Intelligent diagnosis auxiliary method and system based on prostate cancer lesion and storage medium

Abstract

The application discloses a prostate cancer lesion-based intelligent diagnosis auxiliary method and system and a storage medium, relates to the tumor treatment data processing technical field, and comprises the following steps: acquiring current patient case data and image data of a lesion site; searching for and acquiring reference image data sets, statistically analyzing the probability of the lesion tissue appearing at different positions in space based on the reference image data sets, and generating a probability distribution graph of the lesion tissue; calculating and generating a radiation dose distribution graph of the three-dimensional space of the current lesion area according to particle implantation site coordinates; and determining implantation coordinates of superimposed particles according to the probability distribution graph and the radiation dose distribution graph, and marking the implantation coordinates in the image data of the current patient. Through the above scheme, the implantation coordinates of the superimposed particles are determined, the implantation positions of the radioactive particles are more suitable for the distribution of the lesion tissue, the cancer cells of the lesion tissue are killed, the spread of the cancer cells is slowed down or even blocked, and the treatment effect is improved.

Description

Technical Field

[0001] This application relates to the field of tumor treatment data processing technology, and specifically to an intelligent diagnostic assistance method, system, and storage medium based on prostate cancer lesions. Background Technology

[0002] Prostate cancer is a common malignant tumor of the male reproductive system, mainly occurring in the cells of the prostate gland. Prostate cancer brachytherapy is a minimally invasive treatment that precisely implants radioactive particles (such as iodine-125) into the prostate to kill cancer cells using localized radiation. It is suitable for patients with early-stage localized prostate cancer, those who cannot tolerate surgery, or those who need to preserve organ function. It is characterized by minimal trauma and rapid recovery.

[0003] Currently, in the treatment of prostate cancer patients using particle implantation, the prostate and urethra must first be marked. Then, using ultrasound imaging and a particle implantation device, particles are implanted into the prostate. In current procedures, particles are implanted evenly across the entire prostate. Later, when lesions or suspected lesions are found, additional particles are implanted at the corresponding locations based on the doctor's experience. This process relies on the doctor's subjective judgment, which can lead to over-implantation in localized areas, negatively impacting treatment effectiveness. Summary of the Invention

[0004] In the treatment of prostate cancer using radioactive particle implantation, the inability to accurately determine the location of lesions due to the difficulty in precise imaging of local areas, coupled with the significant deviation that can occur when doctors manually determine the particle implantation point, presents a problem. The first objective of this application is to propose an intelligent diagnostic assistance method based on prostate cancer lesions. This method can automatically perform calculations and analyses based on the lesion's condition and mark reference implantation sites around the lesion for the superimposed radioactive particles, assisting doctors in particle implantation and improving treatment outcomes. To implement the above intelligent diagnostic assistance method, the second objective of this application is to provide an intelligent diagnostic assistance system based on prostate cancer lesions. Finally, to facilitate the widespread use of the above intelligent diagnostic assistance method, the third objective of this application is to propose a computer-readable storage medium, the specific scheme of which is as follows:

[0005] A smart diagnostic aid method based on prostate cancer lesions, comprising:

[0006] Obtain the current patient's medical records and imaging data of the lesion site;

[0007] The image data of patients with similar lesion locations to the current patient's case data are retrieved from the large database of prostate cancer cases and stored as reference cases and reference image datasets, respectively.

[0008] Based on the reference image dataset, the probability of lesion tissue appearing at different locations in space is statistically analyzed. Combined with the lesion tissue morphology and location already determined for the current patient, a probability distribution map of the lesion tissue appearing at various locations in three-dimensional space is generated.

[0009] The distribution of blood vessels and the direction of blood flow through the lesion site are identified from the imaging data of the current lesion site. Based on the coordinates of the particle implantation site and the radiation dissipation dose with the blood flow, a three-dimensional radiation dose distribution map of the current lesion area is generated.

[0010] The coordinates of the implantation site of the superimposed particles are determined based on the probability distribution map and the radiation dose distribution map of the current patient's lesion site, and then marked in the current patient's imaging data.

[0011] The radiation dose at the implantation site coordinates of each superimposed particle satisfies the OARs constraint conditions.

[0012] Using the above technical solution, when formulating a radioactive particle implantation plan for the current patient, the shape and distribution of lesions in other patients with similar case data can be referenced. Combined with the current patient's established imaging data, the potential lesion distribution location can be inferred, thereby determining the implantation site coordinates of the superimposed particles. This allows the implantation location of the radioactive particles to better match the distribution location of the lesion, which helps to kill cancer cells in the lesion, slow down or even block the spread of cancer cells, and improve the treatment effect.

[0013] Furthermore, a search was conducted in the large database of prostate cancer cases for patients with similar case data to the current patient, including:

[0014] Obtain quantitative data on the current patient's lesion tissue;

[0015] Based on a large database of prostate cancer cases, correlation analysis was used to obtain the influencing factors related to the above quantitative data.

[0016] Search the large database of prostate cancer cases for patients who match the above-mentioned influencing factors and output the results;

[0017] The quantitative data includes Gleason classification data, PSA level data, prostate cancer staging data, or data used to characterize lesion features after being weighted and fused by a set algorithm.

[0018] The influencing factors include age, genetic history, treatment history, specific hormone levels, lifestyle characteristics, living environment characteristics, and level of inflammatory infection.

[0019] The above technical solutions can accurately and effectively identify patients with similar cases to the current patient. Based on the data of existing patients, doctors can customize a more reasonable and precise particle implantation plan for the current patient.

[0020] Furthermore, a probability distribution map of the occurrence of lesion tissue at various locations in three-dimensional space is generated, including:

[0021] A three-dimensional coordinate system is established with reference to the current patient's image data, and the three-dimensional space of a set size is divided into multiple statistical regions;

[0022] A specific site in the prostate is selected as a reference benchmark. The positions of lesions in each image data in the reference image dataset are statistically analyzed in the above three-dimensional coordinate system. The number of times lesions appear in each statistical region is counted and compared with the number in the reference image dataset to generate a first probability distribution map.

[0023] Based on the shape and size of the lesion tissue contained in each image data in the reference image dataset, the edge contour features of the lesion tissue cells are obtained by training a deep neural network.

[0024] The visible edge contour of the lesion tissue is extracted from the current patient's image data. The current visible edge contour is expanded and improved using the aforementioned edge contour features. A second probability distribution map is generated based on the degree of matching between the visible edge contour and the edge contour features.

[0025] The first probability distribution map and the second probability distribution map are weighted and merged to generate the probability distribution map.

[0026] By using the above technical solutions, the distribution probability of lesion tissue in three-dimensional space can be obtained by improving the lesion tissue morphology of similar patients and the lesion tissue morphology of the current patient. This method has higher reliability and can help doctors find the optimal position coordinates of superimposed particles, thereby improving the treatment effect.

[0027] Furthermore, generating a probability distribution map of the occurrence of lesion tissue at various locations in three-dimensional space also includes:

[0028] Acquire reference images taken at each diagnosis for each reference case and form multiple sets of diffusion images based on the reference case ID;

[0029] The diffusion direction and speed of the concentrated lesion tissue in each group of diffusion images are calculated and analyzed in three-dimensional space over time, and the probability of the generated lesion tissue spreading in each direction in three-dimensional space is statistically analyzed.

[0030] Based on the current morphology of the patient's lesions, estimate the spread area and location of the current patient's lesions within a set time period, and generate a third probability distribution map;

[0031] The first probability distribution map, the second probability distribution map, and the third probability distribution map are weighted and merged to generate the probability distribution map.

[0032] By using the above technical solutions, the direction and speed of the spread of lesions in similar diseases can be analyzed to predict the current direction and speed of the spread of lesions in patients, thereby making the predicted location and distribution of lesions more accurate.

[0033] Furthermore, generating a probability distribution map of the occurrence of lesion tissue at various locations in three-dimensional space also includes:

[0034] Acquire the imaging data collected during the current patient's previous diagnoses, compare and analyze the diffusion direction and diffusion speed of the lesion tissue, and generate a diffusion prediction model;

[0035] Based on the diffusion prediction model, the distribution location of the lesion tissue in three-dimensional space after a set time is generated, and probability correction parameters are generated.

[0036] The probability distribution map is generated by correcting the third probability distribution map using the aforementioned probability correction parameters.

[0037] Using the above technical solution, the distribution range of lesions at a set time in the future can be predicted based on the current diffusion trend of the patient's lesion tissue. Using the above distribution range as a correction reference can make the results of the above probability distribution map more accurate.

[0038] Furthermore, a three-dimensional radiation dose distribution map of the current lesion area is generated, including:

[0039] To obtain the radiation dose of a single radioactive particle and its radiation range in three-dimensional space;

[0040] To obtain the escaping dose of radiation from radioactive particles as they flow through the bloodstream;

[0041] Based on the position coordinates of each radioactive particle and the distribution of blood vessels, and in conjunction with the radiation dose and radiation range mentioned above, the radiation dose at each location in the three-dimensional space of each lesion area is calculated, and the radiation dose distribution map is generated.

[0042] The escaping dose is inversely proportional to the distance of the radioactive particle from the blood vessel and directly proportional to the blood flow in the blood vessel.

[0043] The above technical solution can accurately calculate the range of radiation dissipation caused by the flow of radioactive particles through the blood, thus making the calculation of radiation dose in the lesion area more accurate and avoiding excessive radiation dose in local areas.

[0044] Furthermore, based on the currently determined coordinates of the particle implantation site and the egress dose of radiation flowing through the bloodstream, it also includes:

[0045] The distance between the lesion tissue and the blood vessels at the lesion site is obtained based on the current patient's imaging data;

[0046] Based on the diffusion prediction model, the time points at which the lesion tissue compresses the blood vessels at the lesion site are obtained and the compression amplitude is calculated.

[0047] Based on the timing and magnitude of the compression exerted by the lesion on the blood vessels at the lesion site, a correction data is generated to correct the blood flow data of the aforementioned blood vessels.

[0048] The above technical solution fully considers the impact of lesion tissue diffusion on normal blood flow, and thus makes accurate corrections to the range and speed of radioactive particle radiation dissipation, making the location coordinates of superimposed particle implantation more precise, avoiding local radiation dose in the lesion area from exceeding the limit or falling below the set dose, and ensuring the treatment effect.

[0049] To implement the above-mentioned intelligent diagnostic assistance method based on prostate cancer lesions, this application also proposes an intelligent diagnostic assistance system based on prostate cancer lesions, comprising:

[0050] The first data acquisition unit is configured to acquire the current patient's medical records and imaging data of the lesion site;

[0051] The second data acquisition unit is configured to connect to a large database of prostate cancer cases. It is used to search the large database of prostate cancer cases for patients and their corresponding lesion sites that are similar to the current patient's case data, and store them as reference cases and reference image datasets respectively.

[0052] The lesion location generation unit is configured to statistically analyze the probability of lesion tissue appearing at different locations in space based on the reference image dataset, and generate a probability distribution map of lesion tissue appearing at various locations in three-dimensional space by combining the lesion tissue morphology and location already determined for the current patient.

[0053] The radiation dose distribution generation unit is configured to identify the distribution of blood vessels and the direction of blood flow through the lesion site from the image data of the current patient's lesion site, and generate a three-dimensional radiation dose distribution map of the current lesion area based on the currently determined particle implantation site coordinates and the radiation dissipation dose with the blood flow.

[0054] The implantation site annotation unit is configured to determine the implantation site coordinates of the superimposed particles based on the probability distribution map combined with the radiation dose distribution map of the current patient's lesion site, and to annotate them in the current patient's image data.

[0055] Furthermore, the lesion location generation unit includes:

[0056] The statistical region construction sub-unit is configured to establish a three-dimensional coordinate system with reference to the current patient's image data, and divide the three-dimensional space of a set size into multiple statistical regions;

[0057] The first probability distribution generation subunit is configured to select a set site in the prostate as a reference benchmark, count the position of the lesion tissue contained in each image data in the reference image dataset in the above three-dimensional coordinate system, count the number of times the lesion tissue appears in each statistical region and compare it with the number in the reference image dataset to generate the first probability distribution map.

[0058] The second probability distribution generation subunit is configured to obtain the edge contour features of lesion tissue cells based on the shape and size of the lesion tissue contained in each image data in the reference image dataset, extract the visible edge contour of the lesion tissue from the current patient's image data, expand and improve the current visible edge contour using the above edge contour features, and generate a second probability distribution map based on the degree of matching between the visible edge contour and the edge contour features.

[0059] The probability fusion subunit weights and fuses the first probability distribution map and the second probability distribution map to generate the probability distribution map.

[0060] A computer-readable storage medium having a computer program module loaded thereon, which, when executed by a processor, is used to implement the intelligent diagnostic assistance method based on prostate cancer lesions as described above.

[0061] The above technical solutions facilitate the promotion and use of the aforementioned intelligent diagnostic assistance methods.

[0062] This application includes at least one of the following beneficial effects:

[0063] (1) By referring to the shape and distribution of lesion tissues of other patients with similar case data to the current patient, and combining the existing image data of the current patient, the potential lesion tissue distribution location can be inferred, thereby determining the implantation site coordinates of the superimposed particles, so that the implantation location of the radioactive particles is more in line with the distribution of the lesion tissue, which helps to kill cancer cells in the lesion tissue, slow down or even block the spread of cancer cells, and improve the treatment effect.

[0064] (2) By incorporating the diffusion speed and direction of the lesion tissue itself into the calculation of the particle implantation location coordinates, the implanted radioactive particles can more effectively block the spread of cancer cells and improve the effect of particle implantation in the treatment of prostate cancer. Attached Figure Description

[0065] Figure 1 This is a schematic diagram illustrating the steps of the intelligent diagnostic assistance method of this application;

[0066] Figure 2 A schematic diagram illustrating a method for generating the probability distribution of lesion tissue occurrence at various locations in three-dimensional space;

[0067] Figure 3 This is a schematic diagram showing the connection of the functional units of the intelligent diagnostic auxiliary system of this application.

[0068] Figure reference numerals: 100, first data acquisition unit; 200, second data acquisition unit; 300, lesion location generation unit; 400, radiation dose distribution generation unit; 500, implantation site annotation unit. Detailed Implementation

[0069] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0070] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0071] A smart diagnostic aid method based on prostate cancer lesions is proposed to assist doctors in determining the location of the lesions in a patient, thereby enabling more precise implantation of superimposed particles at their coordinates. Figure 1 As shown, the method mainly includes the following steps:

[0072] S100: Acquire the current patient's medical records and imaging data of the lesion site;

[0073] S200: Search the large database of prostate cancer cases for patients and their corresponding lesion locations that are similar to the current patient's case data, and store them as reference cases and reference image datasets respectively.

[0074] S300: Based on the reference image dataset, the probability of lesion tissue appearing at different locations in space is statistically analyzed, and combined with the lesion tissue morphology and location already determined for the current patient, a probability distribution map of the lesion tissue appearing at various locations in three-dimensional space is generated.

[0075] S400: Based on the imaging data of the current patient's lesion site, the distribution of blood vessels and the direction of blood flow through the lesion site are confirmed. Based on the currently determined particle implantation site coordinates and the radiation dissipation dose with the blood flow, a three-dimensional radiation dose distribution map of the current lesion area is generated.

[0076] S500, based on the probability distribution map and the radiation dose distribution map of the current patient's lesion site, determine the coordinates of the implantation site of the superimposed particles and mark them in the current patient's imaging data.

[0077] In step S100 above, the current patient's case data mainly includes the following: patient name, age, physical data, current case data generation time, chief complaint information, medical history, family history, examination parameters, diagnostic opinions, and treatment plan. The examination parameters include PSA concentration level, prostate volume, and adjacent lymph node size; the diagnostic opinions include data such as nodular isodense lesions at the apex of the prostate, lymph node enlargement, and PRIMARY score; the treatment plan includes current medication data, radiotherapy and chemotherapy data, etc. Imaging data includes one or more of MRI, ultrasound, and CT images.

[0078] In step S200, searching the prostate cancer case database for patients with similar case data to the current patient further includes:

[0079] S210, Obtain quantitative data of the current patient's lesion tissue. The aforementioned quantitative data includes Gleason grade data, PSA level data, prostate cancer staging data, or data formed by weighted fusion using a set algorithm to characterize the lesion features. For example, data formed by weighted fusion of the volume and shape of the lesion tissue after taking feature values. The purpose is to form a specific evaluation index.

[0080] S211, based on a large database of prostate cancer cases, uses correlation analysis to obtain influencing factors related to the aforementioned quantitative data. These correlation analyses include principal component analysis (PCA) or multivariate analysis, used to identify case data whose correlation with the quantitative data exceeds a set threshold as influencing factors.

[0081] In detail, the influencing factors include age, genetic history, treatment history, specific hormone levels, lifestyle characteristics, living environment characteristics, and level of inflammatory infection.

[0082] S212, the system searches the prostate cancer case database for patients matching the aforementioned influencing factors and outputs the relevant case data. For example, if the influencing factors are selected as the current patient's age, PSA level, duration of disease, and prior use of estradiol, the system automatically filters the prostate cancer case database for patients similar to these factors and outputs the relevant case data. Based on this technical approach, similar patients can be accurately and effectively identified. Using the data from existing cases, doctors can customize a more reasonable and precise particle implantation plan for the current patient.

[0083] In step S300, a probability distribution map of the occurrence of lesion tissue at various locations in three-dimensional space is generated, such as... Figure 2 As shown, it includes:

[0084] S310: A three-dimensional coordinate system is established with reference to the current patient's image data, and the three-dimensional space of a set size is divided into multiple statistical regions. These multiple statistical regions are also three-dimensional regions, and statistical calculations are performed in the form of multiple layered images during the statistical process.

[0085] S311, Select a specific site in the prostate as a reference benchmark, such as the bladder neck as a reference site, count the position of the lesion tissue contained in each image data in the reference image dataset in the above three-dimensional coordinate system, count the number of times the lesion tissue appears in each statistical region and compare it with the number in the reference image dataset to generate a first probability distribution map.

[0086] S320: Based on the shape and size of the lesion tissue contained in each image data in the reference image dataset, the edge contour features of the lesion tissue cells are obtained by training a deep neural network, such as a convolutional neural network (CNN).

[0087] S321, extract the visible edge contour of the lesion tissue from the current patient's image data, expand and improve the current visible edge contour using the aforementioned edge contour features, and generate a second probability distribution map based on the degree of matching between the visible edge contour and the edge contour features. The above expansion and improvement steps include extracting contours from the current patient's image data, then performing contour matching, and finally obtaining and filling in the missing contours in the image data based on the matched contour features.

[0088] S330, the first probability distribution map and the second probability distribution map are weighted and fused to generate the probability distribution map. For example, the probability of lesion tissue appearing at a certain coordinate position in the three-dimensional coordinate system is: P1=W1*P2+W2*P3, where P2 represents the probability of lesion tissue appearing at the above position in patients with similar cases, P3 represents the probability of lesion tissue appearing at the above position after the image data of the current patient's lesion tissue is improved, and W1 and W2 are the weights of the values, usually W2>W1.

[0089] In practical applications, after radioactive particles are implanted at the site of a prostate lesion, the lesion tissue in the prostate will naturally diffuse, and the radioactive particles themselves have a certain decay period, such as iodine-125 ( 125The half-life of I) is 59.4 days; therefore, the diffusion of lesion tissue within at least 60 days must be considered during particle implantation. To this end, in this optimized embodiment, to further determine the current location of the patient's lesion tissue and improve the accuracy of radioactive particle implantation, step S300, which generates a probability distribution map of the lesion tissue appearing at various locations in three-dimensional space, further includes:

[0090] S340, acquire reference images taken at each diagnosis for each reference case and form multiple sets of diffusion images based on the reference case ID.

[0091] S341, Calculate and analyze the diffusion direction and speed of the lesion tissue in the three-dimensional space of each group of diffusion images, and statistically calculate the probability of the lesion tissue spreading in each direction in the three-dimensional space.

[0092] S342, combining the current lesion tissue morphology of the patient, estimate the diffusion area and location of the current patient's lesion tissue within a set time period, and generate a third probability distribution map.

[0093] S331, the first probability distribution map, the second probability distribution map, and the third probability distribution map are weighted and fused to generate the probability distribution map.

[0094] The above technical solution can predict the direction and speed of the spread of lesions in the current patient by analyzing the spread direction and speed of lesions in similar diseases, thereby making the predicted location distribution of lesions more accurate. The above solution is well adapted to the spread characteristics of lesions.

[0095] Considering the current patient's lesion tissue's own diffusion characteristics, if the current patient's case data contains image data collected during previous diagnoses, then further generating a probability distribution map of the lesion tissue appearing at various locations in three-dimensional space also includes:

[0096] S350: Acquire image data collected during each of the patient's previous diagnoses, compare and analyze the diffusion direction and speed of the lesion tissue, and generate a diffusion prediction model. This diffusion prediction model includes the direction of lesion tissue diffusion into three-dimensional space and the corresponding diffusion speed curve.

[0097] S351, Based on the diffusion prediction model, generate the distribution position of the lesion tissue in three-dimensional space after a set time, and generate probability correction parameters;

[0098] S352, the third probability distribution map is corrected using the above probability correction parameters, and then step S331 is performed to generate the probability distribution map.

[0099] The above technical solution can predict the distribution range of lesions at a set time in the future based on the current diffusion trend of the patient's lesion tissue. Using the above distribution range as a correction reference can make the results of the above probability distribution map more accurate.

[0100] To ensure treatment safety, the radiation dose at each implantation site coordinate of the superimposed particles meets the OARs (Organs at Risk) constraints. During the particle implantation surgery, the system marks the number of particles implanted at each coordinate. The aforementioned OARs (Organs at Risk) constraints refer to a series of restrictions that must be followed when formulating a treatment plan to protect the patient's normal tissues and organs during radiotherapy. In the embodiments of this application, the distance between implanted particles is set to 0.5-1 cm, and the prescription dose for particle implantation is set according to the size of the prostate and the size of the lesion.

[0101] Theoretically, radiation from radioactive particles refers to the process of energy propagating in the form of waves or particles. Blood irradiated by these particles will also possess a certain degree of radioactivity. That is, when the particles are implanted near blood vessels, such as the prostatic venous plexus, the radiation dose will dissipate.

[0102] In step S400, a three-dimensional radiation dose distribution map of the current lesion area is generated, including:

[0103] To obtain the radiation dose of a single radioactive particle and its radiation range in three-dimensional space;

[0104] To obtain the escaping dose of radiation from radioactive particles as they flow through the bloodstream;

[0105] Based on the position coordinates of each radioactive particle and the distribution of blood vessels, and in conjunction with the radiation dose and radiation range mentioned above, the radiation dose at each location in the three-dimensional space of each lesion area is calculated, and the radiation dose distribution map is generated.

[0106] The escaping dose is inversely proportional to the distance between the radioactive particle and the blood vessel, and directly proportional to the blood flow in the blood vessel. That is, the closer the radioactive particle is to the blood vessel and the greater the blood flow in the blood vessel, the higher the escaping dose.

[0107] Based on the above technical solution, the range of radiation dissipation caused by the flow of radioactive particles through the blood can be accurately calculated. This makes the calculation of radiation dose in the lesion area more accurate and avoids excessive radiation dose in local areas.

[0108] To make the calculation of the above-mentioned escaping dose more accurate, step S400 also includes:

[0109] S410 obtains the distance between the lesion tissue and the blood vessels at the lesion site based on the current patient's imaging data.

[0110] S411, Based on the diffusion prediction model, obtain the time point at which the lesion tissue compresses the blood vessels at the lesion site and calculate the compression amplitude.

[0111] S412, Based on the time point and magnitude of compression of the blood vessels at the lesion site by the lesion tissue, generate correction data to correct the blood flow data of the aforementioned blood vessels.

[0112] The above technical solution fully considers the impact of lesion tissue diffusion on normal blood flow, and thus makes accurate corrections to the range and speed of radioactive particle radiation dissipation, making the location coordinates of superimposed particle implantation more precise, avoiding local radiation dose in the lesion area from exceeding the limit or falling below the set dose, and ensuring the treatment effect.

[0113] To implement the above-mentioned intelligent diagnostic assistance method based on prostate cancer lesions, this application also discloses an intelligent diagnostic assistance system based on prostate cancer lesions, such as... Figure 3 As shown, it mainly includes the following functional units: first data acquisition unit 100, second data acquisition unit 200, lesion location generation unit 300, radiation dose distribution generation unit 400, and implantation site marking unit 500.

[0114] The first data acquisition unit 100 is configured to acquire the current patient's medical record data and image data of the lesion site, which are directly exported from the medical institution's medical record database. The second data acquisition unit 200 is configured to connect to a large database of prostate cancer cases via a communication network, and is used to search the large database for patients with similar medical records and their corresponding image data of the lesion site. The data is then stored locally as reference cases and reference image datasets for easy retrieval.

[0115] The lesion location generation unit 300 is configured to statistically analyze the probability of lesion tissue appearing at different locations in space based on a reference image dataset, and combine this with the already determined lesion tissue morphology and location for the current patient to generate a probability distribution map of the lesion tissue appearing at various locations in three-dimensional space. In practical applications, the above probability distribution map is stored as a two-dimensional data table of coordinate data and associated probability values.

[0116] The radiation dose distribution generation unit 400 is configured to identify the distribution of blood vessels and the direction of blood flow through the lesion site from the imaging data of the current patient's lesion site. This identification is performed by the physician. Then, based on the already determined coordinates of the particle implantation site and the radiation dissipation dose with the blood flow, a three-dimensional radiation dose distribution map of the current lesion area is generated. In practical applications, the above calculation process is automatically completed by a pre-set program algorithm. That is, after determining the initial radiation intensity of the implanted radioactive particle, the blood flow rate in the blood vessel, and the coordinates of the particle implantation position, the radiation dose distribution in three-dimensional space is automatically calculated by the program algorithm.

[0117] The implantation site annotation unit 500 is configured to determine the implantation site coordinates of the superimposed particles based on the probability distribution map and the radiation dose distribution map of the current patient's lesion site, and to annotate them in the current patient's image data. In practical applications, to make the implantation site coordinates more intuitive, the system of this application is also equipped with a three-dimensional image display module to display the relative positional relationship between the lesion tissue and the implanted particles, as well as the number of particles currently implanted at each implantation site coordinate.

[0118] In detail, the lesion location generation unit 300 specifically includes: a statistical region construction subunit, a first probability distribution generation subunit, a second probability distribution generation subunit, and a probability fusion subunit.

[0119] The statistical region construction subunit is configured to establish a three-dimensional coordinate system based on the current patient's image data and divide the three-dimensional space of a set size into multiple statistical regions. The first probability distribution generation subunit is configured to select a set site in the prostate as a reference benchmark according to the user's (doctor's) operation instructions. The set site can be several reference sites on the prostatic artery. The reference image data is adjusted, such as tilted or magnified, so that the reference image overlaps with the reference sites in the current patient's lesion tissue image. The positions of the lesion tissue contained in each image data in the reference image dataset are counted in the above three-dimensional coordinate system. The number of times the lesion tissue appears in each statistical region is counted and compared with the number in the reference image dataset to generate the first probability distribution map.

[0120] The second probability distribution generation subunit is configured to obtain the edge contour features of lesion tissue cells based on the shape and size of the lesion tissue contained in each image data in the reference image dataset, using a deep neural network for training. It then extracts the visible edge contour of the lesion tissue from the current patient's image data, expands and refines the current visible edge contour using the aforementioned edge contour features, and generates a second probability distribution map based on the degree of matching between the visible edge contour and the edge contour features. The deep neural network used is a CNN model. The probability fusion subunit is configured to weightedly fuse the first and second probability distribution maps to generate the probability distribution map. In specific implementation, the weight values ​​corresponding to the first and second probability distribution maps can be set by the user.

[0121] Finally, this application also discloses a computer-readable storage medium on which a computer program module is loaded. When the program module is executed by a processor, it is used to implement the intelligent diagnostic assistance method based on prostate cancer lesions as described above.

[0122] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. An intelligent diagnostic assistance system based on prostate cancer lesions, characterized by, The application relates to a method for determining the implantation site of a superimposed particle in a prostate cancer treatment, comprising the following steps: a first data acquisition unit (100) is configured to acquire case data of a current patient and image data of a lesion site; a second data acquisition unit (200) is configured to be connected with a prostate cancer case database, and is used for searching for image data corresponding to a patient and a lesion site of the patient which are similar to the case data of the current patient from the prostate cancer case database, and storing the image data as reference case data and reference image data set respectively; a lesion site generation unit (300) is configured to statistically obtain the probability of a lesion tissue appearing at different positions in space based on the reference image data set, and to generate a probability distribution diagram of the lesion tissue appearing at different positions in a three-dimensional space by combining a determined lesion tissue form and position of the current patient; a radiation dose distribution generation unit (400) is configured to confirm the blood vessel distribution and blood flow direction flowing through the lesion site from the image data of the lesion site of the current patient, and to generate a radiation dose distribution diagram of the three-dimensional space of the current lesion region according to the coordinates of the particle implantation site and the dispersion dose of the radiation along with the blood flow; an implantation site marking unit (500) is configured to determine the implantation site coordinates of the superimposed particle according to the probability distribution diagram and the radiation dose distribution diagram of the lesion site of the current patient, and to mark the implantation site coordinates in the image data of the current patient. The lesion site generation unit (300) comprises: a statistical region construction subunit configured to establish a three-dimensional coordinate system by taking the image data of the current patient as a reference, and to divide a three-dimensional space with a set size into a plurality of statistical regions; a first probability distribution generation subunit configured to select a set position in the prostate as a reference benchmark, to statistically obtain the position of the lesion tissue in the three-dimensional coordinate system in each image data in the reference image data set, to count the number of times that the lesion tissue appears in each statistical region and compare the number of times with the number of the reference image data set, and to generate a first probability distribution diagram; a second probability distribution generation subunit configured to obtain the edge contour features of the lesion tissue cells based on a deep neural network according to the shape and size of the lesion tissue in each image data in the reference image data set, to extract the visual edge contour of the lesion tissue from the image data of the current patient, to extend and perfect the current visual edge contour by using the edge contour features, and to generate a second probability distribution diagram according to the matching degree of the visual edge contour and the edge contour features; a probability fusion subunit configured to weight and fuse the first probability distribution diagram and the second probability distribution diagram to generate the probability distribution diagram. The radiation dose at the implantation site coordinates of each superimposed particle is limited to meet the constraint condition of a critical organ.

2. The intelligent diagnostic aid system of claim 1, wherein, In the second data acquisition unit (200), the searching for the patient similar to the current patient from the prostate cancer case database comprises the following steps: acquiring quantitative data of the lesion tissue of the current patient; obtaining an influence factor related to the quantitative data based on the prostate cancer case database by using a correlation analysis method; searching for a case patient matched with the influence factor from the prostate cancer case database and outputting the case patient. The quantitative data include Gleason grading data, PSA level data, prostate cancer staging data, or data formed by weighting and fusing set algorithms to represent lesion characteristics. The influence factors include age, genetic history, treatment history, specific hormone level, living habit characteristics, living environment characteristics, and inflammation infection level.

3. The intelligent diagnostic aid system of claim 1, wherein, In the lesion position generation unit (300), the probability distribution map of the lesion tissue appearing at each position in the three-dimensional space is further generated, and the method further includes: Reference images taken at each diagnosis of each reference case are acquired, and a plurality of sets of diffusion image sets are formed according to the reference case IDs; The diffusion direction and speed of the lesion tissue in the three-dimensional space over time in each set of the diffusion image sets are calculated and analyzed, and the probability of the lesion tissue diffusing in each direction in the three-dimensional space is statistically generated; In combination with the lesion tissue morphology of the current patient, the diffusion area and position of the lesion tissue of the current patient within a set time period are estimated, and a third probability distribution map is generated; The first probability distribution map, the second probability distribution map, and the third probability distribution map are weighted and fused to generate the probability distribution map.

4. The intelligent diagnostic aid system of claim 3, wherein, In the lesion position generation unit (300), the probability distribution map of the lesion tissue appearing at each position in the three-dimensional space is further generated, and the method further includes: Image data collected at each diagnosis of the current patient is acquired, and the diffusion direction and speed of the lesion tissue are compared and analyzed to generate a diffusion prediction model; According to the diffusion prediction model, the distribution position of the lesion tissue in the three-dimensional space after a set time period is generated, and a probability correction parameter is generated; The probability distribution map is generated by correcting the third probability distribution map using the probability correction parameter.

5. The intelligent diagnostic aid system of claim 4, wherein, In the radiation dose distribution generation unit (400), a radiation dose distribution map of the three-dimensional space of the current lesion area is generated, including: The radiation dose of a single radioactive particle and its radiation range in the three-dimensional space are acquired; The dispersion dose formed by the radiation of the radioactive particle along with the blood flow is acquired; According to the position coordinates of each radioactive particle and the distribution position of the blood vessels, the radiation dose and the radiation range are combined with the dispersion dose to calculate the radiation dose at each position in the three-dimensional space of each lesion area, and the radiation dose distribution map is generated; The dispersion dose is inversely proportional to the length of the radioactive particle from the blood vessels and proportional to the blood flow of the blood vessels.

6. The intelligent diagnostic aid system of claim 5, wherein, In the radiation dose distribution generation unit (400), according to the current determined particle implantation site coordinates and the dispersion dose of the radiation along with the blood flow, the method further includes: The distance between the lesion tissue and the blood vessels of the lesion site is acquired based on the image data of the current patient; The time point at which the lesion tissue causes compression to the blood vessels of the lesion site and the compression amplitude are acquired based on the diffusion prediction model; According to the time point at which the lesion tissue causes compression to the blood vessels of the lesion site and the compression amplitude, a correction data is generated to correct the blood flow data of the blood vessels.

Images