A system and method for determining the equivalent biological dose of targeted α-nucleoside therapy based on a voxel model.

Abstract

This invention discloses a voxel-based system and method for determining the equivalent biological dose of targeted alpha nuclide therapy. The system constructs a patient-specific three-dimensional voxel anatomical model and activity distribution map through an image processing module; an intelligent source term module, based on the decay characteristics of the selected alpha nuclide, calls a pre-built dose point kernel database to generate an equivalent total dose point kernel; a dose calculation module performs rapid convolution calculation based on cumulative activity and the equivalent total dose point kernel to obtain the three-dimensional physical absorbed dose; and a biological dose module integrates a micro-dose kinetic model to convert the physical dose into a three-dimensional biological dose distribution map equivalent to conventional external beam radiotherapy. This invention overcomes the technical shortcomings of standard models, such as large errors, time-consuming Monte Carlo simulations, reliance on multi-timepoint dynamic imaging, and the inability of physical dose to reflect the biological effects of alpha particles. It achieves efficient, accurate, and clinically practical individualized biological dose assessment, significantly improving the accuracy and safety of treatment plans.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of radiological and nuclear medicine technology, in particular to a voxel model-based targeted alpha nuclide therapy equivalent biological dose measurement system and method. BACKGROUND

[0002] Targeted alpha nuclide therapy (TAT) can precisely kill tumor cells while minimizing damage to surrounding normal tissues due to its high linear energy transfer (LET) and short-range physical properties, and it has great potential in dealing with micro-metastases. In order to achieve individualized precision therapy, it is crucial to accurately assess the radiation dose received by tumors and normal organs in patients.

[0003] Currently, the technical system for evaluating the biological effective dose of TAT mainly originates from the paradigm for beta nuclide therapy, which has the following systematic defects when dealing with the unique physical and biological characteristics of alpha nuclides:

[0004] 1. The dosimetry model based on standardized phantoms cannot reflect the real patient individual anatomy and drug distribution heterogeneity. Traditional internal irradiation dosimetry mainly relies on standardized models based on reference human phantoms, such as the S-value methodology proposed by the Medical Internal Radiation Dose Committee (MIRD) and software such as OLINDA / EXM and IDAC-Dose. This method ignores the huge differences between the sizes, shapes, positions of real patients and the standard phantom, as well as the morphology of tumors. The standard model usually assumes that the radioactive drug is uniformly distributed in the organ, but this is seriously inconsistent with clinical practice. Especially in tumor tissues, the uptake of drugs often has high heterogeneity. For alpha particles with a range of only a few cell diameters, this non-uniform activity distribution at the microscopic scale will cause a large gradient change in energy deposition, forming "cold" and "hot" areas. The calculation model based on the homogenization assumption will severely smooth out this critical dose heterogeneity, which may cause underestimation of tumor dose or misestimation of risk organ dose.

[0005] 2. Over-reliance on idealized dynamic imaging data is seriously out of line with clinical practice. Although the existing individualized dosimetry system can perform accurate dose calculation, its time-integrated model is originally designed for dynamic PET / CT and other multi-time-point imaging. In clinical practice, it is often difficult to perform multi-time-point scanning on alpha nuclide therapy patients due to patient tolerance, equipment time, radiation protection, and other reasons. Therefore, the existing technology has obvious deficiencies in accuracy and convenience of the estimation time integral when dealing with static or a few time-point SPECT / CT images that are more common in clinical practice.

[0006] 3. The simulation method for the decay characteristics of alpha nuclides is inefficient, and lacks a general and optimized computing framework. Different alpha nuclides have different decay characteristics. For example, 211 The decay chain of At is relatively simple, but there is a problem of a huge difference in the half-life of daughter nuclides; while 225 Ac and 223 Ra, which are representative of alpha nuclides, have complex decay chains and face the complexity of multiple alpha particle emission. The traditional Monte Carlo simulation method needs to simulate the entire decay chain process in each source voxel, and track the transport of each charged particle and photon. This process is extremely computationally intensive, and a single simulation may take tens or even hundreds of hours, which completely cannot meet the timeliness requirements of clinical decision-making. The prior art lacks a framework that can intelligently select the optimal simplified processing strategy according to the nuclide decay characteristics. For example, for 211 At, which has a relatively simple decay chain, and for 225 Ac, which has a complex decay chain, different physical approximations and simplified models may be needed to balance computational efficiency and accuracy. Existing systems usually use a "one-size-fits-all" direct simulation method, or are only optimized for specific nuclides by hand, which lacks flexibility and universality, resulting in low computational efficiency or instability when dealing with certain nuclides.

[0007] 4. The dose evaluation endpoint does not match the strong biological effect of alpha particles, and the clinical guidance value is limited. Most existing dosimetry systems provide the final result as physical absorbed dose (unit: Gy). However, the biological effect of alpha particles is much higher than that of photons or electrons of the same dose, and is closely related to the dose rate. Physical dose cannot intuitively reflect its strong biological killing effect. Although some studies have introduced the concept of equivalent biological dose (EQDX), many models simply apply radio-biological parameters (such as alpha / beta value) from external beam radiotherapy, without fully considering the continuous low dose rate irradiation mode and its unique microscopic dose distribution of TAT, which leads to a large deviation between the predicted cell survival rate and the actual situation. SUMMARY

[0008] In view of the deficiencies of the prior art, the present application provides a voxel model-based targeted alpha nuclide therapy equivalent biological dose determination system and method, which realizes efficient, accurate, and clinically practical individualized biological dose evaluation, significantly improves the accuracy and safety of the treatment plan, and solves the technical defects of the prior art such as large error of standard model, time-consuming Monte Carlo simulation, dependence on multi-time point dynamic imaging, and inability of physical dose to reflect the biological effect of alpha particles.

[0009] To achieve the above purpose, the technical solutions of the present application are as follows:

[0010] A voxel model based targeted alpha nuclide therapy equivalent biological dose determination system, comprising:

[0011] An image processing and individualized modeling module for importing and processing computed tomography (CT) images and single photon emission computed tomography (SPECT) or positron emission tomography (PET) images of a patient, generating a patient-specific three-dimensional voxel anatomical model and a three-dimensional voxel-level activity distribution map at at least one time point;

[0012] An intelligent source term processing module in communication with the image processing and individualized modeling module for receiving a nuclide selection instruction of a user interface and generating an equivalent total dose point kernel representing energy deposition characteristics of an entire decay chain of the alpha nuclide based on pre-stored decay chain data of the alpha nuclide;

[0013] A dose calculation module in communication with the image processing and individualized modeling module and the intelligent source term processing module for calculating a three-dimensional physical absorbed dose distribution based on the three-dimensional voxel anatomical model, the three-dimensional voxel-level activity distribution map and the equivalent total dose point kernel;

[0014] A biological dose calculation module in communication with the dose calculation module for integrating a microdose kinetics model to convert the three-dimensional physical absorbed dose distribution into a three-dimensional equivalent biological dose distribution map with a conventional fractionated external beam radiotherapy as a reference standard.

[0015] Preferably, the intelligent source term processing module comprises: a nuclide database for storing decay chain information of a plurality of alpha nuclides, the decay chain information including daughter nuclide species, half-life, decay branching ratio and alpha particle energy; a dose point kernel basis database for storing absorbed dose point kernels of alpha particles of different energies in different standard tissue models pre-calculated by Monte Carlo simulation; and a model selection and calculation unit for selecting and executing a corresponding simplified calculation model to generate the equivalent total dose point kernel according to a complexity of a decay chain of the selected alpha nuclide.

[0016] Preferably, the model selection and calculation unit is configured to:

[0017] For alpha nuclides with complex decay chains, a dose point kernel scaling model is executed to generate the equivalent total dose point kernel by weighting and superimposing absorbed dose point kernels of each alpha-emitting daughter nuclide in the nuclide decay chain according to its decay branching ratio :

[0018] ;

[0019] wherein, is a decay branching ratio of the i-th daughter nuclide; is an alpha particle energy released by the i-th daughter nuclide; The absorbed dose point kernel corresponding to the energy and tissue type is called from the dose point kernel database.

[0020] Preferably, the absorbed dose point kernels in the dose point kernel database are established by:

[0021] A plurality of standard tissue models are constructed, each tissue model having a specific element composition and physical density;

[0022] An isotropic alpha particle point source is defined at the geometric center of each tissue model;

[0023] Through Monte Carlo simulation, the energy deposition in a spherical shell of different radius centered on the alpha particle point source is statistically calculated, and finally the point kernel function of the absorbed dose with distance is obtained .

[0024] Preferably, the image processing and individual modeling module is integrated with a time integration algorithm for calculating the total cumulative activity of a voxel based on limited static imaging data; the image processing and individual modeling module is configured to:

[0025] The activity values measured at multiple time points for the same voxel are curve fitted to obtain a time-activity curve of the voxel;

[0026] The time-activity curve is integrated from zero time to infinity time to calculate the total cumulative activity of the voxel .

[0027] Preferably, the dose calculation module uses dose convolution method to calculate the three-dimensional physical absorbed dose distribution; the dose calculation module is configured to:

[0028] The total cumulative activity of each voxel is multiplied by the equivalent total dose point kernel to quickly obtain the absorbed dose of each voxel .

[0029] Preferably, the biological dose calculation module is configured to:

[0030] A microdose kinetics model is called to calculate the cell survival fraction S according to the radiobiological parameters of the target cells and the physical absorbed dose distribution;

[0031] Based on the cell survival fraction S and the biological parameters of the reference radiotherapy, the three-dimensional equivalent biological dose distribution map is calculated by inverse operation of the linear quadratic model.

[0032] The application also discloses a voxel model-based targeted alpha nuclide therapy equivalent biological dose determination method, comprising the following steps:

[0033] S100: Image acquisition, processing and individualized modeling;

[0034] Obtaining single photon emission computed tomography (SPECT) and computed tomography (CT) fusion images, or positron emission tomography (PET) and CT fusion images, of the patient at at least two time points after injection of a targeted alpha-emitting radionuclide radiopharmaceutical;

[0035] Segmenting and resampling the CT images to construct a patient-specific three-dimensional voxel anatomical model, wherein each voxel contains tissue type, density and spatial location information;

[0036] Attenuation correction, scatter correction and activity calibration are performed on the SPECT or PET images, and they are spatially registered with the CT images to generate a voxel-level three-dimensional activity distribution map that is spatially aligned with the three-dimensional voxel anatomical model;

[0037] S200: Intelligent source term generation;

[0038] Receiving a user-specified alpha-emitting radionuclide type;

[0039] Querying a pre-stored radionuclide database to obtain complete decay chain data of the alpha-emitting radionuclide, including half-lives, decay branching ratios and emitted alpha particle energies of each daughter radionuclide;

[0040] Intelligently selecting a simplified calculation model according to the complexity of the decay chain of the alpha-emitting radionuclide;

[0041] Based on the selected simplified calculation model and a pre-stored dose point kernel database, generating an equivalent total dose point kernel representing the spatial distribution of energy deposition of the entire decay chain of the alpha-emitting radionuclide under unit cumulative activity;

[0042] S300: Absorbed dose calculation based on dose point kernels

[0043] Based on the three-dimensional activity distribution map, the activity values of the same voxel at different time points are curve-fitted by a time integration algorithm, and the time-activity curve obtained by fitting is integrated to calculate the total cumulative activity of each voxel;

[0044] Using the equivalent total dose point kernel generated in step S200, the total cumulative activity is subjected to spatial energy deposition calculation to obtain a three-dimensional physical absorbed dose distribution;

[0045] S400: Equivalent biological dose calculation;

[0046] Inputting the three-dimensional physical absorbed dose distribution and corresponding dose rate information obtained in step S300 into a microdosimetric kinetic model;

[0047] calculating cell survival fraction by the microdosimetric kinetic model based on radiobiological parameters of the target region;

[0048] converting the cell survival fraction into a three-dimensional equivalent biological dose distribution map with conventional fractionated external radiotherapy as a reference standard;

[0049] S500: result output;

[0050] outputting and displaying the three-dimensional equivalent biological dose distribution map and / or the three-dimensional physical absorbed dose distribution map for clinical treatment evaluation.

[0051] Preferably, in the S200 step, when the α nuclide is a complex decay chain nuclide with multiple α-emitting daughter nuclei, the simplified calculation model is a dose point kernel scaling model; the specific steps of generating the equivalent total dose point kernel include:

[0052] calling the dose point kernels corresponding to the energy and tissue type of each α-emitting daughter nucleus in the decay chain from the dose point kernel database, and performing weighted superposition according to the decay branching ratio, the calculation formula being:

[0053] ;

[0054] wherein, is the equivalent total dose point kernel, is the decay branching ratio of the i-th daughter nuclide; is the energy of α particles released by the i-th daughter nuclide; is the corresponding basic dose point kernel.

[0055] Preferably, in the S400 step, the microdosimetric kinetic model is an improved model based on a linear quadratic model, and the calculation formula of cell survival fraction S is:

[0056] ;

[0057] wherein, D is the physical absorbed dose, and are model parameters corrected by microdosimetric parameters.

[0058] This invention provides a voxel-based system for determining the equivalent biological dosimetry of targeted alpha-nucleoside therapy, offering the following advantages: By constructing a patient-specific three-dimensional voxel anatomical model and integrating it with voxel-level activity distribution maps generated by quantitative SPECT / PET, the system fully preserves information on individual patient differences and tumor heterogeneity during calculations. It can accurately calculate microscopic dose variations caused by uneven activity distribution and different tissue densities. This avoids the risk of underestimating the dose in cold tumor zones or misestimating the dose in hot zones of high-risk organs due to the "homogenization" assumption, providing an unprecedentedly accurate data foundation for predicting treatment efficacy and toxicity.

[0059] By introducing a time-integration algorithm adapted to the design of single or a few routine clinical static SPECT / CT images, the cumulative absorbed dose over the entire treatment cycle can be calculated based on limited time-point data, simplifying the clinical data acquisition process and achieving accurate dose assessment based on static images.

[0060] By introducing an intelligent source term processing module, based on the characteristics of the nuclide decay chain, a dose point kernel scaling model or an equivalent source term optimization model is intelligently selected and applied. Through a pre-computation database and the principle of linear superposition, an "equivalent total dose point kernel" is generated, simplifying the complex decay chain simulation into efficient convolution or dot product operations. This achieves a dual improvement in computational efficiency and physical accuracy.

[0061] By integrating a micro-dose kinetic model, the three-dimensional physical dose distribution and dose rate information are converted into a three-dimensional equivalent biological dose distribution map, thereby making a more accurate judgment on the treatment effect and the risk of damage to normal tissues, and improving the clinical guidance value of dose assessment. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in this invention or the prior art, the accompanying drawings used in the description of this invention or the prior art will be briefly introduced below.

[0063] Figure 1 A schematic diagram of the system modules of this invention. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0065] Example 1, as Figure 1 As shown, this invention discloses a voxel-based system for determining the equivalent biological dose of targeted alpha nucleoside therapy. The system employs a modular design, with hardware deployed on a high-performance computing workstation or server cluster, and software integrating image processing, a Monte Carlo engine, and a radiobiology model. Specifically, it includes:

[0066] Image processing and individualized modeling module, located at the front end of the system, as the entrance of data input and preprocessing. For the original, noisy clinical images into accurate, physical computing can be used for individualized digital patient model. Input SPECT / CT or PET / CT DICOM format raw data of at least two time points after injection of alpha radionuclide drugs. Output individualized three-dimensional voxel anatomical model file (including geometry, material and segmentation label information) and registered, quantitative three-dimensional activity distribution map (one or more time points).

[0067] Its processing process includes: image registration: using mutual information-based rigid registration algorithm, the SPECT images at different time points are accurately aligned to the CT image space of a reference time point (usually the first time point), ensuring that the coordinate system of functional image and anatomical image is the same.

[0068] CT image segmentation and modeling: isotropic resampling (voxel size is usually set to 1-2mm 3 ) is performed on the reference CT image to balance the calculation accuracy and resource consumption. Threshold segmentation and region growing algorithm are applied, combined with atlas-guided automatic segmentation tool, to classify voxels into different tissue types (such as soft tissue, bone, lung tissue, etc.). Assign each tissue type voxel with corresponding element composition (such as soft tissue simulation H:10.1%, C:11.1%, N:2.6%, O:76.2%) and mass density (such as soft tissue 1.04g / cm 3 ), to construct a three-dimensional voxelized anatomical model containing material properties.

[0069] SPECT image quantification: use CT generated μ-map to correct SPECT image. Estimate and subtract scattered photon contribution using dual-energy window method or convolution subtraction. Correct according to acquisition parameters and radionuclide half-life. By scanning the planar source or cylindrical phantom with known activity, establish the system sensitivity factor (counts / s / MBq), convert SPECT pixel value to absolute activity concentration (Bq / mL), and generate three-dimensional activity distribution map.

[0070] Intelligent source term processing module, in communication connection with the image processing and individualized modeling module, for receiving radionuclide selection instructions of the user interface, and generating a highly optimized equivalent source term for representing the energy deposition characteristics of its entire decay chain according to the user-selected alpha radionuclide, to replace the time-consuming step-by-step decay simulation. Its core database includes radionuclide database and dose point kernel database.

[0071] Among them, the radionuclide database is stored in XML or SQLite format, which is used to store radionuclide information such as 225 Ac, 223 Ra, 211Complete decay chain information for various alpha nuclides such as At, including daughter nuclide types, half-life, decay branching ratio, and alpha particle energy; the dose point nucleus database is pre-calculated and stored using PHITS / Geant4 Monte Carlo simulation. The absorbed dose point nucleus in the dose point nucleus database is established in the following ways: (1) Collect alpha particles of different energies released by the decay of all alpha nuclides in the range of 4-10 MeV; (2) Construct various standard tissue models for clinical needs, such as soft tissue, bone (cortical bone), kidney, tumor, and red bone marrow. Each tissue model has different elemental composition and physical density. (3) Define an isotropic point source at the geometric center of each tissue model. Set a series of high-density, concentric spherical shells around the point source as statistical regions for energy deposition. The radius of these spherical shells starts from 1 μm and gradually increases with logarithmic or linear intervals until they completely cover the maximum range of alpha particles. (4) Perform independent Monte Carlo simulations for each tissue type and each energy point to finally obtain the absorbed dose point nucleus. .

[0072] Its decision-making and computational logic includes:

[0073] Receiving user input: The user selects a nuclide from a drop-down menu in the graphical interface, such as " 225 Ac”.

[0074] Model selection strategy: If the nuclide decay chain length is ≤2 and the daughter nuclide half-life is extremely short (<1 second), then the equivalent source term optimization model is used, merging the short-lived daughter nuclide and the parent nuclide into an equivalent alpha particle source. If the nuclide decay chain length is >2 and the daughter nuclide half-life is comparable to or longer than that of the parent nuclide (e.g., ...), then the model selection strategy is changed. 225 If Ac), then the dose point kernel scaling model (DPK scaling model) is used.

[0075] DPK scaling model execution (with) 225 (Ac for example):

[0076] The system reads from the nuclide database 225 All α emitters in the Ac decay chain ( 225 Ac, 221 Fr, 217 At, 213 Bi, 213 Branch ratio of Po) and alpha particle energy .

[0077] From the DPK base database, call and interact with various... and the absorbed dose point nucleus corresponding to the current voxel tissue type .

[0078] Based on the long-term radioactive balance assumption, the equivalent total dose point kernel is generated by weighting and superimposing the absorption dose point kernels of each alpha emitter in the nuclide decay chain according to their decay branching ratios :

[0079] ;

[0080] where, is the decay branching ratio of the ith sub-nuclide; is the energy of the alpha particle emitted by the ith sub-nuclide.

[0081] Finally, the equivalent total dose point kernel is saved as a source term definition file for subsequent dose calculation by the dose calculation module.

[0082] The dose calculation module, in communication connection with the image processing and individualized modeling module and the intelligent source term processing module, is used to calculate a three-dimensional physical absorbed dose distribution based on the three-dimensional voxel anatomical model, the three-dimensional voxel-level activity distribution map, and the equivalent total dose point kernel; the processing process thereof includes:

[0083] Cumulative activity calculation: for each voxel, the least squares method is used to perform single exponential or double exponential function fitting on the activity values A(t) at multiple time points to obtain the time-activity curve A(t) of the voxel. Integrating A(t) from t=0 to t=∞, the total cumulative activity of the voxel is obtained , with the unit of MBq·h.

[0084]

[0085] Dose convolution calculation:

[0086] Given that the range of alpha particles (~80 μm) is much smaller than the image voxel size (1 mm), the energy deposition is limited within the source voxel. Therefore, the intra-voxel self-absorption approximation is adopted. The absorbed dose D of each voxel is directly calculated by the following formula:

[0087]

[0088] where, is the value of the equivalent total dose point kernel at the source point, representing the total energy deposited within the source voxel under unit cumulative activity of the entire decay chain.

[0089] This method avoids time-consuming three-dimensional convolution or particle transport simulation, shortening the calculation time from tens of hours to minutes.

[0090] The final output is a three-dimensional physical absorbed dose distribution matrix (unit: Gy), whose grid is consistent with the input voxel model.

[0091] The bio-dose calculation module, communicating with the dose calculation module, integrates a micro-dose kinetic model to convert the three-dimensional physical absorbed dose distribution into a three-dimensional equivalent bio-dose distribution map with conventional fractionated external beam radiotherapy as a reference standard. Its processing includes:

[0092] Inputs: The three-dimensional physical absorbed dose distribution matrix output by the dose calculation module; dose rate information; radiobiological parameters of the target area.

[0093] Application of microdose kinetic model: A modified linear quadratic model is used, with the cell survival fraction S calculated as follows:

[0094]

[0095] Among them, key parameters Determined by the following formula:

[0096]

[0097] In the formula, D is the physical absorbed dose. and These are the model parameters corrected using microdosimetry. The corrected dose mean is a quantity derived from microdosimetry theory, which is directly related to the LET distribution of alpha particles and can be obtained from literature or experimental data. It is used to quantify the high bioavailability of alpha particles. This represents the intrinsic sensitivity of cells to the "linear" component of their radiation-killing effect under low-dose or low-linear-energy-transfer radiation (such as X-rays or gamma rays). It is usually assumed that the reference photon illumination... The values ​​are equal. and It can be obtained from radiobiological experiments targeting specific cell lines, such as prostate cancer LNCaP cells.

[0098] Equivalent biological dose conversion: Substitute the physical dose D and mean dose rate of each voxel into the MK model to calculate the cell survival fraction S of that voxel. Then, through the inverse operation of a linear quadratic model, convert the cell survival fraction S into an equivalent biological dose. (Based on 2Gy / fractionation radiotherapy):

[0099]

[0100] in, and These are the radiobiological parameters of the reference photon.

[0101] The final output is a three-dimensional equivalent biological dose ( Distribution matrix (unit: Gy).

[0102] This invention constructs a patient-specific three-dimensional voxel anatomical model and integrates voxel-level activity distribution maps generated by quantitative SPECT / PET, thus fully preserving information on individual patient differences and tumor heterogeneity during calculations. For alpha particles with a range of only tens of micrometers, their energy deposition is highly concentrated within the source voxel. Traditional dose calculation methods based on macroscopic voxels are prone to "voxel averaging" errors because the voxel size (typically 1-2 mm) is much larger than the alpha particle range. This invention can accurately calculate microscopic dose changes caused by uneven activity distribution and different tissue densities (such as tumors and adjacent bone). This avoids the risk of underestimating the dose in cold tumor areas or misestimating the dose in hot areas of high-risk organs due to the "homogenization" assumption, providing an unprecedentedly accurate data foundation for predicting treatment efficacy and toxicity.

[0103] Furthermore, this invention integrates a time integration algorithm optimized for limited static images into the dose integration module. This algorithm can accurately calculate the total cumulative activity over the entire treatment cycle based on a single or a few (e.g., 2-3) routine clinical static SPECT / CT images through robust curve fitting. This simplifies the data acquisition scheme from the ideal 5-7 time points to a clinically feasible 3 time points, reducing the total scan time from over 60 minutes to less than 1 hour. This not only reduces the direct economic cost of the examination and equipment time cost by approximately 60%, but also significantly improves patient tolerance and compliance, enabling personalized and precise dose assessment to be seamlessly integrated into routine clinical workflows, greatly promoting the widespread application of this technology.

[0104] In addition, by introducing a smart source term processing module, based on the characteristics of the nuclide decay chain ( 225 (Describing the complexity of Ac), the system intelligently selects and applies a dose point kernel scaling model or an equivalent source term optimization model. Through a pre-computation database and the principle of linear superposition, it generates an "equivalent total dose point kernel," simplifying the complex decay chain simulation into efficient convolution or dot product operations. This makes it possible to obtain a comprehensive dose report in a very short time after treatment, achieving "near real-time" clinical dose feedback and completely changing the predicament of traditional methods being unable to guide immediate clinical decision-making.

[0105] By integrating micro-dose kinetic models, the three-dimensional physical dose distribution and dose rate information are converted into a three-dimensional equivalent biological dose distribution map. This conversion allows oncologists to directly use their rich clinical experience in radiotherapy (such as the tumor control probability TCP and normal tissue complication probability NTCP models) to interpret the dose results of TAT, thereby making more accurate judgments on the treatment effect and the risk of normal tissue damage, and formulating truly individualized treatment strategies and follow-up plans.

[0106] Embodiment two, the application also discloses a voxel model-based targeted alpha nuclide treatment equivalent biological dose determination method, which comprises the following steps: 225 Ac-PSMA treatment of metastatic castration-resistant prostate cancer, comprising the following steps:

[0107] System initialization and data preparation: start the voxel model-based targeted alpha nuclide treatment equivalent biological dose determination system described in embodiment one. Load the necessary databases, including: tissue material library (containing the element composition and density of soft tissue, bone, lung and other tissues), alpha nuclide decay database, pre-calculated dose point kernel database, and radiobiological parameter library (containing the alpha / beta values of different cell lines, 、 and other parameters).

[0108] S100: image acquisition, processing and individualized modeling;

[0109] S101: clinical image acquisition:

[0110] A metastatic castration-resistant prostate cancer patient is intravenously injected with 225 Ac-PSMA-617, and after treatment, whole-body scanning is performed at three time points of 4 hours (t1), 24 hours (t2) and 72 hours (t3) after administration using a SPECT / CT system with a medium-energy collimator. Among them,

[0111] The CT scanning parameters are: tube voltage 120 kV, automatic tube current modulation, layer thickness 1.0 mm, pitch 1.0. The acquired CT data are used for anatomical positioning and attenuation correction. The SPECT acquisition parameters are: energy window setting is 218 keV ( 225 Ac daughter body 221 Fr characteristic X-ray) ± 10%. The image reconstruction is performed using an iterative reconstruction algorithm (such as OSEM, 10 iterations, 4 subsets).

[0112] S102: image preprocessing and registration:

[0113] The SPECT reconstructed images at the three time points are respectively rigidly registered with the high-resolution CT images acquired at the t1 time point, so as to ensure that all functional images are located in the same anatomical coordinate system.

[0114] Quantitative correction is performed on the SPECT images; specifically including:

[0115] Attenuation correction; the attenuation map (μ-map) generated by the CT data is used for correction.

[0116] Scattering correction; the double-energy window method is used for scattering estimation and deduction.

[0117] Resolution recovery; introduce point spread function model in iterative reconstruction process for partial correction.

[0118] Activity calibration; convert the reconstructed SPECT pixel value (counts) to absolute activity concentration value by a calibration factor (unit: Bq / mL / counts) which is previously calibrated by a known activity phantom, finally generate three-dimensional activity distribution map A(x,y,z,t1), A(x,y,z,t2), A(x,y,z,t3), unit Bq / mL.

[0119] S103: Individualized voxel model construction:

[0120] Read the registered t1 time CT image data into the system as a three-dimensional matrix. Resample the original CT image voxels to a uniform isotropic voxel grid of 1mm x 1mm x 1mm.

[0121] According to the CT value (Hounsfield Unit, HU) of each voxel, the system automatically performs tissue segmentation, assigns the voxel to a predefined tissue type, for example:

[0122] HU>300: defined as cortical bone, assigned with corresponding element composition and density of 1.92 g / cm³.

[0123] 50< HU ≤ 300: defined as soft tissue (muscle, organ parenchyma), assigned with soft tissue element composition and density of 1.04 g / cm³.

[0124] -1000< HU ≤-200: defined as lung tissue, assigned with lung tissue element composition and density of 0.26 g / cm³.

[0125] At the same time, the doctor draws the tumor target area (such as prostate cancer pelvic lymph node metastasis) and risk organs (such as bilateral kidneys, salivary glands, liver, red bone marrow) on the CT image, and generates corresponding three-dimensional volume of interest labels.

[0126] Finally, a patient-specific three-dimensional voxel anatomy model containing geometric information (position, size), material information (tissue type, density) and structure information (ROI label) is generated.

[0127] S200: Intelligent source item generation;

[0128] S201: Nuclide identification and model selection:

[0129] The user selects the treatment nuclide as "Ac-225" in the system interface. After receiving the instruction, the intelligent source item processing module queries the built-in α nuclide decay database, identifies 225 Ac is a nuclide with long and complex decay chain ( 225 Ac→221 Fr → 217 At→ 213 Bi → 213 Po → 209 Pb → 209 Bi), and automatically activates and executes the calculation path of the dose point kernel scaling model.

[0130] S202: DPK scaling model implementation:

[0131] Decay chain information reading: Precise reading from a nuclide database 225 All α emitter nuclei in the Ac decay chain ( 225 Ac, 221 Fr, 217 At, 213 Bi, 213 The half-life and decay branching ratio of Po ) and the energy of the alpha particles it releases ( ).

[0132] Basic DPK call: The system accesses its pre-computed dose point kernel base database. This database is pre-built through the following steps:

[0133] 1. Construct various standard tissue models (such as soft tissue, bone (cortical bone), kidney, tumor, and red bone marrow, etc.);

[0134] 2. At the geometric center of each tissue model, an isotropic alpha particle point source is set;

[0135] 3. Use Monte Carlo simulation software (such as PHITS or Geant4) to simulate the transport of alpha particles in tissues at different energies (covering the range of 4-10 MeV);

[0136] 4. Around the alpha particle point source, set up a series of concentric spherical statistical grids starting from 1 μm and extending to cover the maximum range of the alpha particle (approximately 80 μm).

[0137] 5. Record the energy deposited within each spherical shell and calculate the absorbed dose point of the alpha particle in that tissue. That is, the dose rate at a distance r from the point source.

[0138] DPK superposition: Based on the physical assumption that daughter and parent radionuclides achieve long-term radioactive equilibrium during treatment (i.e., equilibrium factor). ≈1), the basic DPKs of all α-emitting daughter nuclei in the decay chain are linearly superimposed according to their branch ratios to generate a representative... 225 The equivalent total dose point nucleus of the entire Ac decay chain :

[0139] ;

[0140] where i runs through all the alpha emitting daughter nuclei, and are the decay branching ratio and alpha particle energy of the ith daughter nucleus, respectively. This is saved as a data file as the optimized source term definition for subsequent calculation.

[0141] S300: Absorbed dose calculation based on dose point kernels;

[0142] S301: Cumulative activity calculation:

[0143] Read the activity distribution maps A(t1), A(t2), A(t3) generated in step S100 at three time points.

[0144] For each voxel, the system takes its activity values at different time points, and performs a least square fitting with a double exponential function to obtain the time-activity curve equation of the voxel (in practical application, if some voxels have too large statistical noise leading to non-converged fitting, the system will automatically degrade to single exponential function fitting; if there are only two valid time points, the trapezoidal method is used for approximate integration. These strategies ensure the stable operation of the system under various data quality in clinical practice):

[0145]

[0146] where, and are the exponential decay factors, where the parameter λ is the effective clearance constant. Usually associated with a fast process. This process is often the early distribution of the drug, fast clearance or initial metabolism. For example, the drug is quickly cleared from the blood to the tissue, or the rapid washout for some tissues. Usually associated with a slow process. This process represents the final retention and clearance phase of the drug. For drugs targeting tumors, this usually reflects the specific binding and biological retention of the drug to the target (such as PSMA of prostate cancer cells). This slow process directly determines the deposition of most of the radiation dose. Represents the proportion or size of the part of the activity involved in the fast process at the initial time. Represents the proportion or size of the part of the activity involved in the slow process at the initial time. Describes the dynamics of the fast changing part of the drug in the body. Describes the dynamics of the slow changing part of the drug in the body.

[0147] Integrate the TAC of each voxel from time zero to infinity to calculate the total cumulative activity of the voxel, unit: MBq·h.

[0148]

[0149] S302: Dose convolution calculation:

[0150] Since the range of alpha particles is much smaller than the size of the image voxel (1 mm), their energy is almost completely deposited inside the source voxel, so the in- voxel self-absorption calculation can be performed without energy deposition convolution between voxels.

[0151] By adopting a simplified form of the dose convolution method, the total cumulative activity of each voxel, A(x, y, z), is multiplied by the equivalent total dose point kernel generated in step S200 The value at "r = 0" (i.e., the dose coefficient for energy completely deposited inside the source point) is multiplied, and the three-dimensional physical absorbed dose distribution D(x, y, z) of each voxel is quickly calculated.

[0152]

[0153] where, represents the total energy deposited inside the source voxel under the unit cumulative activity of the entire decay chain. This step avoids time-consuming full Monte Carlo simulation, and the calculation can be completed within a few minutes.

[0154] S400: Equivalent biological dose calculation;

[0155] S401: Physical dose preprocessing:

[0156] Since 225 The contribution of beta and gamma rays in the Ac decay chain can be ignored compared to alpha particles, and the total absorbed dose D(x, y, z) can be approximately equal to the contribution dose of alpha particles .

[0157] S402: Microdose kinetics model calculation:

[0158] The built-in microdose kinetics model is called. For the prostate cancer cell line LNCaP, its radiobiological parameters (such as the alpha / beta ratio of reference photons , , where can be taken as a typical value according to the literature).

[0159] The cell survival fraction S of the microdose kinetics model is calculated according to the formula:

[0160]

[0161] where, ≈ ; ; where, is the corrected dose mean, which is the core microscopic physical quantity representing the quality of α particle radiation. Its value can be obtained from the α particle energy spectrum and LET value of Ac from literature or experimental database. 225 Ac's α particle energy spectrum and its LET value are obtained from literature or experimental database.

[0162] The corrected dose mean of each voxel is calculated by the following formula: The average dose rate (estimated by total dose divided by effective treatment time) is substituted into the above formula to calculate the survival fraction S of representative cells in the voxel.

[0163] S403: Equivalent biological dose conversion:

[0164] The calculated cell survival fraction S is converted into equivalent biological dose EQD2 with conventional fractionated external beam radiotherapy (2 Gy / fraction) as reference by inverse operation of the linear quadratic model.

[0165]

[0166] wherein, is the linear term coefficient under reference photon irradiation. The system performs this calculation for all voxels, and finally generates a three-dimensional equivalent biological dose (EQD2) distribution map.

[0167] S500: Result output and display;

[0168] The final three-dimensional physical absorbed dose distribution and three-dimensional equivalent biological dose (EQD2) distribution are superimposed on the CT anatomical image of the patient in the form of a color cloud map, and are subjected to multi-planar reconstruction and three-dimensional rendering in the display and analysis module. Thus, the physician can view the dose-volume histogram within the ROI of the tumor target and the risk organ. The average dose of the tumor and and the maximum point dose of the normal organ (such as the kidney) and other key dosimetric parameters are obtained. Based on the three-dimensional equivalent biological dose (EQD2) distribution, combined with the clinical experience of external beam radiotherapy, the tumor control probability and normal tissue complication probability are evaluated, which provides accurate and intuitive quantitative basis for the effectiveness and safety evaluation of the current treatment and the adjustment of the subsequent treatment plan.

[0169] The above examples are only used to illustrate the technical solutions of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A voxel model based targeted alpha nuclide treatment isoeffect biodosimetry system, characterized by: Comprise: an image processing and individualized modeling module for importing and processing computed tomography (CT) images and single photon emission computed tomography (SPECT) or positron emission tomography (PET) images of a patient, generating a patient-specific three-dimensional voxel anatomical model and a three-dimensional voxel-level activity distribution map at at least one time point; an intelligent source term processing module in communication with the image processing and individualized modeling module for receiving a user interface nuclide selection instruction and generating an equivalent total dose point kernel representing energy deposition characteristics of an entire decay chain of an alpha nuclide based on pre-stored alpha nuclide decay chain data; a dose calculation module in communication with the image processing and individualized modeling module and the intelligent source term processing module for calculating a three-dimensional physical absorbed dose distribution based on the three-dimensional voxel anatomical model, the three-dimensional voxel-level activity distribution map and the equivalent total dose point kernel; a biological dose calculation module in communication with the dose calculation module for integrating a microdose kinetics model to convert the three-dimensional physical absorbed dose distribution into a three-dimensional equivalent biological dose distribution map with a conventional fractionated external radiotherapy as a reference standard.

2. The voxel model based targeted alpha-nucleus therapy equivalent biological dose determination system of claim 1, wherein: The intelligent source term processing module comprises: a nuclide database for storing decay chain information of multiple alpha nuclides, the decay chain information including daughter nuclide species, half-life, decay branching ratio and alpha particle energy; a dose point kernel base database for storing pre-calculated absorbed dose point kernels of alpha particles of different energies in different standard tissue models through Monte Carlo simulation; a model selection and calculation unit for selecting and executing a corresponding simplified calculation model to generate the equivalent total dose point kernel according to the complexity of the decay chain of the selected alpha nuclide.

3. The voxel model based targeted alpha-nucleus therapy equivalent biological dose determination system of claim 2, wherein: The model selection and calculation unit is configured to: For alpha nuclides with complex decay chains, a dose point kernel scaling model is performed to generate an equivalent total dose point kernel by weighting and superimposing the absorption dose point kernels of each alpha emitting daughter nucleus in the nuclide decay chain according to their decay branching ratios : ； wherein, is the decay branching ratio of the ith sub-nucleus; is the alpha particle energy released by the ith sub-nucleus; is the absorbed dose point kernel of corresponding energy and tissue type called from the dose point kernel database.

4. The voxel model based targeted alpha-nucleus therapy equivalent biological dose determination system of claim 2, wherein: The absorbed dose point kernels in the dose point kernel base database are established by: constructing multiple standard tissue models, each tissue model having a specific element composition and physical density; defining an isotropic alpha particle point source at the geometric center of each tissue model; The point kernel function of absorbed dose versus distance is obtained by Monte Carlo simulation of the energy deposition in different radius spherical shells centered on an alpha particle point source .

5. The voxel model based targeted alpha-nucleus therapy equivalent biological dose determination system of claim 3, wherein: The image processing and individualized modeling module integrates a time integration algorithm for calculating the total cumulative activity of a voxel based on limited static imaging data; the image processing and individualized modeling module is configured to: curve fitting activity values measured at multiple time points for the same voxel to obtain a time-activity curve of the voxel; integrating the time-activity curve from zero time to infinity time to calculate the total cumulative activity of the voxel .

6. The voxel model based targeted alpha-nucleus therapy equivalent biological dose determination system of claim 5, wherein: The dose calculation module calculates the three-dimensional physical absorbed dose distribution using a dose convolution method; the dose calculation module is configured to: The total accumulated activity of each voxel is calculated by multiplying the total activity of each voxel by the total time of the scan The equivalent total dose point kernel is calculated by multiplying the total accumulated activity of each voxel by the total time of the scan The absorbed dose of each voxel is calculated by multiplying the equivalent total dose point kernel by the total time of the scan .

7. The voxel model based targeted alpha-nucleus therapy equivalent biological dose determination system of claim 1, wherein: The biological dose calculation module is configured to: call a microdose kinetics model to calculate a cell survival fraction S according to radiobiological parameters of target cells and the physical absorbed dose distribution; calculate the three-dimensional equivalent biological dose distribution map through inverse operation of a linear quadratic model based on the cell survival fraction S and biological parameters of a reference radiotherapy.

8. A voxel model based targeted alpha nuclide treatment isoeffect biodosimetry method, characterized by: Comprise the following steps: S100: image acquisition, processing and individualized modeling; acquiring single photon emission computed tomography (SPECT) and computed tomography (CT) fusion images or positron emission tomography (PET) and CT fusion images of a patient at at least two time points after injection of a targeted alpha-emitting radionuclide; segmenting and resampling the CT images to construct a patient-specific three-dimensional voxel anatomical model, wherein each voxel contains tissue type, density and spatial location information; performing attenuation correction, scatter correction and activity calibration on the SPECT or PET images, and spatially registering the SPECT or PET images with the CT images to generate a three-dimensional voxel-level activity distribution map spatially aligned with the three-dimensional voxel anatomical model; S200: intelligent source term generation; receiving a user-specified alpha-emitting radionuclide; querying a pre-stored radionuclide database to obtain complete decay chain data of the alpha-emitting radionuclide, the decay chain data including half-lives, decay branching ratios and emitted alpha particle energies of each daughter radionuclide; intelligently selecting a simplified calculation model according to the complexity of the decay chain of the alpha-emitting radionuclide; generating an equivalent total dose point kernel representing the spatial distribution of energy deposition of the entire decay chain of the alpha-emitting radionuclide under unit cumulative activity based on the selected simplified calculation model and a pre-stored dose point kernel database; S300: absorbed dose calculation based on dose point kernels performing curve fitting on activity values of the same voxel at different time points by a time integration algorithm based on the three-dimensional activity distribution map, and integrating the fitted time-activity curve to calculate the total cumulative activity of each voxel; performing spatial energy deposition calculation on the total cumulative activity using the equivalent total dose point kernel generated in step S200 to obtain a three-dimensional physical absorbed dose distribution; S400: equivalent biological dose calculation; inputting the three-dimensional physical absorbed dose distribution and corresponding dose rate information obtained in step S300 into a microdose kinetics model; calculating a cell survival fraction by the microdose kinetics model based on radiobiological parameters of the target region; converting the cell survival fraction into a three-dimensional equivalent biological dose distribution map with conventional fractionated external beam radiotherapy as a reference standard; S500: result output; outputting and displaying the three-dimensional equivalent biological dose distribution map and / or the three-dimensional physical absorbed dose distribution for clinical treatment evaluation.

9. The voxel model based targeted alpha-nucleus therapy equivalent biological dose determination method of claim 8, wherein: In the S200 step, when the alpha-emitting radionuclide is a complex decay chain radionuclide with multiple alpha-emitting daughter nuclei, the simplified calculation model is a dose point kernel scaling model; The specific steps of generating the equivalent total dose point kernel include: calling dose point kernels corresponding to energies and tissue types of each alpha-emitting daughter nucleus in the decay chain from the dose point kernel database, and performing weighted superposition according to the decay branching ratios, with the formula being: ； wherein, is the equivalent total dose point nucleus, is the decay branching ratio of the i-th sub-nucleus; is the energy of the alpha particle released by the i-th sub-nucleus; is the corresponding base dose point nucleus.

10. The method of claim 8, wherein the voxel model-based targeted alpha-nuclide therapy equivalent biological dose determination method is characterized by: In the step S400, the microdose kinetics model is an improved model based on a linear quadratic model, and a calculation formula of a cell survival fraction S is: ; where D is the physical absorbed dose, and are the model parameters corrected by microdosimetric parameters.

Images