Rapid dose assessment method and device for personalized targeted radionuclide therapy

Abstract

This application relates to the field of dosimetry technology, and in particular to a method and apparatus for rapid dose assessment in personalized targeted radionuclide therapy. The method includes: acquiring the patient's labeled radionuclide, drug ligand, and image data; selecting corresponding standard S-values ​​based on the labeled radionuclide and drug ligand; segmenting the first-phase CT image data to obtain a segmentation mask for the target organ; calculating the individualized S-value based on the segmentation mask and generating an individualized voxel model; registering SPECT image data and CT image data to obtain organ-level source term parameters and voxel-level source term parameters; and obtaining organ-level dose assessment results and voxel-level dose assessment results based on the organ-level and voxel-level source term parameters, respectively, to comprehensively evaluate the dose distribution of the source and target organs. This solves problems such as the lack of calibration of dose conversion coefficients and the difficulty in balancing calculation speed and accuracy in voxel models, reduces errors caused by individual differences, and ensures calculation accuracy and speed.

Description

Technical Field

[0001] This application relates to the field of dosimetry technology, and in particular to a method and apparatus for rapid dose assessment of individualized targeted radionuclide therapy. Background Technology

[0002] Personalized precision medicine is a medical model that customizes treatment strategies based on patients' physiological characteristics and disease heterogeneity. It can drive the medical model from "general application to the population" to "optimal application to the individual," achieving efficient use of resources and innovation in treatment paradigms.

[0003] Targeted radionuclide therapy (TRT) has become an important means of cancer treatment due to its high selectivity for tumor lesions and is one of the key development directions of personalized precision medicine. Against this backdrop, radiation dosimetry, as the cornerstone of precision treatment, provides crucial scientific evidence for optimizing treatment regimens and balancing tumor response rates with organ toxicity by quantifying the distribution and energy deposition of radiopharmaceuticals in the target area and normal tissues. Studies show that patients in the personalized dosimetry group experienced an increase in the average tumor absorbed dose of approximately 107.1 Gy, a median overall survival extended to 26.6 months, and an overall response rate improved to 50%, representing a significant improvement compared to the standard dosimetry group, without increasing the risk of adverse reactions. With the development of targeted radionuclide therapy and therapeutic imaging, accurate patient-specific dose measurement is becoming increasingly necessary to meet the needs of these advanced treatment methods.

[0004] Personalized internal radiation dose assessment systems are a core tool for achieving precise control of patient radiation dose. However, related technologies still face many challenges: (1) the dose conversion coefficient lacks localized calibration (e.g., insufficient adaptation of organ parameters based on European and American populations); (2) rapid calculation methods in related technologies ignore individual differences; (3) efficiency and accuracy are difficult to balance (high-resolution voxel model calculations are too time-consuming); (4) fragmented multimodal technology collaborative frameworks and asynchronous data flow. Therefore, developing a system architecture that can efficiently generate personalized voxel models and balance calculation speed and accuracy has significant clinical value for achieving personalized dose optimization. Summary of the Invention

[0005] This application provides a method, device, electronic device, and storage medium for rapid dose assessment of individualized targeted radionuclide therapy, which addresses issues such as the lack of localized calibration of dose conversion coefficients, the neglect of individual differences by rapid calculation methods, and the low computational efficiency of high-resolution voxel models. It reduces systematic errors caused by racial differences, shortens computation time while ensuring computational accuracy, and lowers deployment costs.

[0006] The first aspect of this application provides a method for rapid dose assessment of individualized targeted radionuclide therapy, comprising the following steps:

[0007] Acquire the patient's labeled radionuclide, drug ligand, and imaging data, wherein the imaging data includes SPECT imaging data and CT imaging data;

[0008] The corresponding standard S value is obtained from a preset population-specific SAF database based on the labeled nuclide and drug ligand, wherein the preset population-specific SAF database is constructed based on a reference human model of a preset region;

[0009] The first-phase CT image data is segmented to obtain a segmentation mask of the target organ. Based on the segmentation mask, an individualized S-value is calculated according to the standard S-value and the target correction factor. An individualized voxel model of the patient is generated based on the segmentation mask.

[0010] The SPECT image data and the CT image data are registered to obtain organ-level source term parameters and voxel-level source term parameters. Based on the individualized S-value, organ-level dose assessment results are generated according to the organ-level source term parameters using the preset MIRD calculation method and the preset MCs calculation method. The voxel-level source term parameters and the individualized voxel model are input into the preset fast Monte Carlo model to obtain voxel-level dose assessment results. The target dose assessment results are obtained according to the organ-level dose assessment results and the voxel-level dose assessment results.

[0011] Optionally, in some embodiments, obtaining the corresponding standard S value from a preset population-specific SAF database based on the labeled nuclide and drug ligand includes:

[0012] Based on the labeled nuclide and the drug ligand, determine whether the preset SAF database contains a standard S value that matches the labeled nuclide and the drug ligand;

[0013] If the standard S value exists in the preset SAF database, and the version of the standard S value meets the preset version verification conditions, then the standard S value is output.

[0014] Optionally, in some embodiments, after determining whether a standard S-value matching the labeled nuclide and the drug ligand exists in the preset SAF database, the process includes:

[0015] If the standard S value is not found in the preset SAF database, the particle energy spectrum and branching ratio are obtained from the preset radionuclide decay database based on the labeled nuclide and the drug ligand, and a new S value is calculated based on the particle energy spectrum and the branching ratio. The new S value is then added to the preset population-specific SAF database.

[0016] Optionally, in some embodiments, calculating the individualized S-value based on the segmentation mask, according to the standard S-value and the target correction factor, includes:

[0017] The actual mass of the target organ is calculated based on the segmentation mask, and a target correction factor is determined based on the actual mass and a preset standard mass.

[0018] The individualized S-value is calculated based on the target correction factor and the standard S-value.

[0019] Optionally, in some embodiments, registering the SPECT image data and the CT image data to obtain organ-level source term parameters and voxel-level source term parameters includes:

[0020] The multi-temporal CT image data are registered with the first-temporal CT image data as a reference to obtain the phase transformation matrix;

[0021] Based on the phase conversion matrices, the SPECT image data and the CT image data are registered to obtain the multi-phase count rate of each voxel, and based on the preset count-activity conversion factor, the multi-phase count rate of each voxel is converted into an activity-time curve.

[0022] The organ-level source term parameters and the voxel-level source term parameters are obtained based on the activity-time curve.

[0023] A second aspect of this application provides a rapid dose assessment device for personalized targeted radionuclide therapy, comprising:

[0024] The first acquisition module is used to acquire the patient's labeled radionuclide, drug ligand, and image data, wherein the image data includes SPECT image data and CT image data;

[0025] The second acquisition module is used to acquire the corresponding standard S value from a preset population-specific SAF database based on the labeled nuclide and drug ligand, wherein the preset population-specific SAF database is constructed based on a reference human model of a preset region.

[0026] The segmentation module is used to segment the first-phase CT image data to obtain a segmentation mask of the target organ. Based on the segmentation mask, an individualized S-value is calculated according to the standard S-value and the target correction factor, and an individualized voxel model of the patient is generated based on the segmentation mask.

[0027] The evaluation module is used to register the SPECT image data and the CT image data to obtain organ-level source term parameters and voxel-level source term parameters. Based on the individualized S-value, it uses a preset MIRD calculation method and a preset MCs calculation method to generate organ-level dose evaluation results according to the organ-level source term parameters. It also inputs the voxel-level source term parameters and the individualized voxel model into a preset fast Monte Carlo model to obtain voxel-level dose evaluation results. Finally, it obtains target dose evaluation results based on the organ-level dose evaluation results and the voxel-level dose evaluation results.

[0028] Optionally, in some embodiments, the second acquisition module includes:

[0029] The judgment unit is used to determine whether a standard S value matching the labeled nuclide and the drug ligand exists in the preset SAF database based on the labeled nuclide and the drug ligand.

[0030] The output unit is used to output the standard S value when the standard S value exists in the preset SAF database and the version of the standard S value meets the preset version verification conditions.

[0031] Optionally, in some embodiments, after determining whether a standard S-value matching the labeled nuclide and the drug ligand exists in the preset SAF database, the determining unit includes:

[0032] The update subunit is configured to, when the standard S value does not exist in the preset SAF database, obtain the particle energy spectrum and branching ratio from the preset radionuclide decay database based on the labeled radionuclide and the drug ligand, calculate a new S value based on the particle energy spectrum and the branching ratio, and add the new S value to the preset population-specific SAF database.

[0033] Optionally, in some embodiments, the segmentation module includes:

[0034] A determining unit is configured to calculate the actual mass of the target organ based on the segmentation mask, and determine a target correction factor based on the actual mass and a preset standard mass.

[0035] A calculation unit is used to calculate the individualized S-value based on the target correction factor and the standard S-value.

[0036] Optionally, in some embodiments, the evaluation module includes:

[0037] The first registration unit is used to register multi-temporal CT image data with the first temporal CT image data as a reference to obtain the phase transformation matrix;

[0038] The second registration unit is used to register the SPECT image data and the CT image data based on the phase conversion matrices to obtain the multi-phase count rate of each voxel, and to convert the multi-phase count rate of each voxel into an activity-time curve based on a preset count-activity conversion factor.

[0039] The generation unit is used to obtain the organ-level source term parameters and the voxel-level source term parameters based on the activity-time curve.

[0040] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a rapid dose assessment method for individualized targeted radionuclide therapy as described in the above embodiments.

[0041] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a rapid dose assessment method for individualized targeted radionuclide therapy as described in the above embodiments.

[0042] Therefore, this application has the following beneficial effects:

[0043] (1) This application addresses the population anatomical bias, clinical high-load computational time bottleneck, and novel targeted drugs (such as...) that exist in traditional dose conversion systems. 225 Ac-PSMA, 64 To meet the rapid translation needs from preclinical studies to Phase III trials of Cu-DOTATATE, this application's embodiment constructs a Chinese population-specific dose conversion coefficient library, which achieves multi-dimensional optimization through a hierarchical heterogeneous architecture design. This architecture maintains the rigor of MIRD paradigm-based dose calculations for marketed targeted drugs while significantly improving the dose evaluation efficiency of novel targeted drugs in preclinical and Phase III trials.

[0044] (2) The individualized dose conversion coefficient calculation method proposed in this application realizes the migration from a population model to individualized calculation through multimodal image fusion technology. This technical approach ensures clinical operability while correcting for changes in dose conversion coefficients caused by anatomical variations in individual patients, and helps to improve the accuracy of patient dose evaluation.

[0045] (3) This application innovatively integrates heterogeneous computing and dynamic source term reconstruction technology to achieve a breakthrough in the efficiency and accuracy of nuclear medicine dose assessment. The GPU-accelerated Monte Carlo system based on the CUDA architecture reduces the computation time for a single case from 60 hours to 0.5 hours, an acceleration of more than 80 times. Through a dual strategy of CT baseline elastic registration and SPECT rigid registration, the problem of spatiotemporal mismatch in multi-phase images is overcome, and the voxel activity integration time is reduced to 1 / 10 of the traditional method by combining piecewise fitting-trapezoidal integral algorithm. The fully automated process significantly reduces the time spent on clinical operations, combining high efficiency and clinical applicability, and provides a systematic solution for precision diagnosis and treatment.

[0046] (4) This application integrates multiple differentiated dose assessment schemes. Clinicians can perform rapid organ dose assessment and conduct precise sub-organ dose assessment based on patient-specific voxel models, achieving a dual improvement in clinical efficiency and computational accuracy. By deeply integrating core technologies such as image processing, activity modeling, and dose calculation, an automated dose assessment system is established. The system significantly reduces the human-computer interaction requirements in the source information generation and dose calculation stages, and all calculation modules have achieved individualized algorithm optimization, providing reliable technical support for precision nuclear medicine diagnosis and treatment.

[0047] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0048] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0049] Figure 1 This is a flowchart of a method for rapid dose assessment of individualized targeted radionuclide therapy provided according to embodiments of this application;

[0050] Figure 2 This is a schematic diagram illustrating the principle of constructing a pre-defined population-specific SAF database according to an embodiment of this application;

[0051] Figure 3 This is a schematic diagram illustrating the principle of calculating an individualized S-value according to an embodiment of this application;

[0052] Figure 4 This is a schematic diagram illustrating the principle of a preset fast mapping model provided according to an embodiment of this application;

[0053] Figure 5 This is a schematic diagram of a multimodal tightly coupled system architecture provided according to an embodiment of this application;

[0054] Figure 6This is a block diagram of a rapid dose assessment device for personalized targeted radionuclide therapy provided according to an embodiment of this application;

[0055] Figure 7 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0056] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0057] Before introducing the rapid dose assessment method for individualized targeted radionuclide therapy according to the embodiments of this application, let's first introduce the relevant technical principles.

[0058] In current clinical practice, patient dose assessment is primarily based on the MIRD (Medical Internal Radiation Dose) method, also known as the organ S-value method. This method obtains the average organ dose by multiplying the cumulative organ activity by the organ S-value (which describes the average absorbed dose per unit of cumulative activity in the target area) provided in the MIRD manual. However, its core limitations are twofold: First, the organ S-value is calculated based on a standard reference human model, ignoring the anatomical differences and spatial heterogeneity of radionuclide distribution among individual patients, leading to systematic errors in individualized dose assessment. Second, the reference human model upon which the relevant S-value calculation relies is based on typical physical parameters of Caucasians, and this lack of anatomical adaptation and localized correction further amplifies the systematic errors in dose assessment between tumor target areas and normal tissues. Taking the liver as an example, the mass of the liver in the Chinese male reference human model is 1.46 kg, while the mass of the liver in the ICRP (International Commission on Radiological Protection) male reference human model is 2.36 kg, resulting in a calculation bias of approximately 10%-25% in the photon absorption fraction.

[0059] Monte Carlo simulations (MCs) are considered the gold standard for reliable dose calculation in clinical settings. Combined with advanced medical molecular imaging techniques (e.g., PET / CT and SPECT / CT), they can accurately acquire individualized voxel-level three-dimensional dose distribution maps for each patient. However, their slow computation speed and high cost limit their application in routine clinical procedures. Related technologies have improved accuracy and efficiency by combining various resolutions and energy cutoffs to optimize simulation parameters, but these still require approximately 60 hours of computation time. This high computational load is difficult to adapt to clinical scenarios, especially given the large patient population in China, making practical application infeasible.

[0060] The Voxel S-Value (VSV) method treats each voxel as an individual source voxel and a target voxel of adjacent voxels. It uses dose-conversion coefficients (MCs) to calculate the voxel S-value (the dose conversion coefficient for a specific nuclide and voxel size) and dose kernel matrix M. Through convolution with the three-dimensional source term distribution data, it accurately calculates the energy transfer and deposition within each voxel. Due to its low computational cost and high speed, it is considered an efficient alternative to MCs. VSV calculations typically assume that the point source is located in an infinitely homogeneous soft tissue or aqueous medium. However, because it does not consider the heterogeneity of tissues and organs, it often introduces errors at tissues or tissue boundaries with significant differences in density from water.

[0061] Current mainstream nuclear medicine dosimetry systems (such as HERMES MEDICAL SOLUTIONS, MIM, PMOD, and QDOSE) generally employ the publicly disclosed technical frameworks mentioned above to achieve dose prediction functions. However, they commonly suffer from a disconnect between algorithm performance and clinical needs: On the one hand, while the Monte Carlo method can ensure the accuracy of dose calculation, its massive computational load leads to single-case analyses taking several hours (e.g., HERMES); on the other hand, while fast analytical algorithms (such as the MIRD method and voxel S-value method) can shorten the time to minutes, their simplistic assumptions about heterogeneous tissues and non-uniform activity distribution introduce systematic biases (e.g., QDOSE and MIM). Furthermore, due to commercial data control purposes, the core servers of these systems are deployed overseas, requiring the cross-border transmission of sensitive patient images and dose data, which is susceptible to privacy leaks due to vulnerabilities or malicious attacks on overseas servers.

[0062] Personalized dosing calculations for targeted radiopharmaceuticals rely on a multimodal technology framework. Its core lies not only in optimizing dosing algorithms but also in the organic coupling of key modules such as precise medical image segmentation, dynamic modeling of personalized voxel models (tissue density and activity distribution mapping), and pharmacokinetic parameter inversion based on multi-timepoint molecular imaging data. However, related technologies generally suffer from modular fragmentation (lack of data interface standards leads to difficulties in cross-platform migration), high operational complexity (requiring manual switching between multiple software programs and parameter resetting), and an imbalance between computational efficiency and accuracy (most integrated platforms use fast analytical algorithms instead of Monte Carlo simulations to shorten process time), severely limiting their clinical applicability.

[0063] In summary, the relevant technologies have the following drawbacks:

[0064] (1) Data model mismatch: dual deviation between population characteristics and individualized parameters.

[0065] I. Population bias: The dose conversion factor (such as the S value) depends on the Caucasian anatomical database and differs significantly from key parameters such as organ volume in Asian populations, resulting in large dose assessment errors.

[0066] II. Individual inaccuracies: The anatomical parameters of the reference standard human model are fixed, and the dose conversion coefficient changes caused by individual patient anatomical variations are not corrected.

[0067] (2) Limitations of the computational model: the contradiction between the organ-level uniformity assumption and the voxel-level non-uniformity requirement.

[0068] I. Lack of spatial distribution information: Most relevant dosimetry systems are based on the assumption that radionuclides are uniformly distributed within organs, using cumulative organ activity as the source term input to calculate organ-level uniform dose distribution results. This ignores the non-uniform spatial distribution characteristics of high expression in the tumor target area, leading to underestimation of target dose and the risk of undertreatment due to overestimation of organ dose. There is a lack of dose calculation based on voxel-level non-uniform source terms (such as SPECT / CT multi-time point activity distribution).

[0069] (3) Imbalance between algorithm performance and accuracy: the opposition between Monte Carlo and analytical methods.

[0070] I. High precision comes at a cost: Although Monte Carlo simulation can support voxel-level dose calculation, the calculation time for a single case can be as long as 60 hours, and it requires high-cost hardware such as CPU clusters.

[0071] II. Defects of Fast Algorithms: Although analytical algorithms (MIRD method, voxel S-value method) compress the time to the minute level, they simplify tissue non-homogeneity (CT value-density mapping error > 10%) and geometric approximation, causing systematic bias.

[0072] (4) Fragmented system architecture: Modular redundancy and hardware dependence are the two obstacles.

[0073] I. Fragmented process: Modules such as organ segmentation, voxel modeling, and pharmacokinetic parameter inversion are scattered and independent, requiring manual data export and format conversion, which leads to long operation time and increases the risk of data loss.

[0074] II. High hardware costs: To meet the general Monte Carlo acceleration computing requirements, a multi-node CPU cluster needs to be deployed, which far exceeds the affordability of primary healthcare institutions.

[0075] To address the aforementioned issues, this application provides a rapid dose assessment method for individualized targeted radionuclide therapy. In this method, the patient's labeled radionuclide, drug ligand, and image data are acquired. Based on the labeled radionuclide and drug ligand, corresponding standard S-values ​​are obtained from a pre-defined population-specific SAF database. First-phase CT image data is segmented to obtain a segmentation mask for the target organ. Based on the segmentation mask, an individualized S-value is calculated according to the standard S-value and a target correction factor. An individualized voxel model of the patient is then generated based on the segmentation mask. SPECT and CT image data are registered to obtain organ-level and voxel-level source term parameters. Based on the individualized S-value, an organ-level dose assessment result is generated using pre-defined MIRD and MCs calculation methods based on the organ-level source term parameters. The voxel-level source term parameters and the individualized voxel model are input into a pre-defined rapid Monte Carlo model to obtain a voxel-level dose assessment result. Finally, the target dose assessment result is obtained based on the organ-level and voxel-level dose assessment results. This solves the problems of lack of population suitability, imbalance between computational accuracy and efficiency, and discretization of system architecture in dose assessment. It can reduce systematic errors caused by racial differences, shorten computation time, and reduce deployment costs while ensuring computational accuracy.

[0076] Specifically, Figure 1 This is a flowchart illustrating a rapid dose assessment method for individualized targeted radionuclide therapy provided in an embodiment of this application.

[0077] like Figure 1 As shown, this method for rapid dose assessment of personalized targeted radionuclide therapy includes the following steps:

[0078] In step S101, the patient's labeled radionuclide, drug ligand, and imaging data are acquired, wherein the imaging data includes SPECT imaging data and CT imaging data.

[0079] SPECT image data can be replaced with PET image data.

[0080] In step S102, the corresponding standard S value is obtained from a preset population-specific SAF database based on the labeled nuclide and drug ligand. The preset population-specific SAF database is constructed based on a reference human model of a preset region.

[0081] Preferably, in this application embodiment, the preset population-specific SAF database is a dose conversion coefficient library for the Chinese population pre-established by relevant personnel, and the preset region's reference human model is a Chinese reference human model pre-established by relevant personnel.

[0082] It should be noted that the dose conversion factor S-value is the ratio of the absorbed dose to the source organ activity per unit cumulative activity over time. The calculation process for the S-value is as follows:

[0083] (a) Divide the energy range of 10 keV-4 MeV into 200 equally logarithmically spaced energy points, and construct specific absorption fractions for monoenergetic photons, electrons, and alpha, respectively. (Specific Absorption Fraction, SAF). SAF is the ratio of the energy absorbed by the target organ to its mass, which can be calculated using the following formula.

[0084]

[0085] in, target organs R The ratio of absorption fraction, For deposition in target organs R Energy in MeV. The initial energy (MeV) for particles emitted by the source organ. The mass (kg) of the target organ.

[0086] (ii) Based on the decay energy spectrum and branching ratio of the radionuclide, SAF is accumulated to obtain the target organ. R The organ S value is determined according to the following formula.

[0087]

[0088] in, Source organ to target organ R of S value, This refers to the particle energy (MeV) produced by a single decay of a radioactive nuclide. For this energy ray decay branching ratio, This indicates that the nuclide was produced. i A particle of energy.

[0089] In clinical applications, internal radiation dose is first administered by quantitatively obtaining time-distribution data of radioactivity in the source organ using PET / SPECT molecular imaging technology, followed by calculation of cumulative activity based on a biokinetic model. The final target organ dose is obtained by the following formula:

[0090]

[0091] in, For target organ dose, The cumulative activity of the source organ. For source organs j target organs R of S value.

[0092] In the calculation process of S-value, SAF is usually calculated based on the Monte Carlo Simulation (MCs) method. The accuracy of its particle sampling and transport process is directly constrained by anatomical parameters such as organ geometry, spatial distribution characteristics, and mass density.

[0093] Given this characteristic, establishing a digital phantom with representative population anatomy is crucial for improving the accuracy of dose calculation. The dose conversion coefficient study for the Chinese population utilizes a domestically developed Chinese Reference Adult Male / Female voxel model (CRAM / CRAF). This model's height, weight, and major organ mass meet relevant technical requirements, and the spatial topological relationships of the organs conform to the statistical distribution of anatomical characteristics of the Chinese population. This dose conversion coefficient, based on a localized phantom, can effectively reduce dose assessment bias caused by anatomical differences and significantly improve the rationality of dose calculation when combined with the MIRD methodology framework.

[0094] Optionally, in some embodiments, obtaining the corresponding standard S value from a preset population-specific SAF database based on the labeled nuclide and drug ligand includes: determining whether a standard S value matching the labeled nuclide and drug ligand exists in the preset SAF database based on the labeled nuclide and drug ligand; if a standard S value exists in the preset SAF database and the version of the standard S value meets the preset version verification conditions, then outputting the standard S value.

[0095] Furthermore, in some embodiments, after determining whether a standard S value matching the labeled nuclide and drug ligand exists in the preset SAF database, the method includes: if a standard S value does not exist in the preset SAF database, obtaining the particle energy spectrum and branching ratio from the preset nuclide decay database based on the labeled nuclide and drug ligand, calculating a new S value based on the particle energy spectrum and branching ratio, and adding the new S value to the preset population-specific SAF database.

[0096] Specifically, the preset population-specific SAF database in this application embodiment adopts a hybrid architecture design, and realizes data updates through a dynamic iteration mechanism that combines offline pre-storage and online computation, as shown in the attached figure. Figure 2 As shown, the system architecture consists of two core modules.

[0097] For the offline pre-stored module, a pre-stored population-specific SAF database is provided, pre-stored with SAF matrices calculated based on CRAM phantoms; complete decay parameters of radionuclides recommended by ICRP Publication 107 are included; and commonly used clinical radiopharmaceuticals (such as...) are integrated. 177 The Lu-PSMA organ S-value dataset, through the out-of-the-box call of standardized dose conversion coefficients, significantly shortens the routine targeted radiopharmaceutical dose assessment cycle.

[0098] For the online calculation module, an adaptive calculation process is constructed: the user submits the nuclide and ligand, the system retrieves the list of matching organ S values ​​in the pre-stored library, and if they exist and pass the version verification, the dose conversion coefficient in standard format is directly output.

[0099] If no target combination is found, an online report is returned. The coverage integrity of the CRAM phantom organ database is then verified and the SAF matrix offline database is optimized. For defined source organs, the nuclide decay database in ICRP Report 107 is called to obtain the particle energy spectrum and branching ratio. The target organ dose coefficient is generated based on the SAF cumulative equation. Finally, the new calculation results are dynamically incorporated into the preset population-specific SAF database after statistical verification.

[0100] In this application embodiment, the issues of population anatomical feature bias, clinical high-load computational time bottleneck, and novel targeted drugs (such as...) in traditional dose conversion systems are addressed. 225 Ac-PSMA, 64 To meet the need for rapid translation from preclinical studies to Phase III trials, the Chinese population-specific dose conversion coefficient library constructed in this application has achieved multi-dimensional optimization through a hierarchical heterogeneous architecture design. This architecture, while maintaining the rigor of the MIRD paradigm in calculating the dosage of marketed targeted drugs, significantly improves the dose evaluation efficiency of novel targeted drugs in preclinical and Phase III trials.

[0101] In step S103, the first-phase CT image data is segmented to obtain a segmentation mask of the target organ. Based on the segmentation mask, an individualized S-value is calculated according to the standard S-value and the target correction factor, and an individualized voxel model of the patient is generated based on the segmentation mask.

[0102] Optionally, in some embodiments, based on the segmentation mask, calculating an individualized S-value according to a standard S-value and a target correction factor includes: calculating the actual mass of the target organ based on the segmentation mask; determining the target correction factor based on the actual mass and a preset standard mass; and calculating the individualized S-value based on the target correction factor and the standard S-value.

[0103] Specifically, the individualized S-value calculation proposed in this application embodiment realizes the migration from a group model to individualized calculation through multimodal image fusion technology.

[0104] like Figure 3 As shown, firstly, based on the first-phase CT image data in DICOM format, the system automatically analyzes the patient-specific organ topology using a deep learning segmentation engine (3D U-Net architecture). Human-machine collaborative correction is then applied to the automatic segmentation results: the physician manually corrects the contours of key organs (such as ectopic thyroid glands), and the automatic / manual segmentation masks are fused using an algorithm to generate a segmentation contour of the target organ that conforms to the three-dimensional contours of the source / target organs. Subsequently, a pre-stored organ S-value matrix is ​​invoked, and the tissue density database of ICRP Publication 110 is retrieved simultaneously.

[0105] Based on the patient-source / target organ mask, the total organ mass is calculated using the voxel mass integral formula, as follows:

[0106] ;

[0107] in, For the actual quality of the target organ, The density of the material corresponding to the voxel. For voxel position labeling, It is a voxel volume element.

[0108] Subsequently, anatomical difference compensation calculations for organ S-values ​​were performed, and an organ mass ratio correction factor was established. The formula is as follows:

[0109] ;

[0110] in, The target correction factor, For organ mass in the CRAM model, The actual quality of the patient's organs.

[0111] Finally, the organ-standard S-value based on the CRAM model is multiplied by a correction factor. This approach obtains individualized organ S-values. By correcting for dose conversion coefficient variations caused by individual patient anatomical differences, this technique ensures clinical operability and helps improve the accuracy of patient dose assessment.

[0112] In step S104, SPECT image data and CT image data are registered to obtain organ-level source term parameters and voxel-level source term parameters. Based on the individualized S-value, organ-level dose assessment results are generated according to the organ-level source term parameters using the preset MIRD calculation method and the preset MCs calculation method. The voxel-level source term parameters and the individualized voxel model are input into the preset fast Monte Carlo model to obtain voxel-level dose assessment results. The target dose assessment results are obtained based on the organ-level dose assessment results and the voxel-level dose assessment results.

[0113] Furthermore, in some embodiments, registering SPECT image data and CT image data to obtain organ-level source term parameters and voxel-level source term parameters includes: registering multi-temporal CT image data with the first-temporal CT image data as a reference to obtain phase transformation matrices; based on the phase transformation matrices, registering SPECT image data and CT image data to obtain multi-temporal count rates for each voxel, and converting the multi-temporal count rates of each voxel into activity-time curves based on a preset count-activity conversion factor; and obtaining organ-level source term parameters and voxel-level source term parameters based on the activity-time curves.

[0114] It should be noted that the preset quick guessing model is as follows: Figure 4 As shown, this application's embodiment presents a GPU-accelerated Monte Carlo dosing system developed based on a heterogeneous computing architecture. It utilizes the Visual Studio 2022 integrated development environment and the CUDA 10.0 toolchain, achieving multi-level optimization through C++ / CUDA C hybrid programming. Based on the characteristics of memory, thread hierarchy, and hardware hierarchy in the execution model of the GPU graphics card's CUDA programming model, the program framework, particle transport, step size sampling, data structures, and cross-section access were designed and optimized. While maintaining the integrity of the physical process, accelerated by the RTX 3090, it achieves a speedup of over 50 times compared to traditional MCNP and Geant4 single-core simulations, reducing the simulation time for a single case from 60 hours to less than 0.5 hours.

[0115] The computational accuracy of the fast Monte Carlo method directly depends on the quality of the input individualized voxel model and the dynamic source term distribution. To this end, a multi-temporal SPECT / CT data processing module was constructed: based on DICOM format sequential image data, the system first calls a deep learning segmentation network to extract the first-temporal CT image data to obtain the target organ contour, and generates the patient's individualized voxel model through Hounsfield unit-tissue density piecewise linear transformation.

[0116] In constructing the voxel-level dynamic source term distribution, the system needs to overcome two major technical bottlenecks: first, the spatial mismatch of multi-temporal CT images caused by patient position changes and organ deformation during treatment; second, the lack of anatomical structural information in SPECT functional images, and the gradual decrease in drug accumulation in the lesion area as treatment progresses, leading to inaccurate direct registration of cross-temporal functional images. To address the challenge of multi-temporal source term reconstruction, the system employs a dual registration strategy: using the first temporal CT as a reference, an itk-based elastic registration framework is employed, and subsequent temporal CT images are aligned to the same coordinate system using a built-in image processing toolbox; for SPECT functional images, functional-anatomical image registration is achieved based on the established CT temporal transformation matrix, ultimately obtaining the multi-temporal count rate for each voxel. After completing spatiotemporal registration, the count rate of each voxel is converted into an activity-time curve based on the count-activity conversion factor.

[0117] In radiopharmaceutical kinetic modeling, for scenarios with uniform activity distribution within organs, the count rates of all voxels within a specific volume of interest (VOI) are summed to generate organ-level time-activity curves. Subsequently, a nonlinear least squares method is used to fit a single / double exponential decay model, and finally, the time-integrated activity coefficient is obtained through numerical integration to obtain organ-level source term parameters. This process is simple to calculate and has a short computation time.

[0118] However, the reconstruction of source terms with non-uniform voxel distribution faces a step increase in computational complexity, particularly at the voxel level (voxel dimensions are typically around 10). 6 Fitting and integrating (a) samples requires significant computational power, storage space, and time. Therefore, this application's embodiments employ analytical calculation and trapezoidal integral techniques, with the calculation formula as follows:

[0119] ;

[0120] in, To accumulate activity, For the number of time points, For the i At that moment for The activity level below, This is the attenuation coefficient.

[0121] Thus, the voxel-level source term parameters are calculated, and the overall computation time is reduced to 1 / 10 of that of the double exponential fitting.

[0122] Based on coupled data from individualized voxel models and dynamic source term mappings, the system executes GPU-accelerated Monte Carlo simulations after the user sets key parameters through the interactive interface. Upon completion of the simulation, the system outputs a three-dimensional dose distribution matrix with the same resolution as SPECT and automatically generates organ dose reports. The dose-volume histogram (DVH) generated based on voxel-level non-uniform source terms can quantify the spatial heterogeneity distribution of dose within organs, overcoming the statistical limitations of traditional uniform distribution models. Combined with the GPU-accelerated Monte Carlo algorithm, the overall computational efficiency is improved by hundreds of times, providing crucial support for individualized radiopharmaceutical dosing plans.

[0123] In addition, embodiments of this application can also establish such as Figure 5 The multimodal tightly coupled system architecture shown in this application addresses the significant technical hurdles of targeted radiopharmaceutical dosage evaluation, an interdisciplinary field integrating biomolecular imaging, radiation dosimetry, imaging medicine, and computational physics. Driven by increased production capacity of medical radionuclides and the localization of high-end imaging equipment, nuclear medicine is undergoing a paradigm shift from empirical, extensive diagnosis and treatment to precise quantitative assessment. This transformation places stringent demands on nuclear medicine physicians to integrate multidisciplinary knowledge. Traditional dosage evaluation systems, with their multi-platform switching operations (involving multiple independent modules such as multimodal image fusion, biodynamic model construction, and dosage calculation), result in single assessments taking several hours, severely hindering clinical decision-making efficiency. The multimodal tightly coupled intelligent system constructed in this application achieves full-process automation of dosage evaluation through breakthroughs in three core technologies. This system significantly reduces the intensity of manual interaction, compressing individualized dosage evaluation time to less than one hour, providing real-time support for the dynamic optimization of clinical treatment plans.

[0124] The first stage of the multimodal tightly coupled system employs deep learning to achieve automatic registration and anatomical structure segmentation of multimodal images. This algorithm supports dual hardware modes (CPU and GPU), achieving high-speed computation on GPUs with >8GB of video memory. Currently, segmentation of 117 organs and tissue structures takes less than 4 minutes. The system is equipped with an intelligent interactive interface, integrating 3D image visualization VOI cropping and correction tools to assist clinicians in visually verifying and manually correcting tumor target areas and automatic segmentation results.

[0125] For multi-temporal nuclear medicine imaging data, the system automatically calls the image processing module in core technology 3 to achieve accurate registration of multi-temporal nuclear medicine images, and converts the registered data into a three-dimensional activity distribution map. Simultaneously, it constructs a patient-specific voxel model to lay a personalized data foundation for subsequent dose calculation.

[0126] The system automatically acquires cumulative activity data of source / target organs and regions of interest (VOIs) through its built-in exponential fitting and integral calculation modules, and simultaneously outputs source term parameters in both organ-level (uniform distribution) and voxel-level (non-uniform distribution) modes. This module achieves fully automated generation of source term information, completely avoiding the manual parameter conversion step in traditional processes.

[0127] The system integrates three differentiated dose assessment schemes: a localized improved MIRD standard method, direct MCs using a CRAM model, and a GPU-accelerated Monte Carlo high-precision algorithm. Clinicians can perform rapid dose comparisons between the MIRD method and the OLIDA system, and also conduct sub-organ-level precise dose assessments based on patient-specific voxel models, achieving a dual improvement in clinical efficiency and computational accuracy.

[0128] This application's embodiments establish an automated dose assessment system by deeply integrating core technologies such as image processing, activity modeling, and dose calculation. Compared to similar foreign platforms, the system significantly reduces the human-computer interaction requirements in the source information generation and dose calculation stages, and all calculation modules have achieved localized algorithm optimization, providing reliable technical support for precision nuclear medicine diagnosis and treatment.

[0129] Therefore, the embodiments of this application solve the problem of multimodal data flow interruption, integrate modules such as intelligent image segmentation, multimodal registration, cumulative activity assessment, dose calculation, and patient information management, construct an automated and tightly coupled workflow, achieve seamless cross-platform data transmission through standardized protocols, and ultimately form an integrated closed-loop system covering "data input-dose modeling-clinical decision-making", providing efficient and reliable technical support for individualized precision treatment.

[0130] According to the embodiment of this application, the method for rapid dose assessment of individualized targeted radionuclide therapy involves acquiring the patient's labeled radionuclide, drug ligand, and image data. Based on the labeled radionuclide and drug ligand, the corresponding standard S-value is obtained from a pre-defined population-specific SAF database. First-phase CT image data is segmented to obtain a segmentation mask for the target organ. Based on the segmentation mask, an individualized S-value is calculated according to the standard S-value and the target correction factor. An individualized voxel model of the patient is then generated based on the segmentation mask. SPECT image data and CT image data are registered to obtain organ-level source term parameters and voxel-level source term parameters. Based on the individualized S-value, an organ-level dose assessment result is generated using a pre-defined MIRD calculation method and a pre-defined MCs calculation method based on the organ-level source term parameters. The voxel-level source term parameters and the individualized voxel model are input into a pre-defined rapid Monte Carlo model to obtain a voxel-level dose assessment result. Finally, the target dose assessment result is obtained based on the organ-level dose assessment result and the voxel-level dose assessment result. This solves the problems of lack of population suitability, imbalance between computational accuracy and efficiency, and discretization of system architecture in dose assessment. It can reduce systematic errors caused by racial differences, shorten computation time, and reduce deployment costs while ensuring computational accuracy.

[0131] Next, referring to the accompanying drawings, a rapid dose assessment device for individualized targeted radionuclide therapy according to an embodiment of this application is described.

[0132] Figure 6 This is a block diagram of a rapid dose assessment device for individualized targeted radionuclide therapy according to an embodiment of this application.

[0133] like Figure 6 As shown, the personalized targeted radionuclide therapy dose rapid assessment device 10 includes: a first acquisition module 100, a second acquisition module 200, a segmentation module 300, and an assessment module 400.

[0134] The first acquisition module 100 is used to acquire the patient's labeled radionuclide, drug ligand and image data, wherein the image data includes SPECT image data and CT image data.

[0135] The second acquisition module 200 is used to acquire the corresponding standard S value from a preset population-specific SAF database based on the labeled nuclide and drug ligand, wherein the preset population-specific SAF database is constructed based on a reference human model of a preset region.

[0136] The segmentation module 300 is used to segment the first-phase CT image data to obtain the segmentation mask of the target organ. Based on the segmentation mask, the individualized S-value is calculated according to the standard S-value and the target correction factor, and the individualized voxel model of the patient is generated according to the segmentation mask.

[0137] The evaluation module 400 is used to register SPECT image data and CT image data to obtain organ-level source term parameters and voxel-level source term parameters. Based on the individualized S-value, it uses the preset MIRD calculation method and the preset MCs calculation method to generate organ-level dose evaluation results according to the organ-level source term parameters. It also inputs the voxel-level source term parameters and the individualized voxel model into the preset fast Monte Carlo model to obtain voxel-level dose evaluation results. Finally, it obtains the target dose evaluation results based on the organ-level dose evaluation results and the voxel-level dose evaluation results.

[0138] Optionally, in some embodiments, the second acquisition module 200 includes a judgment unit and an output unit.

[0139] The judgment unit is used to determine whether a standard S value matching the labeled nuclide and drug ligand exists in the preset SAF database based on the labeled nuclide and drug ligand.

[0140] The output unit is used to output the standard S value when the standard S value exists in the preset SAF database and the version of the standard S value meets the preset version verification conditions.

[0141] Optionally, in some embodiments, after determining whether a standard S value matching the labeled nuclide and drug ligand exists in the preset SAF database, the determination unit includes: an update subunit.

[0142] The update subunit is used to obtain the particle energy spectrum and branching ratio from the preset radionuclide decay database based on the labeled radionuclide and drug ligand when the standard S value does not exist in the preset SAF database, calculate the new S value based on the particle energy spectrum and branching ratio, and add the new S value to the preset population-specific SAF database.

[0143] Optionally, in some embodiments, the segmentation module 300 includes a determination unit and a calculation unit.

[0144] The determining unit is used to calculate the actual quality of the target organ based on the segmentation mask, and to determine the target correction factor based on the actual quality and the preset standard quality.

[0145] The calculation unit is used to calculate the individualized S-value based on the target correction factor and the standard S-value.

[0146] Optionally, in some embodiments, the evaluation module 400 includes: a first registration unit, a second registration unit, and a generation unit.

[0147] The first registration unit is used to register multi-phase CT image data with the first phase CT image data as a reference to obtain the phase transformation matrix.

[0148] The second registration unit is used to register SPECT image data and CT image data based on each phase transformation matrix to obtain the multi-phase count rate of each voxel, and to convert the multi-phase count rate of each voxel into an activity-time curve based on a preset count-activity conversion factor.

[0149] The generation unit is used to obtain organ-level source term parameters and voxel-level source term parameters based on the activity-time curve.

[0150] It should be noted that the explanation of the aforementioned embodiment of the rapid dose assessment method for individualized targeted radionuclide therapy also applies to the rapid dose assessment device for individualized targeted radionuclide therapy in this embodiment, and will not be repeated here.

[0151] According to the embodiments of this application, the device for rapid dose assessment of individualized targeted radionuclide therapy acquires the patient's labeled radionuclide, drug ligand, and image data. Based on the labeled radionuclide and drug ligand, it retrieves the corresponding standard S-values ​​from a preset population-specific SAF database. The device segments the first-phase CT image data to obtain a segmentation mask for the target organ. Based on the segmentation mask, it calculates an individualized S-value according to the standard S-value and the target correction factor. Then, it generates an individualized voxel model of the patient based on the segmentation mask. It registers SPECT and CT image data to obtain organ-level and voxel-level source term parameters. Based on the individualized S-value, it uses preset MIRD and MCs calculation methods to generate organ-level dose assessment results based on the organ-level source term parameters. The voxel-level source term parameters and the individualized voxel model are input into a preset rapid Monte Carlo model to obtain voxel-level dose assessment results. Finally, it obtains the target dose assessment results based on the organ-level and voxel-level dose assessment results. This solves the problems of lack of population suitability, imbalance between computational accuracy and efficiency, and discretization of system architecture in dose assessment. It can reduce systematic errors caused by racial differences, shorten computation time, and reduce deployment costs while ensuring computational accuracy.

[0152] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0153] The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0154] When the processor 702 executes the program, it implements the method for rapid dose assessment of individualized targeted radionuclide therapy provided in the above embodiments.

[0155] Furthermore, electronic devices also include:

[0156] Communication interface 703 is used for communication between memory 701 and processor 702.

[0157] The memory 701 is used to store computer programs that can run on the processor 702.

[0158] The memory 701 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0159] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0160] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0161] The processor 702 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0162] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for rapid dose assessment of individualized targeted radionuclide therapy.

[0163] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that 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. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0164] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0165] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0166] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0167] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0168] 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. A method for rapid dose assessment in personalized targeted radionuclide therapy, characterized in that, Includes the following steps: Acquire the patient's labeled radionuclide, drug ligand, and imaging data, wherein the imaging data includes SPECT imaging data and CT imaging data; The corresponding standard S value is obtained from a preset population-specific SAF database based on the labeled nuclide and drug ligand, wherein the preset population-specific SAF database is constructed based on a reference human model of a preset region; The first-phase CT image data is segmented to obtain a segmentation mask of the target organ. Based on the segmentation mask, an individualized S-value is calculated according to the standard S-value and the target correction factor. An individualized voxel model of the patient is generated based on the segmentation mask. The SPECT image data and the CT image data are registered to obtain organ-level source term parameters and voxel-level source term parameters. Based on the individualized S-value, organ-level dose assessment results are generated according to the organ-level source term parameters using a preset MIRD calculation method and a preset MCs calculation method. The voxel-level source term parameters and the individualized voxel model are input into a preset fast Monte Carlo model to obtain voxel-level dose assessment results. Target dose assessment results are obtained based on the organ-level dose assessment results and the voxel-level dose assessment results. The step of obtaining the corresponding standard S-value from a preset population-specific SAF database based on the labeled nuclide and drug ligand includes: Based on the labeled nuclide and the drug ligand, determine whether the preset SAF database contains a standard S value that matches the labeled nuclide and the drug ligand; If the standard S value exists in the preset SAF database, and the version of the standard S value meets the preset version verification conditions, then the standard S value is output. After determining whether a standard S-value matching the labeled nuclide and the drug ligand exists in the preset SAF database, the process includes: If the standard S value is not found in the preset SAF database, the particle energy spectrum and branching ratio are obtained from the preset radionuclide decay database based on the labeled radionuclide and the drug ligand, and a new S value is calculated based on the particle energy spectrum and the branching ratio, and the new S value is added to the preset population-specific SAF database. The formula for calculating the new S value based on the particle energy spectrum and the branching ratio is as follows: in, target organs R of S value, S The value is the dose conversion factor. The energy (MeV) of the particle produced by a single decay of the labeled nuclide. The decay branching ratio of the labeled nuclide. For the production of the labeled nuclide i Particles of a certain energy The specific absorption fraction of R in the target organ. For the target area, For the source region.

2. The method according to claim 1, characterized in that, The step of calculating the individualized S-value based on the segmented mask, according to the standard S-value and the target correction factor, includes: The actual mass of the target organ is calculated based on the segmentation mask, and a target correction factor is determined based on the actual mass and a preset standard mass. The individualized S-value is calculated based on the target correction factor and the standard S-value.

3. The method according to claim 1, characterized in that, The process of registering the SPECT image data and the CT image data to obtain organ-level source term parameters and voxel-level source term parameters includes: The multi-temporal CT image data are registered with the first-temporal CT image data as a reference to obtain the phase transformation matrix; Based on the phase conversion matrices, the SPECT image data and the CT image data are registered to obtain the multi-phase count rate of each voxel, and based on the preset count-activity conversion factor, the multi-phase count rate of each voxel is converted into an activity-time curve. The organ-level source term parameters and the voxel-level source term parameters are obtained based on the activity-time curve.

4. A rapid dose assessment device for personalized targeted radionuclide therapy, characterized in that, include: The first acquisition module is used to acquire the patient's labeled radionuclide, drug ligand, and image data, wherein the image data includes SPECT image data and CT image data; The second acquisition module is used to acquire the corresponding standard S value from a preset population-specific SAF database based on the labeled nuclide and drug ligand, wherein the preset population-specific SAF database is constructed based on a reference human model of a preset region. The segmentation module is used to segment the first-phase CT image data to obtain a segmentation mask of the target organ. Based on the segmentation mask, an individualized S-value is calculated according to the standard S-value and the target correction factor, and an individualized voxel model of the patient is generated based on the segmentation mask. The evaluation module is used to register the SPECT image data and the CT image data to obtain organ-level source term parameters and voxel-level source term parameters. Based on the individualized S-value, it uses a preset MIRD calculation method and a preset MCs calculation method to generate organ-level dose evaluation results according to the organ-level source term parameters. It then inputs the voxel-level source term parameters and the individualized voxel model into a preset fast Monte Carlo model to obtain voxel-level dose evaluation results. Finally, it obtains the target dose evaluation results based on the organ-level dose evaluation results and the voxel-level dose evaluation results. The second acquisition module includes: The judgment unit is used to determine whether a standard S value matching the labeled nuclide and the drug ligand exists in the preset SAF database based on the labeled nuclide and the drug ligand. The output unit is configured to output the standard S value when the standard S value exists in the preset SAF database and the version of the standard S value meets the preset version verification conditions. After determining whether a standard S-value matching the labeled nuclide and the drug ligand exists in the preset SAF database, the determining unit includes: The update subunit is used to obtain the particle energy spectrum and branching ratio from the preset radionuclide decay database based on the labeled radionuclide and the drug ligand when the standard S value does not exist in the preset SAF database, calculate a new S value based on the particle energy spectrum and the branching ratio, and add the new S value to the preset population-specific SAF database. The formula for calculating the new S value based on the particle energy spectrum and the branching ratio is as follows: in, target organs R of S value, S The value is the dose conversion factor. The energy (MeV) of the particle produced by a single decay of the labeled nuclide. The decay branching ratio of the labeled nuclide. For the production of the labeled nuclide i A type of energy particle, The specific absorption fraction of R in the target organ. For the target area, For the source region.

5. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method for rapid dose assessment of individualized targeted radionuclide therapy as described in any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the rapid dose assessment method for individualized targeted radionuclide therapy as described in any one of claims 1-3.

Images