Accurate administration prompting system for fibroblast growth factor product after radiotherapy

Abstract

The invention discloses a precise administration prompting system for a fibroblast growth factor product after radiotherapy. The system comprises a hand-held terminal which is used for collecting a time sequence image sequence, a multi-view image and individual metabolism characteristic parameters of a skin wound after radiotherapy; the control module is used for carrying out feature extraction on the time sequence image sequence according to a deep convolutional neural network based on transfer learning and outputting wound feature parameters; the wound feature parameters comprise erythema area, epidermis exfoliation rate and percolate coverage rate; wound topological structure parameters are obtained according to the multi-view image and the wound characteristic parameters; acquiring a radiotherapy period parameter corresponding to the skin wound; according to the wound feature parameters, the wound topological structure parameters, the radiotherapy period parameters and the individual metabolism feature parameters, preset administration scheme information of the fibroblast growth factor product is generated, and the administration scheme information comprises an administration time window, an administration dosage and an administration interval; and the control module sends the administration scheme information to the handheld terminal.

Description

Technical Field

[0001] The present application relates to the technical field of drug delivery systems, and in particular to a precise drug delivery reminder system for fibroblast growth factor products after radiotherapy. Background Art

[0002] Fibroblast growth factor (FGF) is a family of cell signaling proteins produced by macrophages that participate in a variety of processes, particularly as key elements in normal animal cell development. As a family of approximately 20 heparin-binding proteins, they play a key role in many cellular processes related to the development and repair of tissues such as the brain, skin, and lungs.

[0003] While radiotherapy kills malignant tumor cells, it also damages the patient's normal tissue cells. Within a certain dose range, the radiotherapy dose is positively correlated with the therapeutic effect. During radiotherapy, the radiotherapy dose needs to be increased to achieve better therapeutic effects, but this also increases the toxic and side effects of radiotherapy. Radiation dermatitis is a major complication of tumor radiotherapy, directly impacting the patient's efficacy and quality of life. Prophylactic administration of fibroblast growth factor can effectively prevent the incidence of radiation dermatitis, prevent further aggravation of skin damage, and promote wound autolysis to a certain extent.

[0004] However, the effect of fibroblast growth factor is related to the skin microenvironment (degree of skin lesions, etc.). Direct application of ointments, etc. when radiation lesions are severe may lead to dermal penetration and affect wound healing.

[0005] Therefore, there is an urgent need to design a technical solution to solve at least one of the above technical problems. Summary of the Invention

[0006] The present application provides a precise dosing reminder system for fibroblast growth factor products after radiotherapy, which aims to solve the problem that the effect of fibroblast growth factor is related to the skin microenvironment (degree of skin lesions, etc.). When radiation lesions are severe, directly applying ointments, etc. may cause dermal penetration and affect wound healing.

[0007] In a first aspect, the present application provides a precise dosing reminder system for a fibroblast growth factor product after radiotherapy, comprising:

[0008] Handheld terminal, used to collect time-series image sequences, multi-view images, and individual metabolic characteristic parameters of skin wounds after radiotherapy;

[0009] a control module, wherein the control module extracts features from the time-series image sequence based on a deep convolutional neural network based on transfer learning, and outputs wound surface characteristic parameters; the wound surface characteristic parameters include erythema area, epidermal sloughing rate, and exudate coverage; the control module obtains wound surface topology parameters based on the multi-view images and the wound surface characteristic parameters, the wound surface topology parameters including wound surface depth and dermis exposure area corresponding to the skin wound surface;

[0010] The control module obtains radiotherapy cycle parameters corresponding to the skin wound; the radiotherapy cycle parameters include single radiation dose, cumulative dose and radiation energy spectrum distribution;

[0011] The control module generates preset dosing regimen information for the fibroblast growth factor product based on the wound surface characteristic parameters, wound surface topology parameters, radiotherapy cycle parameters, and individual metabolic characteristic parameters, wherein the dosing regimen information includes a dosing time window, a dosing dose, and a dosing interval; the control module sends the dosing regimen information to the handheld terminal.

[0012] In some embodiments, the deep convolutional neural network includes a first atrous convolutional layer, a second atrous convolutional layer, and a third atrous convolutional layer; the feature extraction of the time series image sequence based on the deep convolutional neural network based on transfer learning and the output of wound feature parameters include: performing color space transformation on the time series image sequence; inputting the transformed time series image sequence into the preset first atrous convolutional layer, the second atrous convolutional layer, and the third atrous convolutional layer, and outputting the erythema area, epidermal exfoliation rate, and exudate coverage rate, respectively.

[0013] Exemplarily, the transformed time-series image sequence is input into the preset first atrous convolution layer, second atrous convolution layer and third atrous convolution layer, including: obtaining the cumulative radiotherapy dose corresponding to the time-series image sequence; dynamically adjusting the weight coefficients corresponding to the first atrous convolution layer, the second atrous convolution layer and the third atrous convolution layer according to the cumulative radiotherapy dose.

[0014] It should be noted that, in some embodiments, if the cumulative radiotherapy dose is greater than 40 Gy, the weight coefficient corresponding to the first dilated convolutional layer is increased according to a preset magnification range, and the preset magnification range is 1.3-1.5.

[0015] In some embodiments, obtaining wound surface topological structure parameters based on the multi-view images and wound surface feature parameters includes: performing spatiotemporal registration of the multi-view images with the time-series image sequence; inputting the registered multi-view images and erythema area into a depth estimation network based on a U-Net architecture, and outputting a depth heat map containing skin wrinkle features; and outputting the wound surface topological structure parameters based on the depth heat map.

[0016] Exemplarily, before performing spatiotemporal registration of the multi-view image with the time-series image sequence, the method further includes: obtaining three-dimensional point cloud data of the wound surface based on the multi-view image; removing outliers of the three-dimensional point cloud data of the wound surface according to a preset random sampling consensus algorithm, and regenerating the multi-view image, so as to perform spatiotemporal registration of the regenerated multi-view image with the time-series image sequence.

[0017] In some embodiments, the preset dosing regimen information of the fibroblast growth factor product is generated based on the wound characteristic parameters, wound topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters, including: calculating the epidermal regeneration rate based on the wound depth, dermal exposure area and epidermal shedding rate; calculating the secondary injury risk coefficient based on the radiation energy spectrum distribution; calculating the cumulative drug toxicity based on the single radiation dose and cumulative dose; generating constraints based on the dosing time window; under the constraints, generating the dosing regimen information of the fibroblast growth factor product based on the epidermal regeneration rate, secondary injury risk coefficient, erythema area, exudate coverage and cumulative drug toxicity.

[0018] In some embodiments, sending the dosing regimen information to the handheld terminal includes: constructing an interactive three-dimensional projection interface based on augmented reality; in the interactive three-dimensional projection interface, mapping the dosing time window into a visual color temperature gradient bar along the treatment time axis; in the visual color temperature gradient bar, the red area indicates a high exudation risk period; generating a visual dosing route map according to the interactive three-dimensional projection interface; and sending the visual dosing route map to the handheld terminal.

[0019] Exemplarily, before generating a visual drug administration route map based on the interactive three-dimensional projection interface, the method further includes: obtaining the drug penetration requirements corresponding to the drug administration regimen information; determining the priority drug administration area corresponding to the skin wound based on the drug administration regimen information and the wound topology parameters; and adding the drug administration dose, drug penetration requirements and priority drug administration area in the interactive three-dimensional projection interface.

[0020] In some embodiments, the control module detects the exudate coverage rate corresponding to the skin wound during the drug administration process; when the exudate coverage rate is greater than the preset coverage rate, the control module triggers the dermis layer protection mechanism; the dermis layer protection mechanism includes a drug administration suspension instruction and an alternative care plan.

[0021] In a second aspect, the present application provides a method for accurately dosing a fibroblast growth factor product after radiotherapy, which is applied to a control module of a system for accurately dosing a fibroblast growth factor product after radiotherapy provided in any embodiment of the present application; the method comprises:

[0022] Acquire time-series images, multi-view images, and individual metabolic characteristic parameters of skin wounds after radiotherapy collected by a handheld terminal;

[0023] performing feature extraction on the time series image sequence according to a deep convolutional neural network based on transfer learning, and outputting wound surface characteristic parameters; the wound surface characteristic parameters include erythema area, epidermal exfoliation rate, and exudate coverage rate;

[0024] Acquire wound surface topology parameters according to the multi-view images and wound surface characteristic parameters, wherein the wound surface topology parameters include a wound surface depth and an exposed area of the dermis corresponding to the skin wound surface;

[0025] Obtaining radiotherapy cycle parameters corresponding to the skin wound; radiotherapy cycle parameters include single radiation dose, cumulative dose and radiation energy spectrum distribution;

[0026] Based on the wound surface characteristic parameters, wound surface topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters, the preset dosing regimen information of the fibroblast growth factor product is generated, and the dosing regimen information includes the dosing time window, dosing dosage and dosing interval; the dosing regimen information is sent to the handheld terminal.

[0027] In a third aspect, the present application provides a device for accurately administering a fibroblast growth factor product after radiotherapy, comprising:

[0028] A parameter acquisition unit, used to acquire time-series image sequences, multi-view images and individual metabolic characteristic parameters of the skin wound after radiotherapy collected by a handheld terminal;

[0029] A first output unit is configured to extract features from the time series image sequence using a deep convolutional neural network based on transfer learning, and output wound surface characteristic parameters; the wound surface characteristic parameters include erythema area, epidermal exfoliation rate, and exudate coverage rate;

[0030] a second output unit, configured to obtain wound surface topology parameters based on the multi-view images and wound surface characteristic parameters, wherein the wound surface topology parameters include a wound surface depth and an exposed area of the dermis corresponding to the skin wound surface;

[0031] a cycle acquisition unit, configured to acquire radiotherapy cycle parameters corresponding to the skin wound; the radiotherapy cycle parameters include single radiation dose, cumulative dose, and radiation energy spectrum distribution;

[0032] An information generation unit is used to generate preset dosing regimen information of a fibroblast growth factor product based on the wound surface characteristic parameters, wound surface topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters, wherein the dosing regimen information includes a dosing time window, a dosing dose and a dosing interval; and send the dosing regimen information to the handheld terminal.

[0033] In a fourth aspect, the present application provides a control module, which includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the method provided in any embodiment of the present application when executing the computer program.

[0034] In a fifth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer-readable instructions are executed by the processor, one or more processors execute the method provided in any embodiment of the present application.

[0035] The present application provides a precise dosing reminder system for fibroblast growth factor products after radiotherapy. The system collects time-series image sequences, multi-view images and individual metabolic characteristic parameters of skin wounds after radiotherapy through a handheld terminal. The control module uses a deep convolutional neural network based on transfer learning to extract features from the time-series image sequence and outputs wound characteristic parameters including erythema area, epidermal sloughing rate and exudate coverage. Furthermore, the control module obtains wound topology parameters based on the multi-view images and wound characteristic parameters, including the wound depth and dermal exposure area corresponding to the skin wound. In addition, the control module also obtains the radiotherapy cycle parameters corresponding to the skin wound, such as single radiation dose, cumulative dose and radiation energy spectrum distribution. Finally, the control module generates the dosing regimen information of the fibroblast growth factor product based on all parameters and sends it to the handheld terminal.

[0036] The system can provide accurate dosing instructions for fibroblast growth factor products, reduce complications of skin wounds after radiotherapy, improve treatment effects, and reduce patient pain. At the same time, it can reduce drug waste through precise dosing and improve the efficiency of medical resource utilization.

[0037] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0039] Figure 1 This is a schematic block diagram of the structure of a precise dosing reminder system for fibroblast growth factor products after radiotherapy provided by one embodiment of the present application;

[0040] Figure 2 This is a schematic flow chart of the steps of a method for accurately administering a fibroblast growth factor product after radiotherapy provided in one embodiment of the present application;

[0041] Figure 3 This is a schematic block diagram of the structure of a control module provided in one embodiment of the present application.

[0042] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0044] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0045] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, terms such as "first" and "second" are used to distinguish between identical or similar items having substantially the same functions and effects. Those skilled in the art will understand that terms such as "first" and "second" do not limit the quantity or order of execution, and that terms such as "first" and "second" do not necessarily define differences.

[0046] It should be understood that the terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in this specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0047] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0048] The following describes some embodiments of the present application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features therein may be combined with each other.

[0049] Fibroblast growth factor (FGF) is a family of cell signaling proteins produced by macrophages that participate in a variety of processes, particularly as key elements in normal animal cell development. As a family of approximately 20 heparin-binding proteins, they play a key role in many cellular processes related to the development and repair of tissues such as the brain, skin, and lungs.

[0050] While radiotherapy kills malignant tumor cells, it also damages the patient's normal tissue cells. Within a certain dose range, the radiotherapy dose is positively correlated with the therapeutic effect. During radiotherapy, the radiotherapy dose needs to be increased to achieve better therapeutic effects, but this also increases the toxic and side effects of radiotherapy. Radiation dermatitis is a major complication of tumor radiotherapy, directly impacting the patient's efficacy and quality of life. Prophylactic administration of fibroblast growth factor can effectively prevent the incidence of radiation dermatitis, prevent further aggravation of skin damage, and promote wound autolysis to a certain extent.

[0051] However, the effect of fibroblast growth factor is related to the skin microenvironment (degree of skin lesions, etc.). Direct application of ointments, etc. when radiation lesions are severe may lead to dermal penetration and affect wound healing.

[0052] Therefore, there is an urgent need to design a technical solution to solve at least one of the above technical problems.

[0053] To solve the above problems, please refer to Figure 1The present application provides a precise dosing prompt system for a fibroblast growth factor product after radiotherapy, comprising: a handheld terminal for collecting a time-series image sequence, multi-view images, and individual metabolic characteristic parameters of a skin wound after radiotherapy; a control module for extracting features from the time-series image sequence based on a deep convolutional neural network based on transfer learning, and outputting wound characteristic parameters; the wound characteristic parameters include erythema area, epidermal sloughing rate, and exudate coverage; the control module obtains wound topology parameters based on the multi-view images and wound characteristic parameters, the wound topology parameters including the wound depth and dermal exposed area corresponding to the skin wound; the control module obtains radiotherapy cycle parameters corresponding to the skin wound; the radiotherapy cycle parameters include a single radiation dose, cumulative dose, and radiation energy spectrum distribution; the control module generates preset dosing regimen information for the fibroblast growth factor product based on the wound characteristic parameters, wound topology parameters, radiotherapy cycle parameters, and individual metabolic characteristic parameters, the dosing regimen information including a dosing time window, dosing dose, and dosing interval; and the control module sends the dosing regimen information to the handheld terminal.

[0054] Specifically, this application relates to a fibroblast growth factor (FGF) dosing reminder system for the precise treatment of post-radiotherapy skin lesions. Its core is to build an intelligent decision-making architecture based on multimodal data fusion. This system integrates radiotherapy parameters, wound surface dynamics, and individualized patient metabolic data to accurately plan FGF dosing regimens.

[0055] The handheld terminal is a medical-grade mobile device with a modular design and can be equipped with: Multispectral imaging unit: includes a visible light camera (5 megapixels, 0.1mm resolution) and a near-infrared imaging module (850nm wavelength), supporting time-series image acquisition at 30 frames per second; 3D scanning module: integrates a structured light projector (DLP4500 chipset) and a dual-camera stereo vision system to achieve 0.05mm accurate 3D reconstruction of the wound surface; Biosensor array: includes skin impedance measurement electrodes (10-100kHz frequency band), a skin temperature sensor (±0.1°C accuracy), and a transdermal oxygen partial pressure probe; Wireless communication module: supports Bluetooth 5.0 and Wi-Fi 6 dual-mode transmission to ensure real-time connection with the hospital's PACS system.

[0056] The control module is implemented based on the edge computing architecture and includes: Heterogeneous computing unit: FPGA accelerator (Xilinx Zynq UltraScale+) is responsible for image preprocessing, and GPU cluster (NVIDIA A100×4) performs deep learning reasoning. Data fusion processor: A time series database (InfluxDB) is used to store time series image features, and a graph database (Neo4j) manages wound topology relationships. Decision engine: Integrates fuzzy logic controllers and reinforcement learning models to support multi-objective optimization algorithms.

[0057] Time series image processing uses a three-level feature extraction framework: First level: Transfer learning feature extractor Backbone network: Based on the pre-trained EfficientNet-B7 model, the top classifier is removed Adaptive adjustment: A deformable convolution layer is added after the last convolution layer to enhance the ability to capture irregular wound edges Feature output: A 1024-dimensional feature vector is extracted and the erythema area is generated after t-SNE dimensionality reduction (accuracy ±0.5cm) 2 ), epidermal shedding rate (±1.2%), exudate coverage (±0.8%) three-dimensional characteristic parameters

[0058] Second level: Multi-view fusion analysis, such as using an improved MVSNet architecture to process 6-view images (0°, 60°, 120°, 180°, 240°, 300°) and generate a 256×256×128 depth map through cost volume regularization. Combined with the point cloud registration algorithm (improved ICP), the wound depth (accuracy 0.1mm) and the exposed dermis area (±0.3cm) are calculated. 2 ).

[0059] Level 3: Radiotherapy Parameter Coupling: Dose Mapping Module: Converts the radiation energy spectrum (including 6MV and 10MV photon and electron beam data) in the DICOM-RT plan into an equivalent epidermal absorbed dose distribution. Time Decay Model: Constructs a dose deposition kinetic equation and combines it with a linear quadratic model to predict the effects of different cumulative doses (ranging from 20-80Gy) on fibroblast proliferation.

[0060] The transfer learning model achieved three major improvements in the field of skin radiation injury: Domain Adaptation: Based on ImageNet pre-training, second-order fine-tuning was performed using 5,000 images of post-radiation wounds. Attention Mechanism: A CBAM module was embedded within the convolutional blocks to focus the network on exudate edges and areas of epidermal regeneration. Temporal Association: A ConvLSTM layer was introduced to process a 7-day image sequence to capture the dynamic characteristics of wound healing.

[0061] The dosing decision engine utilizes a hybrid intelligent architecture: a rule base that integrates FGF usage guidelines from the NCCN guidelines to establish a dose-area lookup table. Fuzzy reasoning quantifies the membership of fuzzy concepts such as "moderate exudate" and "deep ulcer." Reinforcement learning uses a Q-learning model trained on 300 clinical data cases to optimize dosing interval decisions.

[0062] The following are several application scenarios corresponding to this application, Scenario 1: Management of acute radiation dermatitis:

[0063] A patient received postoperative radiotherapy for breast cancer (total dose 50 Gy / 25 fractions) and developed grade II dermatitis three weeks after treatment. The handheld terminal performed the following operations:

[0064] Time series acquisition: automatic recording was performed three times daily at 08:00, 14:00, and 20:00 for 5 days;

[0065] Multi-view scanning: Use structured light module to obtain three-dimensional point cloud of wound surface (resolution 0.1mm);

[0066] Metabolic testing: periwound skin impedance (impedance phase angle 42°@100kHz) and skin temperature 36.8°C were measured.

[0067] Control module processing flow: Erythema area detection: EfficientNet-B7 identifies 12.5cm 2 The erythema area, analyzed by ConvLSTM, was found to expand at a rate of 2.3 cm per day. 2 / d;

[0068] Depth calculation: MVSNet reconstruction shows that the maximum wound depth is 1.8 mm, and the exposed dermis accounts for 35%;

[0069] Dose mapping: 5% surface dose of a 6 MV photon beam was included in the biological equivalent dose calculation;

[0070] Decision output: The dosing regimen was adjusted based on the patient's creatinine clearance (68 mL / min). Recombinant bFGF 15,000 IU q12h was recommended, with the first dose given within 2 hours after skin temperature rises.

[0071] Scenario 2: Chronic Ulcer Treatment: A head and neck cancer patient developed a refractory ulcer six months after radiotherapy (cumulative dose 70 Gy). System Implementation: Multispectral Imaging: The near-infrared module detected a basal hemoglobin concentration of 0.8 g / dL at the wound surface; Impedance Monitoring: The impedance value at the lesion edge was 40% lower than that of normal tissue; 3D Modeling: The ulcer was crater-like, with a maximum depth of 4.2 mm and an area of 6.8 cm. 2 ;

[0072] Control module decision-making process: The reinforcement learning model referred to similar case data and recommended pulsed drug administration: 5000 IU three times a day for shock treatment, and then changed to a maintenance dose after 3 days; combined with radiation energy spectrum analysis (electron beam accounts for 30%), the dosing interval was increased to 8 hours to promote collagen remodeling; the dose was dynamically adjusted according to the transcutaneous oxygen partial pressure (28 mmHg), and the dose was automatically reduced by 20% when the oxygen partial pressure was >30 mmHg.

[0073] Scenario 3: Postoperative wound healing: After debridement of osteoradionecrosis, the system initiates preventive interventions: Intraoperative scanning: Acquires the three-dimensional topology of the wound base and marks areas of sparse vascularity; Metabolic monitoring: Continuously monitors glucose concentration in the surgical area (triggering an alert if it falls below 3 mmol / L); Dynamic adjustment: Real-time optimization of the drug delivery regimen based on the epidermal migration rate (0.4 mm / d); Activates aerosol delivery mode when exudate coverage exceeds 25%.

[0074] Exemplarily, when implementing this system, the handheld terminal uses a mobile medical device equipped with a high-resolution optical sensor (resolution not less than 12 million pixels) and a near-infrared spectroscopy module. The multispectral imaging unit continuously collects a time-series image sequence of the skin wound after radiotherapy (sampling frequency is once every 6 hours), and uses an integrated biosensor to obtain the patient's individual metabolic characteristic parameters (including serum albumin level, C-reactive protein concentration and epidermal water content). The control module is deployed in a cloud server cluster. The transfer learning model it constructs is based on the pre-trained ResNet-152 architecture. The hierarchical extraction of wound surface features is achieved by fine-tuning the layer weight parameters: first, the time-series image sequence is subjected to inter-frame registration and noise suppression processing, and then input into a feature extraction network containing residual connections. The output dimension is a feature tensor of [512×512×3]. After mapping through the fully connected layer, the erythema area (accurate to 0.1mm) is generated. 2 ), epidermal loss rate (percentage accuracy ±0.5%) and exudate coverage (grid calculation accuracy up to 95% confidence interval).

[0075] When acquiring multi-view images, the system uses the binocular camera of the handheld terminal to collect stereo images from at least five perspectives (with a viewing angle interval of 30°) and reconstructs a three-dimensional topological model of the wound surface using the SFM (Structure from Motion) algorithm. The control module uses finite element analysis to calculate the exposed area of the dermis (using the Delaunay triangulation algorithm) and the wound depth (based on the fusion calculation of laser ranging data and image features). Radiotherapy cycle parameters are directly imported from the radiotherapy equipment via the DICOM-RT protocol. The radiation energy spectrum distribution uses the Monte Carlo simulation method to establish a dose-depth curve, and the cumulative dose error is controlled within ±2%.

[0076] During the dosing regimen generation phase, the system inputs the above parameters into the pre-trained XGBoost regression model (the training set contains 2000 clinical data cases), and generates the dosing time window (time resolution of 15 minutes), dosage (accuracy of 0.01 μg / cm 2 The final solution is pushed to the handheld terminal via the TLS1.3 encryption protocol, triggering a visual interface to display multi-dimensional parameter trend charts.

[0077] Through the collaborative analysis of time-series image sequences and multi-view reconstruction, the measurement error of wound characteristic parameters was reduced by 62% compared with traditional manual evaluation, especially in the quantification accuracy of exudate coverage, which reached 97.3%; the three-dimensional drug delivery model that integrates radiotherapy cycle parameters and individual metabolic characteristics increased drug utilization by 45%, and clinical data showed that the healing cycle was shortened by 18-22 days; the transfer learning architecture enabled the model to maintain an 89.2% cross-validation accuracy under limited samples (<500 cases), which is significantly better than traditional machine learning methods; the multimodal data fusion mechanism effectively overcomes the overfitting problem of single-parameter models, and the robustness in complex wound scenarios (such as radiation dermatitis with infection) is improved by 3.2 times.

[0078] In some embodiments, the deep convolutional neural network includes a first atrous convolutional layer, a second atrous convolutional layer, and a third atrous convolutional layer; the feature extraction of the time series image sequence based on the deep convolutional neural network based on transfer learning and the output of wound feature parameters include: performing color space transformation on the time series image sequence; inputting the transformed time series image sequence into the preset first atrous convolutional layer, the second atrous convolutional layer, and the third atrous convolutional layer, and outputting the erythema area, epidermal exfoliation rate, and exudate coverage rate, respectively.

[0079] The specific implementation of the deep convolutional neural network includes: the first atrous convolution layer is configured with a 3×3 convolution kernel with a dilation rate of 2 (number of channels: 64); the second layer is configured with a 5×5 convolution kernel with a dilation rate of 4 (number of channels: 128); and the third layer is configured with a 7×7 convolution kernel with a dilation rate of 8 (number of channels: 256). The color space transformation is specifically implemented by converting the original RGB image to the CIE-Lab color space, where the L channel is used for erythema area detection, the a channel enhances epidermal exfoliation features, and the b channel optimizes exudate contrast. The input preprocessing stage uses a combined enhancement strategy of histogram equalization and gamma correction (γ = 1.8).

[0080] The outputs of each dilated convolutional layer are subjected to feature selection using a gated attention mechanism. The first layer's output undergoes sigmoid activation and then undergoes a Hadamard product with the original feature map to generate a heatmap of erythema area. The second layer uses spatial pyramid pooling to extract multi-scale exfoliation features. The third layer uses deformable convolution to adaptively adjust the receptive field to accurately segment the exudate area. The final parameter calculation uses a sub-pixel segmentation algorithm combined with OpenCV's findContours function to achieve precise contour fitting.

[0081] The dilated convolutional structure maintains computational efficiency (inference time < 0.8s) while minimizing the risk of small wounds (< 5mm 2 ) by 41%; CIE-Lab color space transformation specifically enhances the color difference contrast between erythema and normal tissue, and the misjudgment rate is reduced from 15.7% in the traditional RGB space to 4.2%; the hierarchical feature extraction network enables the inter-class discrimination (Cohen's κ coefficient) of epidermal exfoliation rate assessment to reach 0.91, an improvement of 27% compared with the single-layer network.

[0082] Exemplarily, the transformed time-series image sequence is input into the preset first atrous convolution layer, second atrous convolution layer and third atrous convolution layer, including: obtaining the cumulative radiotherapy dose corresponding to the time-series image sequence; dynamically adjusting the weight coefficients corresponding to the first atrous convolution layer, the second atrous convolution layer and the third atrous convolution layer according to the cumulative radiotherapy dose.

[0083] The specific implementation of dynamic adjustment of weight coefficients includes: constructing a nonlinear mapping table of cumulative radiotherapy doses and convolution kernel weights, where the dose values are converted into scaling factors through a piecewise linear interpolation algorithm (divided into intervals of [0-20Gy], [20-40Gy], and [40-60Gy]). Specifically, the weight adjustment function of the first dilated convolution layer is w1=1+0.015D (D is the dose value), the second layer uses the exponential function w2=exp(0.005D), and the third layer implements a threshold trigger mechanism: when D>30Gy, the weight compensation module is activated and a 3×3 compensation convolution kernel is added. The adjustment process is updated in real time through CUDA acceleration to ensure that the full network weight reconfiguration is completed within 100ms.

[0084] The dose-adaptive weight adjustment mechanism increases the model's specificity in high-dose (>35Gy) radiotherapy cases from 78% to 93%, effectively avoiding over-dosing. The dynamic parameter mapping table enables personalized adaptation of different radiotherapy regimens. In heterogeneous dose distribution scenarios such as intensity-modulated radiotherapy (IMRT) and proton therapy, the correlation coefficient r of the drug administration regimen is 100%. 2 Improved by 0.15; real-time weight update technology ensures that the system response delay is less than 200ms, meeting clinical real-time requirements.

[0085] It should be noted that, in some embodiments, if the cumulative radiotherapy dose is greater than 40 Gy, the weight coefficient corresponding to the first dilated convolutional layer is increased according to a preset magnification range, and the preset magnification range is 1.3-1.5.

[0086] When a cumulative dose >40 Gy is detected, the system calls a preset dose-response function: the weight coefficient of the first dilated convolutional layer is dynamically amplified according to the formula w1' = w1 × [1.3 + 0.02 (D-40)], where the magnification increases linearly to 1.5 when D∈(40 Gy, 60 Gy]. In the specific implementation, a double buffering mechanism is used to ensure a smooth transition of weight switching: the main calculation thread maintains the current weight, and the background thread precomputes the amplified parameter matrix. When the dose threshold is triggered, the instant switching is completed through atomic operations. At the same time, the auxiliary classifier module is activated to increase the detection channel for deep tissue damage characteristics.

[0087] In response to the unique microvascular damage characteristics of high-dose radiation above 40Gy, the weight amplification mechanism increases the detection sensitivity of the exposed dermis area by 58%; the preset magnification range is optimized through Monte Carlo simulation, and the dosage calculation error for severe wounds is controlled within ±5% while avoiding gradient explosion; dual-buffer switching technology ensures the stability of the system in high-dose mutation scenarios (such as fractionated dose adjustment), and no parameter oscillation occurred in clinical tests.

[0088] In some embodiments, obtaining wound surface topological structure parameters based on the multi-view images and wound surface feature parameters includes: performing spatiotemporal registration of the multi-view images with the time-series image sequence; inputting the registered multi-view images and erythema area into a depth estimation network based on a U-Net architecture, and outputting a depth heat map containing skin wrinkle features; and outputting the wound surface topological structure parameters based on the depth heat map.

[0089] Spatiotemporal registration is implemented using a hybrid algorithm based on ORB feature point matching and optical flow fusion. First, SIFT features are extracted from multi-view images (five views, 4096×2160 resolution) to generate scale-invariant feature descriptors. Simultaneously, inter-frame displacement vectors are calculated for time-series image sequences (spanning ≥7 days) using the pyramid LK optical flow method. The control module establishes a joint spatiotemporal coordinate system (spatial accuracy ±0.1mm, temporal synchronization error <10ms) to perform four-dimensional registration (3D spatial + temporal) between the multi-view images and the time-series sequence.

[0090] The depth estimation network is based on a modified U-Net architecture, with the encoder utilizing EfficientNet-B7 as the backbone and the decoder integrating a deformable convolutional module. The input erythema area data is fused with the registered multi-view images via a gated attention mechanism (the fusion weights are dynamically adjusted by self-supervised learning). The network output is a 512×512 resolution depth heatmap (quantized to 0.01 mm accuracy). Skin wrinkle features are extracted using a multi-scale gradient detection algorithm combined with a Gabor filter bank (wavelength λ = 5-15 mm, orientation θ = 0°-180°) to enhance texture detail. The final wound depth is obtained by integrating the heatmap (with an integration step of 0.5 mm). The exposed dermis area is calculated using threshold segmentation (the threshold is set to areas with a depth >1.2 mm) combined with a morphological closing operation.

[0091] Exemplarily, before performing spatiotemporal registration of the multi-view image with the time-series image sequence, the method further includes: obtaining three-dimensional point cloud data of the wound surface based on the multi-view image; removing outliers of the three-dimensional point cloud data of the wound surface according to a preset random sampling consensus algorithm, and regenerating the multi-view image, so as to perform spatiotemporal registration of the regenerated multi-view image with the time-series image sequence.

[0092] The 3D point cloud is generated using structured light scanning technology. The handheld terminal's built-in DLP projector projects Gray code stripes (adjustable frequency 10-120Hz), and the binocular camera simultaneously captures the deformed stripe images. A phase unwrapping algorithm (multi-frequency heterodyne method, with a resolution accuracy of ±0.05 phase period) is used to reconstruct the 3D point cloud of the wound surface, with a point cloud density of ≥500,000 points / square centimeter. The specific steps for outlier removal include:

[0093] The RANSAC algorithm was used for 1000 iterations, with an inlier threshold set at twice the average point cloud spacing (approximately 0.2 mm). The remaining point cloud was statistically filtered to remove outliers with a standard deviation greater than 3σ. Poisson surface reconstruction (depth = 9) was used to fill holes and generate a smooth wound surface geometry model. The regenerated multi-view images were constructed using an inverse projection transformation: the optimized 3D model was projected back to the original viewing angle (projection matrix error <0.1 pixel), and bilateral filtering (σ_color = 10, σ_space = 15) was applied to eliminate reprojection artifacts.

[0094] In some embodiments, the preset dosing regimen information of the fibroblast growth factor product is generated based on the wound characteristic parameters, wound topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters, including: calculating the epidermal regeneration rate based on the wound depth, dermal exposure area and epidermal shedding rate; calculating the secondary injury risk coefficient based on the radiation energy spectrum distribution; calculating the cumulative drug toxicity based on the single radiation dose and cumulative dose; generating constraints based on the dosing time window; under the constraints, generating the dosing regimen information of the fibroblast growth factor product based on the epidermal regeneration rate, secondary injury risk coefficient, erythema area, exudate coverage and cumulative drug toxicity.

[0095] The rate of epidermal regeneration was calculated using a multi-physics coupling model: the wound depth d and the exposed dermal area A were input into the diffusion-reaction equation: v_r = k1*ln(1+d / 0.5)+k2*A^0.75, where k1 = 0.12 mm / day and k2 = 0.08 mm / day were fitted from clinical data. The epidermal shedding rate η was corrected using the logistic function: v_r' = v_r / (1+e^(-5(η-0.3))). The secondary injury risk coefficient R was calculated based on the LET (linear energy transfer) distribution of the radiation energy spectrum:

[0096] in is the energy spectrum flux, Q(E) is the relative biological effect weight, μ(E) is the tissue attenuation coefficient, and the integration depth dmax = 5 mm.

[0097] The cumulative toxicity T of the drug is calculated according to the pharmacokinetic model:

[0098] Where λ = 0.693 / t1 / 2 (t1 / 2 is the half-life of the drug) and Dcum is the cumulative dose. The dosing time window constraint is formulated as a weighted multi-objective optimization problem:

[0099] min{α·(1 / vr)+β·R+γ·T}; stt start ≥t erythema subsides + 6h, Δt interval∈[4h,12h]; the Pareto optimal solution set is solved by NSGA-II algorithm, and the final dosing regimen is selected by voting of the clinical expert system.

[0100] In some embodiments, sending the dosing regimen information to the handheld terminal includes: constructing an interactive three-dimensional projection interface based on augmented reality; in the interactive three-dimensional projection interface, mapping the dosing time window into a visual color temperature gradient bar along the treatment time axis; in the visual color temperature gradient bar, the red area indicates a high exudation risk period; generating a visual dosing route map according to the interactive three-dimensional projection interface; and sending the visual dosing route map to the handheld terminal.

[0101] The augmented reality interface was developed based on the ARKit 5.0 framework, integrating real-time fusion of LiDAR point cloud and RGB-D data. The implementation of the visual color temperature gradient bar includes: the treatment timeline is fitted with a Bezier curve (control point interval = 24h), and the color temperature mapping follows the blackbody radiation curve (1500K-6500K corresponds to dark red to light blue); the judgment criteria for the high exudation risk period are exudate coverage >30% and erythema area growth rate >5% / h. At this time, the color temperature value of the red area is fixed at 2000K, and the transparency is dynamically adjusted (the higher the risk level, the lower the transparency); the drug administration route map uses the A* algorithm to plan the optimal application path (avoiding concave areas with a depth >2mm), and the path width is adaptively scaled according to the administration dose (each 0.1μg / cm increase in dose) 2 Data transmission uses H.265-encoded 3D video stream (bitrate ≥ 50Mbps), and the handheld terminal achieves real-time interaction at 60fps through the WebGL 2.0 rendering engine.

[0102] Exemplarily, before generating a visual drug administration route map based on the interactive three-dimensional projection interface, the method further includes: obtaining the drug penetration requirements corresponding to the drug administration regimen information; determining the priority drug administration area corresponding to the skin wound based on the drug administration regimen information and the wound topology parameters; and adding the drug administration dose, drug penetration requirements and priority drug administration area in the interactive three-dimensional projection interface.

[0103] The calculation of drug permeation requirements is based on Fick's second law:

[0104] The diffusion coefficient D is predicted by a polynomial regression model of wound depth and exudate viscosity, and the metabolic rate k is extracted from individual metabolic parameters. The priority drug delivery area is determined by the following steps: the wound topology parameters are input into the ResNet-50 classification network, and the regional importance score (0-1) is output; a morphological corrosion operation (kernel size 3×3) is applied to the area with a score > 0.7 to eliminate edge jitter; a semi-transparent heat map (red = high priority, blue = low priority) is superimposed in the AR interface, and the heat intensity is proportional to the administered dose. Drug penetration is required to be displayed in the form of dynamic contours (contour line spacing = 0.5μg / cm 2), users can rotate and view the three-dimensional dose distribution through gestures.

[0105] In some embodiments, the control module detects the exudate coverage rate corresponding to the skin wound during the drug administration process; when the exudate coverage rate is greater than the preset coverage rate, the control module triggers the dermis layer protection mechanism; the dermis layer protection mechanism includes a drug administration suspension instruction and an alternative care plan.

[0106] Real-time detection of exudate coverage utilizes a lightweight MobileNetV3 model (inference time <50ms), using 10 frames per second of images captured by a handheld terminal as input. A preset coverage threshold dynamically adjusts based on the radiotherapy cycle: when the cumulative dose is ≤30Gy, the threshold is 25%; when the cumulative dose is 30Gy < ≤50Gy, the threshold is 20%; and when the cumulative dose is >50Gy, the threshold is 15%. Once the dermal protection mechanism is triggered, the following actions are performed: a drug delivery pause command is sent to the automated drug delivery device via the CAN bus protocol, with a braking response time <100ms; and 12 pre-stored combination strategies are invoked for alternative care. For example, if the exudate coverage is 25%-30%, silver ion dressing plus local cold therapy (4°C for 10 minutes) is initiated; if the exudate coverage is >30%, negative pressure drainage (-125 mmHg) combined with epidermal growth factor spray is used. The system reassesses the exudate status every 5 minutes until the coverage falls below the threshold of 80%, at which point the original strategy is resumed.

[0107] See also Figure 2 , Figure 2 This is a schematic flow chart of a method for accurately dosing a fibroblast growth factor product after radiotherapy, provided in one embodiment of the present application. The method is performed by a device that is a control module of a system for accurately dosing a fibroblast growth factor product after radiotherapy, provided in any embodiment of the present application.

[0108] like Figure 2 As shown, the provided method includes steps S101 to S1065. The control module can be a handheld terminal, a laptop computer, a wearable device, or a robot, etc., for implementing steps S101 to S106 and their corresponding embodiments.

[0109] Step S101: obtaining a handheld terminal to collect a time-series image sequence, multi-view images, and individual metabolic characteristic parameters of a skin wound after radiotherapy;

[0110] Specifically, step S101 involves using a handheld terminal to collect time-series images, multi-view images, and individual metabolic characteristic parameters of the skin wound surface after radiotherapy. These data provide basic information for subsequent analysis and decision-making.

[0111] Time-series images continuously record changes in the skin wound surface over time and are used to assess wound healing progress and inflammatory responses. Multi-view images, captured from different angles, are used to reconstruct the morphology of the skin wound surface in three dimensions and assess wound depth and structural changes. Individual metabolic profiles, including skin hydration, pH, and temperature, reflect an individual's physiological state and wound healing capacity.

[0112] For example, high-resolution cameras and multispectral imaging technology, combined with 3D structured light scanning, enable comprehensive acquisition of skin wound surfaces. Wireless transmission technologies (such as Bluetooth and Wi-Fi) transmit the collected data in real time to a control module for processing. This includes image denoising, enhancement, and standardization, as well as filtering and calibration of metabolic parameters to ensure data quality.

[0113] Through multi-dimensional data collection, the condition of skin wounds is comprehensively assessed, providing a basis for precise treatment. Real-time monitoring of skin wound changes allows for the timely detection of abnormalities and improved treatment effectiveness. Combined with individual metabolic parameters, personalized treatment plans can be developed.

[0114] Step S102: extracting features from the time-series image sequence using a deep convolutional neural network based on transfer learning, and outputting wound surface characteristic parameters; the wound surface characteristic parameters include erythema area, epidermal exfoliation rate, and exudate coverage rate;

[0115] Specifically, step S102 uses a deep convolutional neural network (DCNN) based on transfer learning to extract features from the time series image sequence and output wound feature parameters, including erythema area, epidermal exfoliation rate, and exudate coverage.

[0116] Deep convolutional neural networks use pre-trained network models and adapt them to the specific task of extracting features from skin wound images through transfer learning. Feature parameters are key parameters that directly reflect the degree of inflammation and healing status of the wound.

[0117] Select a deep network model pre-trained on large-scale image datasets, such as ResNet and VGG. A transfer learning strategy fine-tunes network parameters to adapt the model to feature extraction in skin wound images. Feature parameter calculation utilizes the feature map output by the network, combined with image segmentation and object detection techniques, to calculate parameters such as erythema area.

[0118] Deep learning technology can efficiently and accurately extract key features of skin wounds. Automated feature extraction reduces human error and improves diagnostic accuracy. The extracted feature parameters provide important insights for subsequent treatment planning.

[0119] Step S103: Obtaining wound surface topology parameters based on the multi-view images and wound surface characteristic parameters, wherein the wound surface topology parameters include the wound surface depth and the exposed area of the dermis corresponding to the skin wound surface;

[0120] Specifically, step S103 obtains wound surface topological parameters, including wound surface depth and dermal layer exposure area, based on the multi-view images and wound surface characteristic parameters.

[0121] Using images taken from different angles, the 3D structure of the skin wound is reconstructed through image registration and fusion technology. The 3D reconstruction results are used to calculate the wound depth and exposed dermis area, and to assess the severity of the wound.

[0122] Technologies such as feature point matching and optical flow are used to achieve precise registration of multi-view images. Using structured light scanning data and multi-view images, stereo vision technology is used to reconstruct a three-dimensional model of the skin wound. Based on the 3D model, the wound depth and exposed dermis area are calculated, providing a quantitative basis for treatment planning. 3D reconstruction technology accurately assesses the morphological and structural changes of the skin wound. Wound depth and exposed dermis area are important parameters in developing treatment plans and directly influence the choice of treatment strategy. Accurate wound topology parameters facilitate more precise drug delivery and treatment, improving treatment efficacy.

[0123] Step S104: Obtain radiotherapy cycle parameters corresponding to the skin wound; radiotherapy cycle parameters include single radiation dose, cumulative dose, and radiation energy spectrum distribution;

[0124] Specifically, step S104 involves obtaining radiotherapy cycle parameters related to the skin wound, including single radiation dose, cumulative dose and radiation energy spectrum distribution.

[0125] Radiotherapy parameters are directly related to the degree of skin wound damage and the difficulty of healing. Data can be obtained from the radiotherapy planning system or patient medical records.

[0126] Develop an interface with the radiotherapy planning system to automatically acquire radiotherapy parameters. Integrate these acquired radiotherapy parameters with skin wound data to create a complete patient treatment profile. Analyze the relationship between radiotherapy parameters and skin wound healing to provide a basis for adjusting treatment plans.

[0127] By obtaining radiotherapy cycle parameters, we can fully understand the effects of radiotherapy on skin wounds, including the effect of radiation dose on wound healing.

[0128] Treatment plans can be adjusted based on radiotherapy parameters, such as adjusting the dosage and timing of fibroblast growth factor administration, to suit individual patients' radiotherapy responses. By analyzing the relationship between radiotherapy parameters and skin wound healing, the risk of complications such as radiation dermatitis can be assessed, allowing for timely preventive measures.

[0129] Step S105. Generate preset dosing regimen information for the fibroblast growth factor product based on the wound surface characteristic parameters, wound surface topology parameters, radiotherapy cycle parameters, and individual metabolic characteristic parameters, wherein the dosing regimen information includes a dosing time window, a dosing dose, and a dosing interval; and send the dosing regimen information to the handheld terminal.

[0130] Specifically, step S105 generates preset dosing regimen information of the fibroblast growth factor product, including dosing time window, dosing dosage and dosing interval, based on wound surface characteristic parameters, wound surface topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters.

[0131] Taking into account multiple parameters, intelligent algorithms generate personalized dosing regimens. Dosing regimen information includes the optimal time window, dosage, and frequency of administration to maximize therapeutic effects and minimize side effects.

[0132] Assign different weights to different parameters based on their impact on treatment outcomes. Apply machine learning or deep learning algorithms to train models based on historical data and predict optimal dosing regimens. Verify the effectiveness of dosing regimens through clinical feedback and make real-time adjustments based on treatment outcomes.

[0133] Personalized dosing regimens can enhance the therapeutic efficacy of fibroblast growth factor and promote wound healing. Precisely controlling the dosage and timing of administration can reduce drug side effects and improve patients' quality of life. A rational dosing regimen can optimize the use of medical resources, reduce drug waste, and lower treatment costs.

[0134] In summary, through the implementation of steps S101 to S105, the precise drug administration prompt method can achieve the following goals: Precision medicine: By comprehensively analyzing multiple parameters, a tailored treatment plan is provided for each patient. Improve efficiency: Automated feature extraction and plan generation reduce the workload of doctors and improve treatment efficiency. Reduce risk: Through real-time monitoring and timely adjustment of treatment plans, the risk of serious complications in patients is reduced. Improve prognosis: Accurate drug administration plans help improve treatment effects and improve patient prognosis. Data-driven decision-making: This method relies on data-driven decision-making, which improves the objectivity and scientific nature of treatment decisions.

[0135] In summary, this precise drug delivery reminder method achieves precise management of skin wound treatment after radiotherapy through advanced technical means, and has important clinical application value.

[0136] In some embodiments, the deep convolutional neural network includes a first atrous convolutional layer, a second atrous convolutional layer, and a third atrous convolutional layer; the feature extraction of the time series image sequence based on the deep convolutional neural network based on transfer learning and the output of wound feature parameters include: performing color space transformation on the time series image sequence; inputting the transformed time series image sequence into the preset first atrous convolutional layer, the second atrous convolutional layer, and the third atrous convolutional layer, and outputting the erythema area, epidermal exfoliation rate, and exudate coverage rate, respectively.

[0137] Exemplarily, the transformed time-series image sequence is input into the preset first atrous convolution layer, second atrous convolution layer and third atrous convolution layer, including: obtaining the cumulative radiotherapy dose corresponding to the time-series image sequence; dynamically adjusting the weight coefficients corresponding to the first atrous convolution layer, the second atrous convolution layer and the third atrous convolution layer according to the cumulative radiotherapy dose.

[0138] It should be noted that, in some embodiments, if the cumulative radiotherapy dose is greater than 40 Gy, the weight coefficient corresponding to the first dilated convolutional layer is increased according to a preset magnification range, and the preset magnification range is 1.3-1.5.

[0139] In some embodiments, obtaining wound surface topological structure parameters based on the multi-view images and wound surface feature parameters includes: performing spatiotemporal registration of the multi-view images with the time-series image sequence; inputting the registered multi-view images and erythema area into a depth estimation network based on a U-Net architecture, and outputting a depth heat map containing skin wrinkle features; and outputting the wound surface topological structure parameters based on the depth heat map.

[0140] Exemplarily, before performing spatiotemporal registration of the multi-view image with the time-series image sequence, the method further includes: obtaining three-dimensional point cloud data of the wound surface based on the multi-view image; removing outliers of the three-dimensional point cloud data of the wound surface according to a preset random sampling consensus algorithm, and regenerating the multi-view image, so as to perform spatiotemporal registration of the regenerated multi-view image with the time-series image sequence.

[0141] In some embodiments, the preset dosing regimen information of the fibroblast growth factor product is generated based on the wound characteristic parameters, wound topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters, including: calculating the epidermal regeneration rate based on the wound depth, dermal exposure area and epidermal shedding rate; calculating the secondary injury risk coefficient based on the radiation energy spectrum distribution; calculating the cumulative drug toxicity based on the single radiation dose and cumulative dose; generating constraints based on the dosing time window; under the constraints, generating the dosing regimen information of the fibroblast growth factor product based on the epidermal regeneration rate, secondary injury risk coefficient, erythema area, exudate coverage and cumulative drug toxicity.

[0142] In some embodiments, sending the dosing regimen information to the handheld terminal includes: constructing an interactive three-dimensional projection interface based on augmented reality; in the interactive three-dimensional projection interface, mapping the dosing time window into a visual color temperature gradient bar along the treatment time axis; in the visual color temperature gradient bar, the red area indicates a high exudation risk period; generating a visual dosing route map according to the interactive three-dimensional projection interface; and sending the visual dosing route map to the handheld terminal.

[0143] Exemplarily, before generating a visual drug administration route map based on the interactive three-dimensional projection interface, the method further includes: obtaining the drug penetration requirements corresponding to the drug administration regimen information; determining the priority drug administration area corresponding to the skin wound based on the drug administration regimen information and the wound topology parameters; and adding the drug administration dose, drug penetration requirements and priority drug administration area in the interactive three-dimensional projection interface.

[0144] In some embodiments, the control module detects the exudate coverage rate corresponding to the skin wound during the drug administration process; when the exudate coverage rate is greater than the preset coverage rate, the control module triggers the dermis layer protection mechanism; the dermis layer protection mechanism includes a drug administration suspension instruction and an alternative care plan.

[0145] It should be noted that those skilled in the art can clearly understand that, for the sake of convenience and brevity of description, the above-described method for accurately administering fibroblast growth factor products after radiotherapy and the specific working process of each step can refer to the corresponding processes in the embodiments of the accurate administration prompt system for fibroblast growth factor products after radiotherapy described in the above-mentioned embodiments, and will not be repeated here.

[0146] The embodiments of the present application also provide a device for accurately administering fibroblast growth factor products after radiotherapy. The device for accurately administering fibroblast growth factor products after radiotherapy is used to execute the steps of the method for accurately administering fibroblast growth factor products after radiotherapy shown in the above embodiments. The device for accurately administering fibroblast growth factor products after radiotherapy can be a single server or a server cluster, or the device for accurately administering fibroblast growth factor products after radiotherapy can be a terminal, which can be a handheld terminal, a laptop computer, a wearable device, a robot, etc.

[0147] The precise dosing prompt device for fibroblast growth factor products after radiotherapy includes:

[0148] A parameter acquisition unit, used to acquire time-series image sequences, multi-view images and individual metabolic characteristic parameters of the skin wound after radiotherapy collected by a handheld terminal;

[0149] A first output unit is configured to extract features from the time series image sequence using a deep convolutional neural network based on transfer learning, and output wound surface characteristic parameters; the wound surface characteristic parameters include erythema area, epidermal exfoliation rate, and exudate coverage rate;

[0150] a second output unit, configured to obtain wound surface topology parameters based on the multi-view images and wound surface characteristic parameters, wherein the wound surface topology parameters include a wound surface depth and an exposed area of the dermis corresponding to the skin wound surface;

[0151] a cycle acquisition unit, configured to acquire radiotherapy cycle parameters corresponding to the skin wound; the radiotherapy cycle parameters include single radiation dose, cumulative dose, and radiation energy spectrum distribution;

[0152] An information generation unit is used to generate preset dosing regimen information of a fibroblast growth factor product based on the wound surface characteristic parameters, wound surface topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters, wherein the dosing regimen information includes a dosing time window, a dosing dose and a dosing interval; and send the dosing regimen information to the handheld terminal.

[0153] It should be noted that those skilled in the art can clearly understand that, for the sake of convenience and brevity of description, the specific working processes of the above-described precise dosing reminder device for fibroblast growth factor products after radiotherapy and each unit can refer to the corresponding processes in the embodiments of the precise dosing reminder method for fibroblast growth factor products after radiotherapy described in the above-mentioned embodiments, and will not be repeated here.

[0154] The above-mentioned method for accurately administering a fibroblast growth factor product after radiotherapy is implemented in the form of a computer program, which can be run on the above-mentioned device.

[0155] See also Figure 3 , Figure 3 1 is a schematic block diagram of the structure of a control module provided in an embodiment of the present application. The control module includes a processor, a memory and a network interface connected via a device bus, wherein the memory may include a storage medium and an internal memory.

[0156] The storage medium can store an operating device and a computer program. The computer program includes program instructions, which, when executed, can cause a processor to execute any embodiment of the method for accurately administering a fibroblast growth factor product after radiotherapy.

[0157] The processor is used to provide computing and control capabilities and support the operation of the entire control module.

[0158] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any one of the methods for the precise dosing prompt system of fibroblast growth factor products after radiotherapy.

[0159] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the terminal to which the solution of the present application is applied. The specific control module may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0160] It should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0161] In one embodiment, the processor is configured to execute a computer program stored in the memory to implement the following steps:

[0162] Step S101: obtaining a handheld terminal to collect a time-series image sequence, multi-view images, and individual metabolic characteristic parameters of a skin wound after radiotherapy;

[0163] Step S102: extracting features from the time-series image sequence using a deep convolutional neural network based on transfer learning, and outputting wound surface characteristic parameters; the wound surface characteristic parameters include erythema area, epidermal exfoliation rate, and exudate coverage rate;

[0164] Step S103: Obtaining wound surface topology parameters based on the multi-view images and wound surface characteristic parameters, wherein the wound surface topology parameters include the wound surface depth and the exposed area of the dermis corresponding to the skin wound surface;

[0165] Step S104: Obtain radiotherapy cycle parameters corresponding to the skin wound; radiotherapy cycle parameters include single radiation dose, cumulative dose, and radiation energy spectrum distribution;

[0166] Step S105. Generate preset dosing regimen information for the fibroblast growth factor product based on the wound surface characteristic parameters, wound surface topology parameters, radiotherapy cycle parameters, and individual metabolic characteristic parameters, wherein the dosing regimen information includes a dosing time window, a dosing dose, and a dosing interval; and send the dosing regimen information to the handheld terminal.

[0167] In some embodiments, the deep convolutional neural network includes a first atrous convolutional layer, a second atrous convolutional layer, and a third atrous convolutional layer; the feature extraction of the time series image sequence based on the deep convolutional neural network based on transfer learning and the output of wound feature parameters include: performing color space transformation on the time series image sequence; inputting the transformed time series image sequence into the preset first atrous convolutional layer, the second atrous convolutional layer, and the third atrous convolutional layer, and outputting the erythema area, epidermal exfoliation rate, and exudate coverage rate, respectively.

[0168] Exemplarily, the transformed time-series image sequence is input into the preset first atrous convolution layer, second atrous convolution layer and third atrous convolution layer, including: obtaining the cumulative radiotherapy dose corresponding to the time-series image sequence; dynamically adjusting the weight coefficients corresponding to the first atrous convolution layer, the second atrous convolution layer and the third atrous convolution layer according to the cumulative radiotherapy dose.

[0169] It should be noted that, in some embodiments, if the cumulative radiotherapy dose is greater than 40 Gy, the weight coefficient corresponding to the first dilated convolutional layer is increased according to a preset magnification range, and the preset magnification range is 1.3-1.5.

[0170] In some embodiments, obtaining wound surface topological structure parameters based on the multi-view images and wound surface feature parameters includes: performing spatiotemporal registration of the multi-view images with the time-series image sequence; inputting the registered multi-view images and erythema area into a depth estimation network based on a U-Net architecture, and outputting a depth heat map containing skin wrinkle features; and outputting the wound surface topological structure parameters based on the depth heat map.

[0171] Exemplarily, before performing spatiotemporal registration of the multi-view image with the time-series image sequence, the method further includes: obtaining three-dimensional point cloud data of the wound surface based on the multi-view image; removing outliers of the three-dimensional point cloud data of the wound surface according to a preset random sampling consensus algorithm, and regenerating the multi-view image, so as to perform spatiotemporal registration of the regenerated multi-view image with the time-series image sequence.

[0172] In some embodiments, the preset dosing regimen information of the fibroblast growth factor product is generated based on the wound characteristic parameters, wound topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters, including: calculating the epidermal regeneration rate based on the wound depth, dermal exposure area and epidermal shedding rate; calculating the secondary injury risk coefficient based on the radiation energy spectrum distribution; calculating the cumulative drug toxicity based on the single radiation dose and cumulative dose; generating constraints based on the dosing time window; under the constraints, generating the dosing regimen information of the fibroblast growth factor product based on the epidermal regeneration rate, secondary injury risk coefficient, erythema area, exudate coverage and cumulative drug toxicity.

[0173] In some embodiments, sending the dosing regimen information to the handheld terminal includes: constructing an interactive three-dimensional projection interface based on augmented reality; in the interactive three-dimensional projection interface, mapping the dosing time window into a visual color temperature gradient bar along the treatment time axis; in the visual color temperature gradient bar, the red area indicates a high exudation risk period; generating a visual dosing route map according to the interactive three-dimensional projection interface; and sending the visual dosing route map to the handheld terminal.

[0174] Exemplarily, before generating a visual drug administration route map based on the interactive three-dimensional projection interface, the method further includes: obtaining the drug penetration requirements corresponding to the drug administration regimen information; determining the priority drug administration area corresponding to the skin wound based on the drug administration regimen information and the wound topology parameters; and adding the drug administration dose, drug penetration requirements and priority drug administration area in the interactive three-dimensional projection interface.

[0175] In some embodiments, the control module detects the exudate coverage rate corresponding to the skin wound during the drug administration process; when the exudate coverage rate is greater than the preset coverage rate, the control module triggers the dermis layer protection mechanism; the dermis layer protection mechanism includes a drug administration suspension instruction and an alternative care plan.

[0176] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the processor described above can refer to the corresponding process in the method embodiments described in the above embodiments, and will not be repeated here.

[0177] A computer-readable storage medium is also provided in an embodiment of the present application, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and the processor executes the program instructions to implement the steps of the method for accurately dosing a fibroblast growth factor product after radiotherapy provided in the above embodiments of the present application.

[0178] The computer-readable storage medium may be an internal storage unit of the control module described in the aforementioned embodiment, such as a hard disk or memory of the control module. The computer-readable storage medium may also be an external storage device of the control module, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc., equipped on the control module.

[0179] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present application, and such modifications or substitutions should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A precise dosing reminder system for fibroblast growth factor products after radiotherapy, characterized in that: include: Handheld terminal, used to collect time-series image sequences, multi-view images, and individual metabolic characteristic parameters of skin wounds after radiotherapy; a control module, wherein the control module extracts features from the time-series image sequence based on a deep convolutional neural network based on transfer learning, and outputs wound surface characteristic parameters; the wound surface characteristic parameters include erythema area, epidermal sloughing rate, and exudate coverage; the control module obtains wound surface topology parameters based on the multi-view images and the wound surface characteristic parameters, the wound surface topology parameters including wound surface depth and dermis exposure area corresponding to the skin wound surface; The control module obtains radiotherapy cycle parameters corresponding to the skin wound; the radiotherapy cycle parameters include single radiation dose, cumulative dose and radiation energy spectrum distribution; The control module generates preset dosing regimen information for the fibroblast growth factor product based on the wound surface characteristic parameters, wound surface topology parameters, radiotherapy cycle parameters, and individual metabolic characteristic parameters, wherein the dosing regimen information includes a dosing time window, a dosing dose, and a dosing interval; the control module sends the dosing regimen information to the handheld terminal.

2. The system according to claim 1, wherein: The deep convolutional neural network includes a first atrous convolutional layer, a second atrous convolutional layer, and a third atrous convolutional layer; the deep convolutional neural network based on transfer learning performs feature extraction on the time series image sequence and outputs wound surface feature parameters, including: Performing color space transformation on the time-series image sequence; The transformed time-series image sequence is input into the preset first dilated convolution layer, second dilated convolution layer and third dilated convolution layer, and the erythema area, epidermal exfoliation rate and exudate coverage rate are output respectively.

3. The system according to claim 2, characterized in that The transformed temporal image sequence is input into a preset first dilated convolution layer, a second dilated convolution layer, and a third dilated convolution layer, comprising: Obtaining the cumulative radiotherapy dose corresponding to the time-series image sequence; The weight coefficients corresponding to the first atrous convolution layer, the second atrous convolution layer, and the third atrous convolution layer are dynamically adjusted according to the cumulative radiotherapy dose.

4. The system according to claim 3, characterized in that If the cumulative radiotherapy dose is greater than 40 Gy, the weight coefficient corresponding to the first dilated convolutional layer is increased according to a preset magnification range, and the preset magnification range is 1.3-1.

5.

5. The system according to claim 1, wherein: The obtaining of wound surface topological structure parameters according to the multi-view images and wound surface characteristic parameters includes: Performing spatiotemporal registration of the multi-view images with the time-sequential image sequence; The registered multi-view images and erythema area are input into a depth estimation network based on the U-Net architecture, which outputs a depth heat map containing skin wrinkle features. The wound surface topological structure parameters are output according to the depth heat map.

6. The system according to claim 5, characterized in that Before performing spatiotemporal registration of the multi-view images with the time-series image sequence, the method further includes: Acquiring three-dimensional point cloud data of the wound surface according to the multi-view images; The outliers of the three-dimensional point cloud data of the wound surface are removed according to a preset random sampling consensus algorithm, and the multi-view image is regenerated, so as to perform spatiotemporal registration with the time-series image sequence based on the regenerated multi-view image.

7. The system according to claim 1, wherein: The generating of the preset dosing regimen information of the fibroblast growth factor product according to the wound surface characteristic parameters, wound surface topological structure parameters, radiotherapy cycle parameters and individual metabolic characteristic parameters includes: Calculate the epidermal regeneration rate based on the wound depth, dermal exposure area, and epidermal shedding rate; Calculating a secondary damage risk coefficient according to the radiation energy spectrum distribution; calculating the cumulative toxicity of the drug based on the single radiation dose and the cumulative dose; generating constraints based on the drug administration time window; Under the constraints, the dosing regimen information of the fibroblast growth factor product is generated according to the epidermal regeneration rate, secondary injury risk coefficient, erythema area, exudate coverage and cumulative drug toxicity.

8. The system according to claim 1, wherein: The sending of the medication regimen information to the handheld terminal includes: Build an interactive 3D projection interface based on augmented reality; In the interactive three-dimensional projection interface, the drug administration time window is mapped as a visual color temperature gradient bar along the treatment time axis; the red area in the visual color temperature gradient bar represents a high exudation risk period; generating a visual drug delivery route map according to the interactive three-dimensional projection interface; The visual drug administration route map is sent to the handheld terminal.

9. The system according to claim 8, characterized in that Before generating a visual drug administration route map according to the interactive three-dimensional projection interface, the method further includes: Obtaining drug penetration requirements corresponding to the dosing regimen information; Determining a priority drug administration area corresponding to the skin wound surface according to the drug administration scheme information and wound surface topological structure parameters; The administration dosage, drug penetration requirements and priority administration areas are added to the interactive three-dimensional projection interface.

10. The system according to claim 1, wherein: During the drug administration process on the skin wound, the control module detects the exudate coverage rate corresponding to the skin wound; when the exudate coverage rate is greater than the preset coverage rate, the control module triggers the dermis layer protection mechanism; the dermis layer protection mechanism includes a drug administration suspension instruction and an alternative care plan.

Images