Boron concentration distribution model acquisition method based on melanoma, terminal and medium

Abstract

The invention provides a melanoma-based boron concentration distribution model acquisition method, a terminal and a medium, and the method comprises the steps: obtaining a target region, a microneedle administration region and administration concentration of melanoma, and collecting PET image data of the target region at each sampling moment after microneedle administration; collecting micro-needle boron concentration data of the drug administration area at each sampling moment, and constructing a sample data set based on the PET image data and the corresponding micro-needle boron concentration data; based on the sample data set, executing iterative training of a boron concentration distribution model to obtain a trained boron concentration distribution model; according to the method provided by the invention, the simulation result of the boron concentration distribution corresponding to each moment can be quickly and conveniently obtained, the data resolution of the simulation result of the boron concentration distribution is effectively improved, the precision of the boron concentration distribution data is improved, and a data basis is provided for subsequent BNCT analysis and treatment scheme formulation.

Description

Technical Field

[0001] The present application belongs to the technical field of boron concentration acquisition, and relates to a method, terminal and medium for acquiring a boron concentration distribution model based on melanoma, and specifically to a method, terminal and medium for acquiring a boron concentration distribution model based on melanoma. Background Art

[0002] In the field of melanoma treatment, neutron capture therapy (BNCT) has unique advantages over traditional treatment methods (such as photons or carbon ions), including: (1) Melanoma is resistant to traditional photon irradiation, but not to BNCT, and BNCT is effective for both aerobic and anoxic tumor cells; (2) Boron drugs selectively accumulate in melanoma and non-melanoma tumor cells, and can accurately destroy cancer cells with complex shapes and reduce damage to normal tissues; (3) BNCT can deliver high radiation doses to cancer cells in a single irradiation, while traditional photon irradiation requires 6 to 7 weeks to complete the same dose; based on this, BNCT has great potential in the treatment of malignant cutaneous melanoma.

[0003] In the prior art, there are relatively few studies on the distribution of boron concentration in melanoma; although for other types of tumors, neutron excitation spectroscopy, mass spectrometry, positron emission tomography-computed tomography (PET-CT) and other technologies can be used to measure the boron concentration in tissues. However, due to the relatively scattered distribution of melanoma and the small area of ​​a single melanoma, the boron concentration obtained based on PET-CT has problems such as low resolution and long scanning time, which makes it impossible to accurately and quickly obtain high-resolution boron concentration distribution information in melanoma tissues at each moment. Moreover, the boron concentration data obtained based on PET-CT can only reflect the boron concentration distribution at a single moment, and it cannot characterize the spatiotemporal variation characteristics of boron concentration, thereby reducing the BNCT treatment effect of melanoma.

[0004] Therefore, how to accurately and quickly obtain the boron concentration distribution in the melanoma area has become a technical problem that needs to be solved in this field. Summary of the invention

[0005] In view of the shortcomings of the prior art described above, the purpose of the present application is to provide a method for obtaining a boron concentration distribution model based on melanoma, a boron concentration distribution acquisition method, a terminal and a medium, so as to solve the problem that the existing methods cannot quickly and accurately obtain the boron concentration distribution in the melanoma area.

[0006] In a first aspect, the present application provides a method for obtaining a boron concentration distribution model based on melanoma, comprising:

[0007] The target area, microneedle administration area and administration concentration of melanoma are obtained; the target area is a PET acquisition area of ​​boron concentration information, which covers and is larger than the range of the melanoma; the administration area is a boron drug diffusion area after microneedle administration; PET image data of the target area at each sampling time after microneedle administration, and microneedle boron concentration data of the administration area at each sampling time are collected; a sample data set is constructed based on the PET image data and the microneedle boron concentration data at the corresponding sampling time; based on the sample data set, iterative training is performed on a preset boron concentration distribution model to obtain a trained boron concentration distribution model; wherein the boron concentration distribution model is a model that simulates the spatial distribution of boron concentration in the target area at subsequent times based on the PET image data at an initial time.

[0008] In some embodiments, the iterative training, when performed once, includes:

[0009] Extract the image features of the microneedle boron concentration data in the current sample data pair; based on the image features and the image feature prediction results corresponding to the previous sample data pair, obtain the image feature prediction results corresponding to the current sample data pair through a time series prediction method; perform image reconstruction on the current image feature prediction results to obtain a boron concentration distribution simulation image corresponding to the current sample data pair; perform image comparison on the current boron concentration distribution simulation image and the PET image data in the current sample data pair to obtain the image difference between the two, so as to optimize the model parameters based on the image difference.

[0010] In some embodiments, when the iterative training is performed for the first time, the method for obtaining the image feature prediction result corresponding to the previous sample data includes:

[0011] The image features corresponding to the PET image data in the first sample data pair and the image features corresponding to the microneedle boron concentration data in the first sample data pair are extracted respectively; the image features of the two are fused; and the image feature prediction result corresponding to the fused feature is obtained by using the time series prediction method.

[0012] In some embodiments, the method of acquiring the image features includes:

[0013] Using a multi-scale convolution module, feature extraction is performed on each of the microneedle drug delivery data to obtain image features corresponding to each of the microneedle drug delivery data, which are:

[0014]

[0015] Among them, F Micro is the image feature of microneedle boron concentration data, MSCN() is the multi-scale convolution module; is the microneedle boron concentration data; and, using the residual module, extracting the spatial structural features in the PET image data in the first sample data pair to obtain the global spatial features of the PET image data, which is:

[0016]

[0017] Among them, F PET is the image feature corresponding to the PET image data, and ResNet() is the residual module; The PET image data is centered for the first sample data.

[0018] In some embodiments, the method for acquiring the image features further includes:

[0019] Adaptive feature alignment is performed on the image features of each modality, as follows:

[0020] F aligned =AFAM F PET ,F micro

[0021] Among them, F Micro is the image feature of the microneedle boron concentration data; F PET is the image feature corresponding to the PET image data; AFAM is an adaptive feature alignment module.

[0022] In some embodiments, the method for obtaining the microneedle boron concentration data includes:

[0023] In the drug administration area, each sampling point at different sampling distances is determined, and the boron concentration of each sampling point at different sampling times is obtained, so as to construct a sampling data pair after microneedle drug administration based on the sampling distance, sampling time and corresponding boron concentration; based on each of the sampling data pairs, a pre-constructed boron drug diffusion function is fitted to obtain a fitted microneedle boron concentration diffusion model; based on the sampling time of the PET image data, the microneedle boron concentration data corresponding to the sampling time is obtained using the fitted microneedle boron concentration diffusion model.

[0024] In some embodiments, for each sampling point within the microneedle coverage area, at the initial moment, the method for acquiring the microneedle boron concentration data includes:

[0025] Based on the volume of the drug solution and the boron concentration of the microneedle drug delivery, combined with the tissue mass of the sampling point, the boron concentration corresponding to the sampling point is obtained as follows:

[0026]

[0027] Among them, V is the volume of the drug solution for microneedle administration; C is the boron concentration for microneedle administration; and M is the tissue mass at the sampling point.

[0028] In a second aspect, the present application provides a method for obtaining boron concentration distribution based on melanoma, comprising:

[0029] Determine the dosing concentration of microneedle drug delivery, and based on the dosing concentration, determine a trained boron concentration distribution model corresponding to the dosing concentration; obtain PET image data of the target area at an initial time; based on the PET image data, use the boron concentration distribution model to obtain a corresponding boron concentration distribution simulation image at the target time; wherein the boron concentration distribution model is obtained using any of the melanoma-based boron concentration distribution model acquisition methods described above.

[0030] In a third aspect, the present application provides a terminal, comprising: a processor and a memory, wherein the memory is communicatively connected to the processor; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes any of the above-described methods for acquiring a boron concentration distribution model based on melanoma, or executes the above-described method for acquiring a boron concentration distribution based on melanoma.

[0031] In a fourth aspect, the present application provides a computer storage medium storing a computer program, which, when executed by a processor, implements any of the above-described methods for acquiring a boron concentration distribution model based on melanoma, or executes any of the above-described methods for acquiring a boron concentration distribution based on melanoma.

[0032] As described above, the boron concentration distribution model acquisition method based on melanoma, the boron concentration distribution acquisition method, the terminal and the medium provided in the present application, by constructing a sample data pair consisting of PET image data and microneedle boron concentration data at the corresponding moment, and using each sample data pair to iteratively train the boron concentration distribution model to obtain a trained boron concentration distribution model, so that based on the trained boron concentration distribution model, combined with the PET image data at the initial moment, a boron concentration distribution simulation image at a subsequent moment can be obtained; compared with the prior art, the method described in the present application makes full use of the measured boron concentration data after microneedle administration, and the microneedle boron concentration data obtained by fitting based on the measured boron concentration data, that is, on the basis of making full use of the law of microneedle boron concentration change, the spatial distribution characteristics of the boron drug concentration in the PET image data are considered, so that not only can the simulation results of the boron concentration distribution corresponding to each moment be obtained quickly and conveniently, but also the defect of low resolution of PET data is improved, the data resolution of the boron concentration distribution simulation results is effectively improved, and the accuracy of the boron concentration distribution data is improved, thereby providing a better data basis for subsequent BNCT analysis and calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1Shown is a schematic flow chart of the method for acquiring a melanoma-based boron concentration distribution model provided in an embodiment of the present application;

[0034] Figure 2 Shown is a schematic flow chart of a method for acquiring PET image data in an embodiment of the present application;

[0035] Figure 3 Shown is a schematic flow chart of a method for obtaining microneedle boron concentration data in an embodiment of the present application;

[0036] Figure 4 Shown is a flow chart of the iterative training process described in the embodiment of the present application when it is executed;

[0037] Figure 5 Shown is a schematic flow chart of the method for obtaining boron concentration distribution based on melanoma described in the embodiments of the present application;

[0038] Figure 6 Shown is a schematic diagram of the structure of the terminal described in the embodiment of the present application. DETAILED DESCRIPTION

[0039] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0040] It should be noted that in the following description, with reference to the accompanying drawings, several embodiments of the present application are described in the accompanying drawings. It should be understood that other embodiments may also be used, and mechanical composition, structure, electrical and operational changes may be made without departing from the spirit and scope of the present application. The following detailed description should not be considered restrictive, and the scope of the embodiments of the present application is limited only by the claims of the published patents. The terms used here are only for describing specific embodiments and are not intended to limit the present application. Spatially related terms, such as "upper", "lower", "left", "right", "below", "below", "lower", "above", "upper", etc., may be used in the text to facilitate the description of the relationship between an element or feature shown in the figure and another element or feature.

[0041] Furthermore, as used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It should be further understood that the terms "comprises", "includes", "includes" indicate the presence of the described features, operations, elements, components, items, types, and / or groups, but do not exclude the presence, occurrence, or addition of one or more other features, operations, elements, components, items, types, and / or groups. The terms "or" and "and / or" used herein are interpreted as inclusive, or mean any one or any combination.

[0042] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the technical solution in the embodiments of the present invention is further described in detail through the following embodiments and in combination with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the invention.

[0043] Before further describing the present invention in detail, the nouns and terms involved in the embodiments of the present invention are explained. The nouns and terms involved in the embodiments of the present invention are applicable to the following interpretations:

[0044] (1) PET-CT: Positron emission tomography / X-ray computed tomography is a combination of PET and CT. PET and CT are designed as one and controlled by a workstation. During scanning, PET and CT imaging are performed simultaneously as needed, and the workstation fuses the two images together to achieve better identification and positioning.

[0045] (2) Microneedle drug delivery, which uses micron-sized needle tips to pierce the stratum corneum of the skin, forming a temporary mechanical channel, releasing the drug directly into the epidermis or dermis, allowing the drug to enter the systemic circulation through the capillaries.

[0046] In order to solve the technical problems in the prior art, the present application provides a method for obtaining a boron concentration distribution model based on melanoma in a first aspect, which is used to obtain a boron concentration distribution model corresponding to the microneedle drug administration concentration under the microneedle drug administration method.

[0047] Among them, the boron concentration distribution model is a model used to simulate the spatial distribution of boron concentration in the area where the melanoma is located at subsequent moments based on the PET-CT image data at the initial moment (hereinafter referred to as PET image data); the initial moment is a moment when the microneedle is administered or after the microneedle is administered; the subsequent moment is a moment after the initial moment.

[0048] See also Figure 1 , showing a schematic flow chart of the method for acquiring a melanoma-based boron concentration distribution model provided in an embodiment of the present application; Figure 1 As shown, the method comprises the following steps:

[0049] S10, obtaining the target area of ​​melanoma, and determining the microneedle drug delivery area and drug delivery concentration of the same melanoma;

[0050] Wherein, the target area is a PET acquisition area of ​​boron concentration information, and the range of the acquisition area is larger than the range of the melanoma;

[0051] The drug administration area is the diffusion area of ​​the boron drug after microneedle administration, and the range of the diffusion area is larger than the area of ​​the melanoma;

[0052] Specifically, after determining the coverage area of ​​the melanoma, the coverage area is expanded according to a preset expansion ratio, and the expanded area is used as the target area to ensure that the target area can cover the entire melanoma area;

[0053] And, a dosing area for microneedle administration is constructed based on the microneedle administration position as the center and the diffusion radius of the boron drug for microneedle administration; wherein the diffusion radius of the boron drug is a preset value, which can be obtained based on experiments or experience.

[0054] It should be noted that the diffusion radius of the boron drug is positively correlated with the drug administration concentration of the microneedle drug administration, that is, the greater the drug administration concentration, the greater the diffusion radius of the boron drug, and vice versa.

[0055] S20, collecting PET image data of the target area at different sampling times, and collecting microneedle boron concentration data in the drug administration area corresponding to the sampling times;

[0056] Wherein, the PET image data is data acquired through PET acquisition and characterizes the spatial distribution information of boron concentration in the target area;

[0057] The microneedle boron concentration data is data characterizing the spatial distribution information of the boron concentration in the drug administration area after microneedle drug administration.

[0058] Specifically, according to the area range of the target area, PET images of the area range at several sampling moments after microneedle drug administration are collected to obtain PET image data of the target area;

[0059] And, according to the area range of the drug administration area, the microneedle boron concentration data of the area range corresponding to each of the sampling moments is obtained.

[0060] In the present application, the microneedle boron concentration data corresponds to the PET image data, and the sampling times of the two are the same or close, that is, the sampling time difference between the two is less than a time difference threshold.

[0061] In order to enable the PET image data to accurately and comprehensively reflect the dynamic distribution of boron concentration over time after microneedle administration, in a specific embodiment, the PET image data is obtained in a manner such as Figure 2 As shown, including:

[0062] S21, determining each sampling moment, and acquiring the PET image corresponding to each sampling moment;

[0063] The PET image is the original image data collected during the PET scanning process.

[0064] Specifically, according to the preset time interval, the PET sampling time series is constructed as follows:

[0065] T=t0,t1,t2,...t n ,t0=0,t n =T

[0066] Where T is the sampling time series, t0 is the initial sampling time; t i is the i-th sampling moment (i=1,2,...,n).

[0067] After the microneedle drug administration, the PET scan is started, and when the sampling time is reached, the PET image at the corresponding time is collected to obtain the PET image corresponding to each time, which is:

[0068]

[0069] Where x, y, z are the coordinates of the PET image in three-dimensional space; H, W, D are the spatial resolutions of the PET image in the z-axis, x-axis, and y-axis directions, respectively; t i is any moment in the sampling time series T.

[0070] S22, performing image preprocessing on each of the PET image data to obtain preprocessed PET image data;

[0071] In this embodiment, the image preprocessing includes at least coordinate registration; that is, since there may be some differences in the acquisition positions, angles and other acquisition states corresponding to different moments, the acquired PET images may have certain imaging deviations. In order to eliminate the imaging deviations during the image acquisition process, coordinate registration is performed on the PET images acquired at each moment.

[0072] Specifically, based on the coordinates of the PET image at the initial moment, coordinate registration is performed on the PET images collected at other moments except the initial moment to ensure that the PET images at different moments are in the same coordinate system, that is:

[0073]

[0074] in, is the PET image before registration, is the registered PET image; is the PET image at the initial moment, and R is the coordinate registration function, which is used to align the PET images at each moment (except the initial moment) to the coordinate system corresponding to the initial moment.

[0075] It should be noted that, in other embodiments, the image preprocessing may also include other existing preprocessing methods such as image enhancement and image filtering; illustratively, the image preprocessing process also includes:

[0076] The PET images acquired at each time are subjected to standardization processing to eliminate the acquisition system deviation corresponding to different PET images acquired at different acquisition times.

[0077] S23, extracting the boron concentration information in the PET images to obtain PET image data corresponding to each of the PET images.

[0078] Specifically, for a single PET image, extract the boron concentration information corresponding to each pixel in the PET image to obtain the spatial distribution information of the boron concentration in the PET image; use the obtained spatial distribution information of the boron concentration as the PET image data corresponding to the PET image;

[0079] In order to quickly, accurately and conveniently obtain the microneedle boron concentration data corresponding to the sampling time of the PET image data, in a specific embodiment, the microneedle boron concentration data is obtained in a manner such as Figure 3 As shown, including:

[0080] S201, determining sampling points at different sampling distances in the medication area;

[0081] The sampling point is a collection point for collecting boron concentration information in the tissue; the sampling distance is the distance between the sampling point and the microneedle coverage area.

[0082] In a specific embodiment, microneedle drug delivery is performed using a microneedle array, and the area covered by the microneedle array is used as a drug delivery source area; based on the drug delivery source area, different sampling areas for microneedle drug delivery are constructed in sequence according to a preset sampling interval a; that is, an area outside the drug delivery source area and at a distance a from the drug delivery source area is used as the first sampling area; an area outside the drug delivery source area and at a distance 2a from the drug delivery source area is used as the second sampling area, and so on, to obtain the sampling areas of the current microneedle drug delivery; in each of the sampling areas, a number of sampling points are determined by random selection as the sampling points in the corresponding sampling area.

[0083] S202, obtaining the boron concentration of each sampling point at different sampling times, so as to construct a sampling data pair after microneedle drug administration based on the sampling distance, sampling time and corresponding boron concentration;

[0084] Wherein, a single sampling data pair includes the sampling interval, sampling time and corresponding boron concentration value corresponding to the sampling point;

[0085] Specifically, according to a preset time interval, several sampling moments after the end of microneedle drug administration are determined; through tissue sampling and boron concentration measurement, the boron concentration value corresponding to each sampling point at different sampling moments is obtained to construct each sampling data pair.

[0086] It should be noted that the tissue sampling can be achieved by using existing biological tissue sampling methods, and the boron concentration determination can be achieved by using existing in vitro biological determination methods, such as mass spectrometry or chromatography, which can accurately perform quantitative analysis on the boron element content, i.e., the boron concentration.

[0087] S203, fitting a pre-constructed boron drug diffusion function based on each of the sampling data pairs to obtain a fitted microneedle boron concentration diffusion model;

[0088] Wherein, the boron drug diffusion function is used to characterize the spatiotemporal diffusion law of the boron drug in the drug administration area after microneedle administration;

[0089] In this embodiment, in order to facilitate the acquisition of boron concentration data, the diffusion process of the boron drug in the drug administration area is regarded as a uniform diffusion process, and the boron drug diffusion function constructed is:

[0090] B=C(r,t,D)

[0091] Among them, C is the boron drug diffusion function, B is the measured value of boron concentration, r is the sampling distance, t is the sampling time, and D is a preset diffusion coefficient used to reflect the diffusion ability of the drug.

[0092] Each of the sampling data pairs is input into the boron drug diffusion function respectively, so as to fit the boron drug diffusion function by using each of the sampling data pairs to obtain the fitted boron drug diffusion function.

[0093] Exemplarily, the boron drug diffusion function is a polynomial linear function.

[0094] It should be noted that, for each sampling point in the microneedle coverage area, in order to more quickly and conveniently obtain the boron concentration of the sampling point at the initial moment (microneedle administration moment), the boron concentration corresponding to the sampling point is obtained based on the volume and boron concentration of the microneedle administration liquid and the tissue mass of the sampling point:

[0095]

[0096] Among them, V is the volume of the drug solution for microneedle administration; C is the boron concentration for microneedle administration; and M is the tissue mass at the sampling point.

[0097] S204, based on the sampling time of the PET image data, using the fitted microneedle boron concentration diffusion model, obtain the microneedle boron concentration data corresponding to the sampling time.

[0098] Specifically, the sampling time corresponding to the PET image data is obtained, and the sampling time is input into the fitted microneedle boron concentration diffusion model to obtain the changing function between the boron concentration and the sampling distance, that is, the functional relationship between the boron concentration and the sampling distance; based on the functional relationship, the spatial distribution of the microneedle boron concentration in the drug administration area is obtained, and the spatial distribution is used as the microneedle boron concentration data corresponding to the sampling time.

[0099] S30, constructing a sample data set based on the PET image data and the corresponding microneedle boron concentration data;

[0100] Wherein, the sample data set includes a plurality of sample data pairs;

[0101] A single sample data pair includes the PET image data and the microneedle boron concentration data at a sampling time corresponding to the PET image data.

[0102] Acquire the sampling time of the PET image data; and combine the microneedle boron concentration data having the same sampling time and the PET image data into a sample data pair according to the sampling time;

[0103] Wherein, the microneedle boron concentration data is used as input data in the sample pair, and the PET image data is used as label data in the sample pair;

[0104] This step is performed on each PET image data to obtain each sample data pair;

[0105] Each sample data pair is arranged in the time sequence of the corresponding sampling moment to obtain a sample data set.

[0106] S40, training a pre-constructed boron concentration distribution model based on the sample data set to obtain a trained boron concentration distribution model.

[0107] Wherein, the boron concentration distribution model includes a feature extraction module, a time series prediction module, an image reconstruction module and an image comparison module;

[0108] The feature extraction module is used to extract the image features of the boron concentration data to obtain high-dimensional feature information;

[0109] The time series prediction module uses a long short-term memory network (LSTM) to obtain the dependency of boron concentration distribution in the time dimension, and learns the feature transfer and change rules between different time points through the feature sequence of continuous time points;

[0110] The image reconstruction module is a module for converting the boron concentration distribution characteristics into a boron concentration distribution simulation image;

[0111] The image comparison module is a module for comparing the image difference between the boron concentration distribution simulation image and the PET image.

[0112] Specifically, an iterative training process is performed on the boron concentration distribution model using each sample data pair in the sample data set to obtain a trained boron concentration distribution model.

[0113] To facilitate the description of the process, a single sample data pair will be used as an example to describe the single iterative training process.

[0114] like Figure 4 As shown, for a single sample data pair (the i-th sample data pair) as an example, a single iterative training process, when executed, includes:

[0115] Using a feature extraction module, the image features of the microneedle boron concentration data in the current sample data pair are obtained; and the image feature prediction result corresponding to the previous sample data is obtained; wherein the image feature prediction result is a feature processing result obtained after the previous sample data pair is processed using an LSTM network;

[0116] The image feature and the image feature prediction result corresponding to the previous sample data pair are input into the LSTM network to obtain the image feature prediction result corresponding to the current sample data pair;

[0117] Using an image reconstruction module, performing image reconstruction on the current image feature prediction result to obtain a boron concentration distribution simulation image corresponding to the current sample data pair;

[0118] And, based on the image comparison module, image comparison is performed on the current boron concentration distribution simulation image and the PET image in the current sample data pair to obtain the image difference between the two, so as to optimize the parameters of the model based on the image difference.

[0119] Based on the above steps, model training is performed on the boron concentration distribution model using each sample data pair to obtain a trained boron concentration distribution model.

[0120] It should be noted that, when executing the first model iteration training process, the image feature prediction result corresponding to the previous sample data pair is the image feature prediction result obtained based on the first sample data pair;

[0121] Specifically, the method for obtaining the image feature prediction result corresponding to the previous sample data includes:

[0122] The feature extraction module is used to extract the image features corresponding to the PET image data in the first sample data pair and the image features corresponding to the microneedle boron concentration data in the first sample data pair; that is, the PET image data corresponding to the initial time t0 is input into the feature extraction module to obtain the image features corresponding to the PET image data; and the microneedle boron concentration data corresponding to the initial time t0 is input into the feature extraction module to obtain the image features corresponding to the microneedle boron concentration data;

[0123] Feature fusion is performed on the image features corresponding to the PET image data and the image features of the microneedle boron concentration data to obtain the fused features corresponding to the initial time t0; the fused features are input into the LSTM network for processing to obtain the image feature prediction results corresponding to the first sample data pair.

[0124] When executing iterative training of the model corresponding to the second sample data pair, the image feature prediction result corresponding to the first sample data pair is combined with the image feature of the microneedle boron concentration data in the second sample data pair, and is input into the LSTM network for processing to obtain the image feature prediction result corresponding to the second sample data pair; and the image reconstruction module is used to reconstruct the image feature prediction result to obtain a boron concentration distribution simulation image corresponding to the second sample data pair; and the image comparison module is used to compare the boron concentration distribution simulation image corresponding to the second sample data pair with the PET concentration data in the second sample data pair to obtain a corresponding comparison result, and the current boron concentration distribution model is optimized based on the comparison result to obtain a new boron concentration distribution model, and subsequent sample data pairs are used to perform iterative training on the new boron concentration distribution model to obtain a trained model.

[0125] In a more specific embodiment, the LSTM network is:

[0126]

[0127] Among them, F time is the output of the LSTM network, i.e., the image feature prediction result; is the image feature prediction result corresponding to the first sample data pair, that is, the image feature prediction result corresponding to the initial time t0; is the image feature prediction result corresponding to the second sample data pair, that is, the image feature prediction result corresponding to the first sampling time t1; is the image feature prediction result corresponding to the last sample data pair, that is, the last sampling time t n The corresponding image feature prediction results.

[0128] Based on the learned time series features, the time series features are mapped to the image feature prediction results corresponding to each sampling moment through a fully connected layer or a convolutional network to obtain the image feature prediction results corresponding to each sampling moment, which are:

[0129]

[0130] in, is the i-th sampling time t i The image feature prediction result; f is the time series mapping function (usually a multi-layer perceptron or convolutional network); is the i-th sampling time t i The corresponding network parameters.

[0131] In a specific embodiment, the image reconstruction module uses a convolutional neural network including a deconvolution layer, which is:

[0132]

[0133] in, is the i-th sampling time t i The image feature prediction results; Simulated image for the reconstructed boron concentration distribution.

[0134] To ensure that the network can generate accurate and stable results during training, that is, to generate a high-precision boron concentration distribution simulation image, based on mean square error (MSE) loss, attention constraint loss, feature alignment constraint loss, microneedle concentration constraint loss and biological sample constraint loss, in a specific embodiment, a total loss function is constructed as follows:

[0135] L=L PET +λ1L DET +λ2L DIFF +λ3L ATT +λ4L ALI

[0136] Among them, L PET is the image resolution loss, which is used to measure the global error of super-resolution;

[0137] L DET Measure data loss under microneedles to ensure accuracy of local concentration;

[0138] LDIFF To diffuse data loss and ensure the accuracy of concentration in both time and space dimensions;

[0139] L ATT For attention constraint loss, optimize the network's attention to key areas;

[0140] L ALI It is the feature alignment loss, which reduces the feature differences between modalities;

[0141] λ1,λ2,λ3,λ4 are the weights of each constraint.

[0142] In a specific embodiment, the feature extraction module includes a multi-scale convolution module, and the method of using the feature extraction module to extract the image features of the microneedle boron concentration data includes:

[0143] Using the multi-scale convolution module, feature extraction is performed on each of the microneedle drug administration data to obtain the diffusion characteristics of the boron drug in the local tissue, reflecting the multi-scale spatial characteristics of the boron concentration in the drug administration area, which is:

[0144] F Micro =MSCN(Micro ti )

[0145] Among them, F Micro is the image feature of microneedle boron concentration data, MSCN() is a multi-scale convolution module; Micro ti This is the microneedle boron concentration data.

[0146] Furthermore, in order to further improve the attention of the timing prediction module to key areas in the image, the feature extraction module also includes a global attention mechanism (GA) module and a channel attention mechanism (CA) module to dynamically identify key feature areas in the image and assign them higher weights, thereby improving the accuracy of boron concentration reconstruction.

[0147] Among them, the global attention (GA) is a module used to capture global key features and optimize the reconstruction of the overall boron concentration distribution; the channel attention (CA) is a module used to weight feature channels, highlight effective features, and reduce invalid information.

[0148] The two attention mechanisms are weighted and fused as follows:

[0149] F att =GA·F aligned +CA·F aligned .

[0150] Furthermore, the feature extraction module further includes a residual module, and the implementation method of extracting the image features corresponding to the PET image data in the first sample data pair by using the feature extraction module includes:

[0151] Using the Residual module, feature extraction is performed on the PET image data to obtain the global spatial features of the PET image data, which are:

[0152] F PET =ResNet(PET t0 )

[0153] Among them, F PET is the image feature corresponding to the PET image data, ResNet() is the residual module; PET t0 The PET image data is centered for the first sample data.

[0154] In a specific embodiment, the implementation method of performing feature fusion on the image features corresponding to the PET image data and the image features of the microneedle boron concentration data is:

[0155] F Fusion =Conv F PET ,F Micro

[0156] Among them, F Fusion is the fusion feature, and Conv is the feature fusion module.

[0157] Furthermore, due to the difference in distribution of different modal data in the feature space, feature mismatch may occur, thereby affecting the prediction accuracy of the model. In order to improve the matching degree between different modal features, the implementation process of performing feature fusion on the image features corresponding to the PET image data and the image features of the microneedle boron concentration data also includes:

[0158] Adaptive feature alignment is performed on the image features of each modality, that is, the feature contribution of different modalities is adjusted by dynamic weighting to learn the mapping relationship between features, so that the features of different modalities can be aligned in the same feature space, thereby reducing the feature deviation between modalities and achieving the effect of effective fusion of each modality information in the feature space.

[0159] More specifically, the adaptive feature alignment is:

[0160] F aligned =AFAM F PET ,F micro

[0161] Among them, F Micro is the image feature of microneedle boron concentration data; FPET is the image feature corresponding to the PET image data; AFAM is the adaptive feature alignment module.

[0162] Based on the same inventive concept, the present application also provides a method for obtaining boron concentration distribution based on melanoma, which is used to obtain a simulated image of boron concentration distribution at a certain target time after microneedle administration;

[0163] Wherein, the target time is a preset time, which can be any time after microneedle administration;

[0164] The boron concentration distribution simulation image is a simulation data, that is, it is used to simulate the distribution information of the boron concentration in space at a certain moment after microneedle drug administration.

[0165] In this embodiment, the method for obtaining the boron concentration distribution based on melanoma is as follows: Figure 5 As shown, including:

[0166] S1, determining a dosing concentration of microneedle drug delivery; based on the dosing concentration, determining a boron concentration distribution model corresponding to the dosing concentration;

[0167] Among them, the boron concentration distribution model is a model used to simulate the spatial distribution of boron concentration in the melanoma area at subsequent moments based on the PET image data at an initial moment; the initial moment is a moment during or after microneedle administration; the subsequent moment is a moment after the initial moment.

[0168] In this embodiment, the boron concentration distribution model is a pre-trained model; specifically, the model training method of the boron concentration distribution model is to use the boron concentration distribution model acquisition method provided in the above embodiment to perform model training, which will not be repeated here.

[0169] S2, obtaining PET image data of the target area at an initial time; based on the PET image data, using the boron concentration distribution model, obtaining the corresponding boron concentration distribution simulation image at the target time.

[0170] Specifically, a certain moment when the microneedle drug is administered or after the microneedle drug administration is taken as the initial moment; the PET image data corresponding to the initial moment is sampled; the PET image data and the target moment are input into the boron concentration distribution model to output a boron concentration distribution simulation image corresponding to the target moment.

[0171] Based on the same technical concept, the method for acquiring a boron concentration distribution model based on melanoma or the method for acquiring a boron concentration distribution based on melanoma provided in an embodiment of the present invention can be implemented on the terminal side or the server side.

[0172] See also Figure 6 , is an optional hardware structure diagram of a terminal provided in an embodiment of the present invention, and the terminal may be a mobile phone, a computer device, a tablet device, a personal digital processing device, a factory background processing device, etc. The terminal includes: at least one processor 61, a memory 62, at least one network interface 64 and a user interface 63. The various components in the device are coupled together through a bus system 65. It can be understood that the bus system 65 is used to realize the connection and communication between these components. In addition to the data bus, the bus system also includes a power bus, a control bus and a status signal bus.

[0173] The user interface 63 may include a display, a keyboard, a mouse, a trackball, a click gun, keys, buttons, a touch pad or a touch screen.

[0174] It is understood that the memory 62 can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), synchronous static random access memory (SSRAM). The memory characterized by the embodiments of the present invention is intended to include but is not limited to these and any other suitable categories of memory.

[0175] The memory 62 in the embodiment of the present invention is used to store various categories of data to support the operation of the terminal. Examples of these data include: any executable program for operating on the terminal 60, such as an operating system 621 and an application 622; the operating system 621 includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application 622 may include various applications, such as a media player (MediaPlayer), a browser (Browser), etc., for implementing various application services. The method for obtaining a boron concentration distribution model based on melanoma or the method for obtaining a boron concentration distribution based on melanoma provided in the embodiment of the present invention may be included in the application 622.

[0176] The method disclosed in the above embodiment of the present invention can be applied to the processor 61, or implemented by the processor 61. The processor 61 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 61 or the instruction in the form of software. The above processor may be a general processor, a digital signal processor (DSP, Digital Signal Processor), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor 61 can implement or execute the methods, steps and logic block diagrams disclosed in the embodiment of the present invention. The general processor 61 may be a microprocessor or any conventional processor, etc. In combination with the steps of the accessory optimization method provided in the embodiment of the present invention, it can be directly embodied as a hardware decoding processor to execute, or it can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium, which is located in a memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0177] In an exemplary embodiment, the terminal 60 may be implemented by one or more application specific integrated circuits (ASIC), DSP, programmable logic device (PLD), complex programmable logic device (CPLD) to execute the aforementioned method.

[0178] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when called by a processor, implements the method for acquiring a boron concentration distribution model based on melanoma or the method for acquiring a boron concentration distribution based on melanoma provided by the present invention.

[0179] Among them, the computer-readable storage medium can be a tangible device that can hold and store instructions used by the instruction execution device. The computer-readable storage medium can be, for example, (but not limited to) an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, and a mechanical encoding device.

[0180] The computer-readable program characterized herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. A network adapter card or network interface in each computing / processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0181] In summary, the boron concentration distribution model acquisition method based on melanoma, the boron concentration distribution acquisition method, the terminal and the medium provided in the present application, by jointly constructing the different modal information of low-resolution PET images and microneedle drug administration diffusion data into sample data pairs, so that the boron concentration distribution characteristics contained in different data sources are fully expressed and utilized, and by iteratively training the boron concentration distribution model using each sample data pair to obtain the trained boron concentration distribution model, so that the trained boron concentration distribution model can accurately reflect the spatiotemporal law of boron drug diffusion in the area where the melanoma is located; compared with the prior art, the method described in the present application makes full use of the law of microneedle boron concentration changes, and combines the spatial distribution characteristics of boron drug concentration in PET image data, so that not only can the simulation results of the boron concentration distribution corresponding to each moment be obtained quickly and conveniently, but also It also improves the defect of low resolution of PET data, effectively improves the data resolution of the boron concentration distribution simulation results, and improves the accuracy of the boron concentration distribution data; in addition, the present application performs feature alignment on the data sources of each modality, thereby effectively avoiding the problem of information loss and incompleteness under a single modality, effectively reducing the feature mismatch problem between different modal data, and improving the overall feature expression ability, making the reconstructed boron concentration distribution more accurate; and, the introduction of the attention mechanism enables the network to dynamically focus on the key areas of the boron concentration distribution, reducing the waste of resources in non-important areas, thereby improving the accuracy of the reconstruction results in key areas, and improving the efficiency and real-time performance of model reconstruction; and, the design of the joint supervision loss function enables the network to have stability and efficiency during the training process, and can achieve balanced optimization goals under different constraints, effectively preventing overfitting problems. The network architecture of the present invention has good generalization ability under different scenarios and data sources, can adapt to a variety of boron concentration distribution reconstruction tasks, and provides reliable technical support for practical medical applications.

[0182] The present invention realizes high-precision and high-stability reconstruction of boron concentration distribution during BNCT treatment of malignant melanoma through the design of multimodal information fusion, adaptive feature alignment, attention mechanism and joint supervision loss function, which significantly improves the predictability and controllability of subsequent BNCT treatment effects, and has important theoretical value and application prospects.

[0183] The above embodiments are merely illustrative of the principles and effects of the present application and are not intended to limit the present application. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed in the present application shall still be covered by the claims of the present application.

Claims

1. A method for obtaining a boron concentration distribution model based on melanoma, comprising: Obtain the target area of ​​melanoma, microneedle drug delivery area, and drug delivery concentration; The target area is the PET acquisition area of ​​boron concentration information, which covers and is larger than the range of the melanoma; the drug administration area is the area where the boron drug diffuses after microneedle administration; Constructing a sample data set based on the PET image data and the microneedle boron concentration data at the corresponding sampling time; Based on the sample data set, performing iterative training on a preset boron concentration distribution model to obtain a trained boron concentration distribution model; The boron concentration distribution model is a model that simulates the spatial distribution of boron concentration in the target area at subsequent moments based on the PET image data at an initial moment.

2. The method for obtaining a boron concentration distribution model according to claim 1, characterized in that: The iterative training, when executed once, includes: Extracting image features of the microneedle boron concentration data in the current sample data pair; Based on the image feature and the image feature prediction result corresponding to the previous sample data pair, the image feature prediction result corresponding to the current sample data pair is obtained through a time series prediction method; Performing image reconstruction on the current image feature prediction result to obtain a boron concentration distribution simulation image corresponding to the current sample data pair; An image comparison is performed on the current boron concentration distribution simulation image and the PET image data in the current sample data pair to obtain an image difference between the two, so as to optimize the model parameters based on the image difference.

3. The method for obtaining a boron concentration distribution model according to claim 2, characterized in that: When the iterative training is performed for the first time, the method for obtaining the image feature prediction result corresponding to the previous sample data includes: Respectively extracting image features corresponding to the PET image data in the first sample data pair, and extracting image features corresponding to the microneedle boron concentration data in the first sample data pair; Perform feature fusion on the image features of both; The time series prediction method is used to obtain the image feature prediction result corresponding to the fusion feature.

4. The method for obtaining a boron concentration distribution model according to claim 2, characterized in that: The method for acquiring the image features includes: Using a multi-scale convolution module, feature extraction is performed on each of the microneedle drug delivery data to obtain image features corresponding to each of the microneedle drug delivery data, which are: F Micro =MSCN(Micro ti ) Among them, F Micro is the image feature of microneedle boron concentration data, MSCN() is a multi-scale convolution module; Micro ti is the microneedle boron concentration data; and, The residual module is used to extract the spatial structural features in the PET image data in the first sample data pair to obtain the global spatial features of the PET image data, which are: F PET =ResNet(PET t0 ) Among them, F PET is the image feature corresponding to the PET image data, ResNet() is the residual module; PET t0 The PET image data is centered for the first sample data.

5. The method for obtaining a boron concentration distribution model according to claim 4, characterized in that: The method for acquiring the image features also includes: Adaptive feature alignment is performed on the image features of each modality, as follows: F aligned =AFAM F PET ,F micro Among them, F Micro is the image feature of the microneedle boron concentration data; F PET is the image feature corresponding to the PET image data; AFAM is an adaptive feature alignment module.

6. The method for obtaining a boron concentration distribution model according to claim 1, characterized in that: The method for obtaining the microneedle boron concentration data includes: In the drug administration area, each sampling point at different sampling distances is determined, and the boron concentration of each sampling point at different sampling times is obtained, so as to construct a sampling data pair after microneedle drug administration based on the sampling distance, sampling time and corresponding boron concentration; Based on each of the sampling data pairs, a pre-constructed boron drug diffusion function is fitted to obtain a fitted microneedle boron concentration diffusion model; Based on the sampling time of the PET image data, the microneedle boron concentration data corresponding to the sampling time is obtained using the fitted microneedle boron concentration diffusion model.

7. The method for obtaining a boron concentration distribution model according to claim 6, characterized in that: For each sampling point within the microneedle coverage area, at the initial moment, the method for obtaining the microneedle boron concentration data includes: Based on the volume of the drug solution and the boron concentration of the microneedle drug delivery, combined with the tissue mass of the sampling point, the boron concentration corresponding to the sampling point is obtained as follows: Among them, V is the volume of the drug solution for microneedle administration; C is the boron concentration for microneedle administration; and M is the tissue mass at the sampling point.

8. A method for obtaining boron concentration distribution based on melanoma, comprising: Determine a dosing concentration of microneedle drug delivery, and based on the dosing concentration, determine a trained boron concentration distribution model corresponding to the dosing concentration; Acquire PET image data of the target area at an initial time; Based on the PET image data, using the boron concentration distribution model, a corresponding boron concentration distribution simulation image at a target time is obtained; Wherein, the boron concentration distribution model is obtained by using the method for obtaining a melanoma-based boron concentration distribution model according to any one of claims 1 to 7.

9. A terminal, characterized in that: include: A processor and a memory, wherein the memory is communicatively connected to the processor; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes the method for acquiring a boron concentration distribution model based on melanoma as described in any one of claims 1 to 7, or executes the method for acquiring a boron concentration distribution based on melanoma as described in claim 8.

10. A computer storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for obtaining a boron concentration distribution model based on melanoma according to any one of claims 1 to 7 is implemented, or the method for obtaining a boron concentration distribution based on melanoma according to claim 8 is implemented.

Images