparable), the yellow line is the dose received by the lateral ventricle wall, it can be seen that the protection plan is lower than the conventional plan (the two curves can be separated), among which the relative doses of the protective LVW and the protective PTV are low. DETAILED DESCRIPTION

[0040] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0041] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The sequence numbers of the operations, such as S101, S102, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0042] Figure 1 A schematic diagram of a method for preventing brain radiation damage provided by an embodiment of the present invention specifically includes: S1: Obtain brain imaging data of the subject; In one embodiment, the imaging data includes one or more of the following: CT, MRI. In a specific embodiment, the present invention collected clinical data and follow-up information of 354 patients with long-term survival of glioma who underwent radiotherapy at Tiantan Hospital from April 2019 to June 2022 (enhanced MRI was reviewed at 1 month, 4 months, and 10 months after radiotherapy, and at least once a year thereafter). Figure 4As shown, including general demographic information, symptoms, tumor size and location, surgical conditions, histopathology, molecular pathology, radiotherapy target dose, chemotherapy regimen, other treatment conditions, and follow-up MRI information. The risk factors, location characteristics, and occurrence time of RN in LVW were comprehensively statistically analyzed. Finally, the RN-prone population, time of occurrence, location of LVW, and the relationship with radiotherapy dose and target area were obtained. The analysis found that the RN incidence rate was as high as 29.7% (105 / 354). In addition, through the above data analysis, it was found that: 1) The incidence of RN in long-term survival gliomas is high (29.7%); 2) Most RNs occur in LVW (76%); 3) Symptomatic RN of hemiplegia and aphasia is mostly caused by RN in LVW involving functional areas; 4) More than 90% of recurrences in patients with long-term survival gliomas occur around the tumor cavity, and rarely occur in LVW far away from the tumor.

[0043] S2: performing regional delineation on the brain image data to obtain a lateral ventricle wall region; In one embodiment, the delineation of the lateral ventricle wall region is obtained by automatically delineating the target region on the brain image using a network model.

[0044] In one embodiment, the network model includes one or more of the following: CNN, mask RNN, FasterR-CNN, FCN, SegNet, PSPNet, DeepLab v1 / v2 / v3.

[0045] In one embodiment, the training process of the network model is: Obtain patient brain image sets and labels; Delineating the lateral ventricle wall region of the image set to obtain regional features; The regional features are input into the network model to be trained to perform regional feature learning to obtain the network model.

[0046] In a specific embodiment, the LVW automatic delineation technology is established: the radiotherapy positioning MRI fusion conventional sequences (T1, T2, FLAIR and T1 enhancement) of 300 patients with long-term survival gliomas are introduced into the target area delineation system (Eclipse planning system). The LVW is surrounded by cerebrospinal fluid signals, especially on the T2 sequence, the difference between LVW low signal (gray) and cerebrospinal fluid signal (high signal - white) is obvious, and the LVW automatic delineation (wall thickness 3-5mm) can be realized. The best MRI sequence for delineating the lateral ventricle wall is determined, and the LVW automatic delineation technology is realized in the Eclipse target area delineation system through deep learning, training and verification of the data of 300 patients with long-term survival gliomas.

[0047] In a specific embodiment, the target area is delineated on the image data, and the delineation result is as follows: Figure 5As shown in the figure, there is a significant difference in signal between LVW and surrounding cerebrospinal fluid on CT and MRI images. On CT, LVW is medium-density gray, while cerebrospinal fluid is low-density black; on MRI sequences, LVW is medium-signal gray, while cerebrospinal fluid is low-signal black on T1 and high-signal white on T2.

[0048] S3: The lateral ventricular wall area is given a dose lower than the preset dose.

[0049] In one embodiment, the preset exposure dose is a limited exposure dose obtained by calculating the relationship between brain radiation necrosis occurring in the lateral ventricle wall region and different exposure doses.

[0050] In one embodiment, the method for calculating the preset exposure dose includes: Obtain brain images of patients with brain radiation necrosis; Delineating the brain image to obtain the lateral ventricle wall area; Calculate the maximum dose or average dose in the area of ​​brain radiation necrosis in the lateral ventricular wall area; The relationship between brain radiation necrosis and the dose was obtained by fitting the mathematical relationship based on the maximum dose or the average dose; The preset exposure dose is obtained based on the relationship between brain radiation necrosis and exposure dose.

[0051] In one embodiment, S2 is replaced by S2' to delineate the brain imaging area to obtain the tumor target area and the lateral ventricle wall area, and S3 is replaced by S3'. When the lateral ventricle wall area intersects with the tumor target area, the lateral ventricle wall area within the intersection area is given a first irradiation dose lower than a preset dose; and the lateral ventricle wall area outside the intersection area is given a preset second irradiation dose.

[0052] In one embodiment, the first irradiation dose is a dose hotspot of a tumor target area, and the second irradiation dose is calculated in the same manner as a preset irradiation dose.

[0053] In one embodiment, the method for calculating the preset second exposure dose includes: Obtain brain images of patients with brain radiation necrosis; Delineating the brain image to obtain the lateral ventricle wall area; Calculate the maximum dose or average dose in the area of ​​brain radiation necrosis in the lateral ventricular wall area; The relationship between brain radiation necrosis and the dose was obtained by fitting the mathematical relationship based on the maximum dose or the average dose; The preset second irradiation dose is obtained based on the relationship between brain radiation necrosis and the irradiation dose.

[0054] In one embodiment, the lateral ventricle wall region outside the target area in S3' also includes a reduced irradiated volume.

[0055] In one embodiment, the brain image is replaced with a fused image of CT or MRI, and the fused image is delineated to obtain the lateral ventricle wall region.

[0056] In one embodiment, the preset exposure dose is replaced by a preset total exposure dose.

[0057] In one embodiment, the preset second exposure dose is replaced by a preset total exposure dose.

[0058] In one embodiment, the total irradiated dose is calculated as follows: Obtain the brain image of the subject and the dose of each minimum beam to a single voxel; Performing regional delineation on the brain image data to obtain a lateral ventricle wall region sequence; The dose of each minimum beam to a single voxel in the lateral ventricle wall region sequence was input into the topological network structure for calculation, and the total dose of the lateral ventricle wall region was obtained.

[0059] Optionally, the dose of each minimum beam to a single voxel is calculated through a topological network structure to obtain the total dose of a single voxel point in the lateral ventricle wall region sequence, and the total dose of the single voxel point in the lateral ventricle wall region is summed to obtain the total dose of the lateral ventricle wall region.

[0060] Optionally, the total irradiation dose of the lateral ventricle wall region is obtained by weighted averaging the total irradiation doses of a single voxel point.

[0061] In another embodiment, the preset exposure dose is replaced by a preset exposure dose distribution map.

[0062] In another embodiment, the preset second exposure dose is replaced by a preset exposure dose distribution map.

[0063] In another embodiment, the calculation process of the irradiated dose distribution map is: Obtaining brain images of the subject; Performing regional delineation on the brain image data to obtain a tumor target area and a lateral ventricle wall area; Calculate the radiation intensity of the lateral ventricle wall region sequence and the target region sequence to obtain the radiation intensity map of the lateral ventricle wall region and the radiation intensity map of the target region; Inputting the lateral ventricle wall region sequence, the target region sequence, the lateral ventricle wall region radiation intensity map, and the target region radiation intensity map into the neural network to obtain the lateral ventricle wall region sequence radiation dose distribution map and the target region sequence radiation dose distribution map; Optionally, the Shenjiang network is a neural network obtained by training by acquiring a lateral ventricle wall region sequence, a target region sequence, a lateral ventricle wall region illumination intensity map, and a target region illumination intensity map for feature learning.

[0064] In a specific embodiment, LVW is used as an organ at risk for target delineation and dose limitation, and the LVW exposure dose is reduced to the maximum extent without reducing the radiotherapy dose of the tumor target area. Specific LVW protection principles: LVW within the target area should avoid high-dose hot spots to the maximum extent; LVW outside the target area should minimize the exposure volume and dose.

[0065] In a specific embodiment, the data of 76 long-term glioma patients with RN in the lateral ventricle wall that have been collected were used for analysis. First, the RN area was outlined on the CT / MRI fusion images of all patients before radiotherapy (compared with the MRI images of the first RN); second, the radiotherapy dose of the RN area of ​​each patient was fully calculated in the original treatment planning system; finally, the relationship between the occurrence of RN in the LVW and the dose (dose-response curve) was analyzed to find the optimal limiting dose for the occurrence of RN.

[0066] In a specific embodiment, the RN area was manually delineated on the CT / MRI fusion images of 76 LVW patients with RN (compared with the MRI images of the first occurrence of RN), such as Figure 6 As shown in the figure, by comprehensively calculating the overall radiation dose (including the maximum dose and the average dose) of the RN area in the LVW of each patient in the planning system that has completed the treatment, and drawing a graph of the relationship between RN and radiation dose (dose-response curve), the optimal limiting dose for intervention in RN is obtained.

[0067] In one embodiment, the present invention performs statistical analysis of clinical data on patients after radiation exposure, such as Figure 7 As shown, age, damage location, radiation dose, and irradiated volume are statistically significant. With the increase of age, radiation dose, and irradiated volume, the rate of damage increases. The present invention analyzes ALL, A2, O2, and O3 ( Figure 8 ) After receiving radiation, with the increase of time, the number and proportion of patients suffering from radiation damage (with the increase of time, the number first increases and then decreases, and the cumulative ratio increases with the increase of time and then remains unchanged), as well as the proportion of radiation necrosis and pseudoprogression, and the regional proportion of damage in different areas. Most of the injuries occurred within 3 years, and the largest proportion occurred in LVW.

[0068] In one embodiment, the innovation of the present invention is that LVW, which is most prone to RN, is protected as an organ at risk. This is an innovative RN intervention strategy that has not been reported domestically or internationally. In addition, with the advantage of the discipline, a large number of case data are used to explore the best and executable LVW protection plan. The invention cleverly uses the strong difference between LVW and surrounding cerebrospinal fluid signals on CT and MRI, and establishes LVW protection intelligent radiotherapy plan design technology through deep learning and training verification, which greatly reduces labor costs while reducing RN and improving patients' quality of life.

[0069] The disclosed embodiments of the present invention further provide a computer program product or system, including a computer program, which implements the above-mentioned method steps for preventing brain radiation damage when executed by a processor.

[0070] Figure 2 A schematic diagram of a system for preventing brain radiation damage provided by an embodiment of the present invention specifically includes: Acquisition unit: acquires the brain image data of the subject; A delineation unit: delineating the brain image data to obtain a lateral ventricle wall region; Dose unit: Give the lateral ventricular wall area a dose lower than the preset dose.

[0071] Figure 3 A schematic diagram of a device for preventing brain radiation damage provided by an embodiment of the present invention specifically includes: A memory and a processor; the memory is used to store program instructions; the processor is used to call the program instructions, and when the program instructions are executed, any one of the above-mentioned methods for preventing brain radiation damage is performed.

[0072] The disclosed embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, any one of the above-mentioned methods for preventing brain radiation damage is performed.

[0073] An embodiment of the present invention provides a brain intelligent radiotherapy method, comprising: Obtain the brain image of the subject and the dose of each minimum beam to a single voxel; The dose distribution map of the target area sequence, lateral ventricle wall area sequence or other involved organ area sequence is obtained according to the calculation of the irradiated dose distribution; The dose distribution maps of the target area sequence, lateral ventricle wall area sequence or other involved organ area sequence are input into the target model to obtain the predicted results of the irradiated dose of the target area, lateral ventricle wall area or other involved organ area.

[0074] In one embodiment, the target model is a topological network structure model, and the construction process of the topological network structure is: obtaining the dose of each minimum beam to a single voxel, and based on the topological structure, constructing a topological network of the dose distribution map of each minimum beam to a single voxel, the target area sequence, the lateral ventricle wall area sequence or other involved organ area sequence to obtain a topological network structure model.

[0075] In one embodiment, the target model also includes parameter optimization, and the optimization parameters are obtained through the dose distribution map. The loss function is obtained by comparing the optimization parameters with the prediction results. The parameters of the target model are adjusted based on the loss function, and the training is repeated until the loss function remains unchanged to obtain the target model.

[0076] Optionally, the parameter optimization further includes introducing a non-differentiable regularization term, and removing interference information of the sequence or distribution graph through the regularization term.

[0077] In a specific embodiment, an intelligent radiotherapy plan for LVW protection is developed, and an executable radiotherapy plan is directly obtained according to the region of interest (ROI, LVW) and the prescription dose. Artificial intelligence and modeling optimization are used in combination, and the optimal dose distribution is predicted by artificial intelligence, and then modeling optimization is carried out with the predicted results as the target, so as to realize the correlation coupling and feedback adjustment between prediction and optimization, and finally build an artificial intelligence automated radiotherapy plan design. It mainly includes: 1) An optimization model based on a topological network structure, the input is the dose of each minimum beam to a single voxel, and the output is the total dose of the voxel point. The topological transformation converts the treatment plan optimization problem into a neural network training process for dose prediction; 2) Based on a deep learning dose calculation model, it is planned to establish and improve the VMAT treatment plan database based on the electronic signal of the anatomical structure, the LVW, target area information and other organs at risk, analyze the intensity map of each control point, and the corresponding dose distribution, use the intensity map and CT image as input, calibrate the output to the corresponding dose distribution, and train the deep learning network to realize the dose calculation model (200 patient data as a training set, 100 patient data as a validation set). 3) Automatic optimization is achieved based on the automatic derivation algorithm of the calculation graph. In practical applications, radiotherapy plans also need to consider objective factors such as the gantry position and the direction of the radiation field, and some complex and non-differentiable regularization terms need to be introduced, such as the modulation factor representing the degree of modulation. The automatic derivation algorithm based on the computational graph improves the calculation speed and can efficiently process complex regularization terms, and directly obtain the optimization parameters using the predicted dose distribution. On this basis, the integration of the prediction model and the optimization algorithm is completed, and finally an intelligent radiotherapy plan design technology for lateral ventricle wall protection that only requires manual fine-tuning is established.

[0078] In a specific embodiment, a preliminary experiment showed that the average dose of LVW using the artificial protection plan was reduced by 7.4% compared with the conventional radiotherapy plan. Fig. 9 , Fig.10As shown, Fig. 9 In the gray pie chart, LVM accounts for the highest proportion. In the colored pie chart, LVM accounts for a higher proportion than RCM. Fig.10 In the histogram, both the maximum dose and the average dose show that the protective plan dose of LVM is lower, and the conventional plan and protective plan doses of PTV are equal, indicating that protective planning in LVM can achieve the minimum and ideal dose for the tumor.

[0079] The verification results of this verification embodiment show that assigning inherent weights to indications can improve the performance of the method relative to the default setting. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units. A person of ordinary skill in the art may understand that all or part of the steps in the various methods of the above-mentioned embodiments may be completed by instructing the relevant hardware through a program, and the program may be stored in a computer-readable storage medium, and the storage medium may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0080] A person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be implemented by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. The above-mentioned medium storage can be a read-only memory, a disk or an optical disk, etc.

[0081] The above is a detailed introduction to a computer device provided by the present invention. For a person skilled in the art, according to the concept of the embodiments of the present invention, there may be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for preventing brain radiation damage, characterized in that: include: S1. Obtaining brain imaging data of the subject; S2, performing regional delineation on the brain image data to obtain a lateral ventricle wall region; S3. Give the lateral ventricular wall area a dose lower than the preset dose.

2. The method for preventing brain radiation damage according to claim 1, characterized in that: The S2 is replaced by S2' to delineate the brain imaging area to obtain the tumor target area and the lateral ventricle wall area, and the S3 is replaced by S3' when the lateral ventricle wall area intersects with the tumor target area, a first irradiation dose lower than a preset dose is given to the lateral ventricle wall area within the intersection area; and a second irradiation dose preset is given to the lateral ventricle wall area outside the intersection area; Optionally, the first irradiation dose is a dose hotspot of a tumor target area, and the second irradiation dose is calculated in the same manner as a preset irradiation dose; Optionally, S3' further includes reducing the irradiated volume of the lateral ventricle wall area outside the intersection area.

3. The method for preventing brain radiation damage according to claim 1 or 2, characterized in that: The preset exposure dose or the preset second exposure dose is an exposure dose obtained by calculating the relationship between brain radiation necrosis occurring in the lateral ventricle wall area and different exposure doses; Optionally, the calculation method of the preset exposure dose or the preset second exposure dose includes: Obtain brain images of patients with brain radiation necrosis; Delineating the brain image to obtain the lateral ventricle wall area; Calculate the maximum dose or average dose in the area of ​​brain radiation necrosis in the lateral ventricular wall area; The relationship between brain radiation necrosis and irradiation dose can be obtained by fitting mathematical relationships based on maximum dose or average dose; Obtaining a preset exposure dose or a preset second exposure dose based on the relationship between brain radiation necrosis and the exposure dose; Optionally, the brain image in S1 is a fusion image of CT and / or MRI.

4. The method for preventing brain radiation damage according to claim 1, characterized in that: The delineation of the lateral ventricle wall region is obtained by automatically delineating the target region on the brain image through a network model; Optionally, the network model includes one or more of the following: CNN, mask RNN, Faster R-CNN, FCN, SegNet, PSPNet, DeepLab v1 / v2 / v3; Optionally, the training process of the network model is: Obtain patient brain image sets and labels; The brain image set is input into the network model to be trained for model training to obtain the network model.

5. The method for preventing brain radiation damage according to claim 1 or 2, characterized in that: The preset exposure dose is replaced by a preset total exposure dose; Optionally, the preset second exposure dose is replaced by a preset total exposure dose; Optionally, the calculation process of the total irradiated dose is: Obtain the brain image of the subject and the dose of each minimum beam to a single voxel; Performing regional delineation on the brain image data to obtain a lateral ventricle wall region sequence; The lateral ventricle wall region sequence and the dose of each minimum beam to a single voxel are input into the topological network structure for calculation, and the total dose of the lateral ventricle wall region is obtained; Optionally, the dose of each minimum beam to a single voxel is calculated by a topological network structure to obtain a total dose of a single voxel point in the lateral ventricle wall region sequence, and the total dose of the single voxel point in the lateral ventricle wall region is summed to obtain the total dose of the lateral ventricle wall region; Optionally, the total irradiation dose of the lateral ventricle wall region is obtained by weighted averaging the total irradiation doses of a single voxel point.

6. The method for preventing brain radiation damage according to claim 1 or 2, characterized in that: The preset exposure dose is replaced by a preset exposure dose distribution map; Optionally, the preset second exposure dose is replaced by a preset exposure dose distribution map; Optionally, the calculation process of the irradiated dose distribution map is: Obtaining brain images of the subject; Performing regional delineation on the brain image data to obtain a tumor target area and a lateral ventricle wall area; Calculate the radiation intensity of the lateral ventricle wall region sequence and the target region sequence to obtain the radiation intensity map of the lateral ventricle wall region and the radiation intensity map of the target region; Inputting the lateral ventricle wall region sequence, the target region sequence, the lateral ventricle wall region radiation intensity map, and the target region radiation intensity map into the neural network to obtain the lateral ventricle wall region sequence radiation dose distribution map and the target region sequence radiation dose distribution map; Optionally, the Shenjiang network is a neural network obtained by training by acquiring a lateral ventricle wall region sequence, a target region sequence, a lateral ventricle wall region illumination intensity map, and a target region illumination intensity map for feature learning.

7. A brain intelligent radiotherapy method, characterized in that: include: Obtain the brain image of the subject and the dose of each minimum beam to a single voxel; The dose distribution diagram of the target area sequence, the lateral ventricle wall area sequence or other involved organ area sequence is obtained by calculating the irradiated dose distribution according to claim 6; Inputting the dose distribution map of the target area sequence, the lateral ventricle wall area sequence or other involved organ area sequence into the target model to obtain the radiation dose prediction result of the target area, the lateral ventricle wall area or other involved organ area; Optionally, the target model is a topological network structure model, and the construction process of the topological network structure is: obtaining the dose of each minimum beam to a single voxel, and constructing a topological network structure model based on the topological structure for the dose distribution map of each minimum beam to a single voxel, a target area sequence, a lateral ventricle wall area sequence or other involved organ area sequence; Optionally, the target model further includes parameter optimization, obtaining optimized parameters through the dose distribution map, comparing the optimized parameters with the prediction results to obtain a loss function, adjusting the parameters of the target model based on the loss function, and repeating the training until the loss function remains unchanged to obtain the target model; Optionally, the parameter optimization further includes introducing a non-differentiable regularization term, and removing interference information of the sequence or distribution graph through the regularization term.

8. A computer program product comprising a computer program or instructions, characterized in that: The computer program or instructions are executed by a processor to implement the method for preventing brain radiation damage described in any one of claims 1 to 6, or to implement the intelligent brain radiotherapy method described in claim 7.

9. A computer device comprising a memory, a processor and a computer program or instruction stored in the memory, characterized in that: The computer program or instructions are executed by a processor to implement the method for preventing brain radiation damage described in any one of claims 1 to 6, or to implement the intelligent brain radiotherapy method described in claim 7.

10. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: The computer program or instructions are executed by a processor to implement the method for preventing brain radiation damage described in any one of claims 1 to 6, or to implement the intelligent brain radiotherapy method described in claim 7.

Images