Skin lesion treatment light source determination method based on Wood's lamp and related device

Abstract

The embodiment of the invention relates to the technical field of medical equipment, and provides a skin lesion treatment light source determination method based on a Wood's lamp and a related device, and the method comprises the following steps: smearing 5-aminolevulinic acid on a target skin area; acquiring Wood lamp irradiation image data of the target skin area; inputting the Wood lamp irradiation image data into a pre-trained lesion degree identification model to obtain a lesion identification result; according to the lesion recognition result, determining lesion skin division area data; light source output doses of different infection zones are determined according to the lesion skin division zone data, the boundary ranges of an uninfected zone, a mild infection zone and a severe infection zone can be distinguished according to wood lamp irradiation image data of a target skin zone, and a light dose output strategy is dynamically matched according to the fluorescence intensity gradient of the zones, so that the target skin area can be identified. And the sterilization efficiency of a severe infection area is improved while excessive irradiation of normal tissues is avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical equipment, in particular to a skin lesion treatment light source determination method based on Wood's light and related devices. BACKGROUND

[0002] As an important tool for skin lesion detection, Wood's light can excite specific components in skin tissue to produce characteristic fluorescence by emitting 320-400 nm long-wave ultraviolet light, and has unique application value in the field of photodynamic therapy. In clinical practice, Wood's light can effectively detect the abnormal accumulation of protoporphyrin IX in the infected area: due to the lack of ferrous chelatase metabolic pathway in pathogenic microorganisms, the concentration of protoporphyrin IX in the infected area is significantly higher than that in normal tissue, and it presents characteristic coral red fluorescence under ultraviolet excitation. This optical property makes it an important basis for guiding photodynamic therapy. The current mainstream treatment plan uses a fixed dose of red light (630-635 nm) for photodynamic therapy after positioning the infected area through Wood's light fluorescence imaging, and uses active oxygen produced by photosensitive reaction to selectively kill pathogens.

[0003] However, the existing red light treatment light source output uses a homogeneous irradiation mode, which fails to achieve differentiated dose regulation for infected subareas, so it may cause damage to normal tissue due to excessive light dose in uninfected areas, and the sterilization effect may be affected due to insufficient dose in severely infected areas, and the accuracy of light source treatment is not high. SUMMARY

[0004] The embodiments of the present application provide a skin lesion treatment light source determination method based on Wood's light and related devices, which can distinguish the boundary range of uninfected areas, mildly infected areas and severely infected areas according to the Wood's light irradiation image data of the target skin area, and dynamically match the light dose output strategy according to the fluorescence intensity gradient of the subareas, so as to avoid excessive irradiation of normal tissue and improve the sterilization efficiency of severely infected areas.

[0005] The first aspect of the embodiments of the present application provides a skin lesion treatment light source determination method based on Wood's light, which comprises: applying 5-aminolevulinic acid to the target skin area; obtaining Wood's light irradiation image data of the target skin area; inputting the Wood's light irradiation image data into a pre-trained lesion degree recognition model to obtain a lesion recognition result; determining lesion skin division area data according to the lesion recognition result; determining the light source output dose of different infected subareas according to the lesion skin division area data.

[0006] In a possible implementation, the determining of the lesion skin division area data according to the lesion recognition result comprises: reconstructing the lesion recognition result in a three-dimensional space to obtain a lesion skin heat map; decomposing the heat map by using an adaptive grid division algorithm to obtain a plurality of independent detection units; calculating the fluorescence intensity of the detection units to obtain a fluorescence intensity standard deviation and a fluorescence intensity mean value; determining a lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean value; collecting the lesion skin division area types to obtain lesion skin division area data.

[0007] In a possible implementation, the determining of the lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean value comprises: when the fluorescence intensity standard deviation σ < 15 and the fluorescence intensity mean value μ < 50 a.u., marking the lesion skin division area type as an uninfected area; when the fluorescence intensity standard deviation σ < 30 and the fluorescence intensity mean value 50 ≤ μ < 150 a.u., marking the lesion skin division area type as a mild infection area; when the fluorescence intensity standard deviation σ < 50 and the fluorescence intensity mean value μ > 150 a.u., marking the lesion skin division area type as a severe infection area.

[0008] In a possible implementation, the training process of the pre-trained lesion degree recognition model comprises: obtaining first historical Wood's light image data containing lesion area images labeled with protoporphyrin IX fluorescence intensity grading; constructing a first lesion degree recognition model of a cloud data center according to the first historical Wood's light image data; downloading the first lesion degree recognition model to each client node, and each client node respectively trains the first lesion degree recognition model according to client local Wood's light image data to obtain first lesion degree recognition model update parameters; uploading the first lesion degree recognition model update parameters to the cloud data center, and the cloud data center globally aggregates the first lesion degree recognition model according to the first intrusion detection model update parameters to obtain a second lesion degree recognition model; Repeat the steps of sending the first lesion degree identification model to each client node, each client node...obtaining the first lesion degree identification model update parameters to uploading the first lesion degree identification model update parameters to the cloud data...obtaining the second lesion degree identification model until it iterates k times to obtain the lesion degree identification model.

[0009] In one possible implementation, constructing a first lesion severity recognition model of the cloud data center based on the first historical Wood's lamp image data includes: performing data cleaning on the first historical Wood's lamp image data to obtain cleaned first historical Wood's lamp image data; Performing format conversion on the cleaned first historical Wood's lamp image data to obtain second historical Wood's lamp image data; performing feature encoding on the second historical Wood's lamp image data to obtain third historical Wood's lamp image data; determining a network traffic feature matrix based on the third historical Wood's lamp image data; According to the network traffic feature matrix, a CNN neural network is used to construct a first pathology degree recognition model for a cloud data center.

[0010] In this example, by obtaining the Wood's lamp irradiation image data of the target skin area, the fluorescence characteristics are analyzed using the pre-trained lesion degree recognition model. After generating the lesion recognition result, the diseased skin division area data is established by combining three-dimensional reconstruction and dynamic partitioning technology, and finally the differentiated light source dose control of the infected partition is achieved. According to the Wood's lamp irradiation image data of the target skin area, the boundary range of the uninfected area, the mildly infected area and the severely infected area can be distinguished, and the light dose output strategy is dynamically matched according to the fluorescence intensity gradient of the partition. While avoiding excessive irradiation of normal tissue, the sterilization efficiency of the severely infected area is improved, so as to solve the technical problems of "insufficient treatment" and "excessive damage" in traditional homogenized treatment, and the low accuracy of light source treatment.

[0011] A second aspect of an embodiment of the present application provides a device for determining a light source for treating skin lesions based on a Wood's lamp, the device comprising: an acquisition unit, configured to acquire Wood's lamp irradiation image data of a target skin area; a first processing unit, configured to input the Wood's lamp irradiation image data into a pre-trained lesion degree recognition model to obtain a lesion recognition result; A second processing unit is used to determine diseased skin area division data according to the diseased skin identification result; The determination unit is used to determine the light source output dose of different infected areas according to the diseased skin area division data.

[0012] In a possible implementation, in the aspect of determining the lesion skin division area data according to the lesion identification result, the second processing unit is configured to: perform three-dimensional space reconstruction on the lesion identification result to obtain a lesion skin heat map; decompose the heat map by using an adaptive grid division algorithm to obtain a plurality of independent detection units; perform fluorescence intensity calculation on the detection units to obtain a fluorescence intensity standard deviation and a fluorescence intensity mean value; determine a lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean value; collect the lesion skin division area types to obtain lesion skin division area data.

[0013] In a possible implementation, in the aspect of determining the lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean value, the second processing unit is configured to: when the fluorescence intensity standard deviation σ < 15 and the fluorescence intensity mean value μ < 50 a.u., mark the lesion skin division area type as an uninfected area; when the fluorescence intensity standard deviation σ < 30 and the fluorescence intensity mean value 50 ≤ μ < 150 a.u., mark the lesion skin division area type as a mild infection area; when the fluorescence intensity standard deviation σ < 50 and the fluorescence intensity mean value μ > 150 a.u., mark the lesion skin division area type as a severe infection area.

[0014] In a possible implementation, in the aspect of inputting the Wood's light image data into the pre-trained lesion degree identification model to obtain a lesion identification result, the training process of the pre-trained lesion degree identification model in the first processing unit includes: obtain first historical Wood's light image data, the first historical Wood's light image data containing lesion area images labeled with protoporphyrin IX fluorescence intensity grading; construct a first lesion degree identification model of a cloud data center according to the first historical Wood's light image data; distribute the first lesion degree identification model to each client node, and each client node respectively trains the first lesion degree identification model according to client local Wood's light image data to obtain first lesion degree identification model update parameters; upload the first lesion degree identification model update parameters to the cloud data center, and the cloud data center performs global model aggregation on the first lesion degree identification model according to the first intrusion detection model update parameters to obtain a second lesion degree identification model; The steps of repeatedly issuing the first lesion degree identification model to each client node, each client node obtaining first lesion degree identification model update parameters, uploading the first lesion degree identification model update parameters to the cloud data center, and obtaining a second lesion degree identification model are repeated k times to obtain a lesion degree identification model.

[0015] In one possible implementation, in the aspect of constructing the first lesion degree identification model of the cloud data center according to the first historical Wood's lamp image data, the first processing unit is configured to: perform data cleaning on the first historical Wood's lamp image data to obtain cleaned first historical Wood's lamp image data; perform format conversion on the cleaned first historical Wood's lamp image data to obtain second historical Wood's lamp image data; perform feature coding on the second historical Wood's lamp image data to obtain third historical Wood's lamp image data; determine a network traffic feature matrix according to the third historical Wood's lamp image data; construct the first lesion degree identification model of the cloud data center by using a CNN neural network according to the network traffic feature matrix.

[0016] A third aspect of the embodiment of the present application provides a terminal, including a processor, an input device, an output device, and a memory, which are connected to each other, wherein the memory is configured to store a computer program, the computer program includes program instructions, and the processor is configured to invoke the program instructions to execute the steps of the Wood's lamp-based skin lesion treatment light source determination method in the first aspect of the embodiment of the present application.

[0017] A fourth aspect of the embodiment of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to execute some or all steps of the Wood's lamp-based skin lesion treatment light source determination method in the first aspect of the embodiment of the present application.

[0018] A fifth aspect of the embodiment of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute some or all steps of the Wood's lamp-based skin lesion treatment light source determination method in the first aspect of the embodiment of the present application. The computer program product can be a software installation package. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0020] Figure 1 A total flowchart of a skin lesion treatment light source determination method based on a Wood lamp is provided for the embodiments of the present application. Figure 2 A structural diagram of a skin lesion treatment light source determination device based on a Wood lamp is provided for the embodiments of the present application. Figure 3 A structural diagram of a terminal is provided for the embodiments of the present application. Reference signs: 0-preparation unit, 1-acquisition unit, 2-first processing unit, 3-second processing unit, 4-determination unit. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0022] The terms "first", "second", etc. in the specification and claims of the present application and the above drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed or can optionally include other steps or units inherent to the process, method, product or device.

[0023] In the present application, "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The phrase appears in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment to other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in the present application can be combined with other embodiments.

[0024] In order to better understand the business product determination method provided by the embodiments of the present application, the scene of applying the business product determination method will be briefly introduced first. In the process of treating skin lesions by using Wood's light guidance, the interpretation of fluorescence intensity depends on the subjective experience of doctors, lacks a quantitative analysis system, and it is difficult to accurately distinguish the fluorescence intensity threshold boundary of the non-infected area, the mild infected area and the severe infected area. In addition, the light source output adopts a homogeneous irradiation mode in the treatment using red light, and the differential dose regulation of the infected area cannot be realized. Therefore, in the non-infected area of the skin, normal tissue damage may be caused by excessive light dose, while in the severe infected area, the bactericidal effect is affected due to insufficient dose, and then the boundary misjudgment leads to the treatment area deviation, causing the contradictory phenomenon of "over-treatment and insufficient treatment".

[0025] The Wood's light-based skin lesion treatment light source determination method is applied to a Wood's light-based skin lesion treatment light source determination device, Figure 1 The overall flowchart of the Wood's light-based skin lesion treatment light source determination method is shown. As Figure 1 shown, it includes: S0, applying 5-aminolevulinic acid to the target skin area; Wherein, the 5-aminolevulinic acid (ALA) is dissolved in physiological saline or temperature-sensitive gel, and then the 5-aminolevulinic acid liquid is applied to the target skin area. 118 mg of 5-aminolevulinic acid can be dissolved in 2.5 ml of physiological saline or temperature-sensitive gel, and then wet the target skin area, which is used to make the Wood's light analyze the image information of the target area and the photodynamic therapy.

[0026] S1, acquiring Wood's light irradiation image data of the target skin area.

[0027] S2, inputting the Wood's light irradiation image data into a pre-trained lesion degree recognition model to obtain a lesion recognition result.

[0028] S3, determining lesion skin division area data according to the lesion recognition result.

[0029] S4, determining the light source output dose of different infected areas according to the lesion skin division area data.

[0030] Wherein, the present example can excite the characteristic fluorescence of protoporphyrin IX in the skin tissue by the specific waveband ultraviolet light emitted by the Wood's light, and capture the fluorescence image data of the target area by the high-resolution camera.

[0031] The pre-trained lesion degree recognition model can be constructed based on a federated learning framework, which aggregates the annotation data of multiple medical institutions, extracts the spatial features of the fluorescence intensity distribution using a convolutional neural network, and adapts the imaging differences of different devices using a transfer learning technique. The lesion recognition result output by the model can include the concentration gradient information of protoporphyrin IX and the heterogeneity features of recognizing fluorescence distribution.

[0032] In determining the lesion skin division area data, the recognition result can be first reconstructed in three-dimensional space, the two-dimensional fluorescence image is mapped to a skin curved coordinate system, and the target area is divided into 0.5cm*0.5cm detection units through an adaptive grid algorithm. The fluorescence intensity mean and standard deviation can be calculated synchronously in each unit, and the spatial boundaries of the non-infected area, the mild and severe infection areas can be distinguished by combining the double-parameter threshold.

[0033] In the light source dose determination, a dynamic compensation strategy can be used. Based on the area proportion of each infection area in the division area data, a preset dose gradient table can be matched, an exponential decay algorithm can be introduced for the transition zone, and a 0.1mm-level precision light spot regulation can be realized through a programmable micromirror array.

[0034] In this example, by obtaining the Wood's light irradiation image data of the target skin area, analyzing the fluorescence features using the pre-trained lesion degree recognition model, generating the lesion recognition result, and combining the three-dimensional reconstruction and dynamic partitioning technology to establish the lesion skin division area data, the differential light source dose regulation of the infection partition is finally realized. According to the Wood's light irradiation image data of the target skin area, the boundary ranges of the non-infected area, the mild infection area and the severe infection area can be distinguished, the light dose output strategy can be dynamically matched according to the fluorescence intensity gradient of the partition, the overexposure of normal tissues can be avoided, and the sterilization efficiency of the severe infection area can be improved, thereby solving the technical problems of "insufficient treatment" and "excessive damage" in traditional homogeneous treatment, and the accuracy of light source treatment is not high.

[0035] In one possible implementation, the determining lesion skin division area data according to the lesion recognition result comprises: S301, reconstructing the lesion recognition result in three-dimensional space to obtain a lesion skin heat map.

[0036] S302, decomposing the heat map using an adaptive grid division algorithm to obtain a plurality of independent detection units.

[0037] S303, calculating the fluorescence intensity of the detection units to obtain the fluorescence intensity standard deviation and the fluorescence intensity mean.

[0038] S304, determining the lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean, specifically comprising: when the fluorescence intensity standard deviation σ < 15 and the fluorescence intensity mean value μ < 50 a.u., mark the lesion skin divided area type as uninfected area; when the fluorescence intensity standard deviation σ < 30 and the fluorescence intensity mean value 50 ≤ μ < 150 a.u., mark the lesion skin divided area type as mild infection area; when the fluorescence intensity standard deviation σ < 50 and the fluorescence intensity mean value μ > 150 a.u., mark the lesion skin divided area type as severe infection area.

[0039] S305, set the lesion skin divided area type set to obtain lesion skin divided area data.

[0040] In the three-dimensional space reconstruction stage, the three-dimensional heat map of the lesion area can be generated by fusing multi-view Wood's light image data and skin surface curvature parameters and using a stereo matching algorithm.

[0041] For the adaptive grid division algorithm, a dynamic quadtree partitioning strategy can be used to automatically adjust the grid density according to the gradient change of the heat map: high-density grids of 0.3 cm x 0.3 cm can be generated in areas with intense fluorescence intensity fluctuations, while low-density grids of 0.8 cm x 0.8 cm can be used in uniform areas. After the fluorescence intensity data in each independent detection unit is denoised by Gaussian filtering, the mean value μ and the standard deviation σ are calculated synchronously, where μ reflects the overall protoporphyrin IX accumulation level of the unit, and σ represents the uniformity of fluorescence distribution, which can reduce the computational complexity while ensuring the adaptive analysis capability for complex lesion morphology.

[0042] In determining the area type, the two-parameter threshold in step S304 can be associated and verified with the clinical pathology database: when the μ of a certain unit < 50 a.u. and σ < 15, it indicates that the protoporphyrin IX metabolism of the area is normal and the distribution is uniform, and it is determined as uninfected area; while when μ > 150 a.u., even if σ is high (< 50), it is still marked as severe infection area, because the local accumulation of high concentration protoporphyrin IX has exceeded the cell metabolic capacity. This determination logic effectively distinguishes the clinical scenarios of "diffuse low infection" and "focal severe infection", and can overcome the misjudgment risk of single mean value threshold method. The finally generated divided area data is stored in a hierarchical topology structure, including the geometric boundary, center coordinates and adjacent area correlation of each infection area, providing structured input for the light source dose regulation module.

[0043] In one possible implementation, the training process of the pre-trained lesion severity recognition model includes: S201, acquire first historical Wood's light image data, the first historical Wood's light image data contains lesion region image labeled with protoporphyrin IX fluorescence intensity grading; S202, construct a first lesion degree recognition model of a cloud data center according to the first historical Wood's light image data; S203, distribute the first lesion degree recognition model to each client node, each client node respectively trains the first lesion degree recognition model according to the client local Wood's light image data to obtain first lesion degree recognition model update parameters; S204, upload the first lesion degree recognition model update parameters to the cloud data center, and the cloud data center performs global model aggregation on the first lesion degree recognition model according to the first intrusion detection model update parameters to obtain a second lesion degree recognition model; S205, repeat the steps of distributing the first lesion degree recognition model to each client node, each client node... obtaining first lesion degree recognition model update parameters to the cloud data center... obtaining second lesion degree recognition model until iteration k times, obtaining a lesion degree recognition model.

[0044] In the example, the collaborative modeling of the data "available but invisible" is realized by using the federated learning framework, the value of the Wood's light image data distributed in medical institutions at all levels can be fully utilized, and the legal risk of sensitive medical data transmission across institutions can be avoided. The model continuously absorbs data characteristics of different imaging conditions and device models during iteration, which improves the robustness of low-quality images in primary medical institutions, and suppresses the overfitting problem caused by a single data source through a dynamic parameter aggregation mechanism. The finally generated model can cross the device gap and stably output high-precision lesion recognition results in various clinical scenarios, providing a reliable analysis basis for subsequent precise treatment.

[0045] In one possible implementation, the first lesion degree recognition model of the cloud data center is constructed according to the first historical Wood's light image data, including: S2021, data cleaning is performed on the first historical Wood's light image data to obtain cleaned first historical Wood's light image data; S2022, format conversion is performed on the cleaned first historical Wood's light image data to obtain second historical Wood's light image data; S2023, feature encoding is performed on the second historical Wood's light image data to obtain third historical Wood's light image data; S2024, determine a network traffic feature matrix according to the third historical Wood's light image data; S2025, constructing a first lesion degree recognition model of the cloud data center by using a CNN neural network according to the network traffic feature matrix.

[0046] In constructing the initial lesion recognition model, the original historical data can be first subjected to multi-stage preprocessing: in the data cleaning stage, image noise can be eliminated by an adaptive filtering algorithm, and at the same time, the histogram equalization technique can be used to compensate for the difference in illumination conditions of different medical institutions, to generate standardized first historical data. In the format conversion link, the heterogeneous image formats such as DICOM and JPEG can be uniformly converted into HDF5 format with 256-bit depth, to ensure the consistency of the input for subsequent processing. This process preserves key clinical annotation information (such as protoporphyrin IX grading labels) through a metadata mapping table, to avoid distortion of data semantics in the format conversion process.

[0047] In the feature encoding stage, multi-scale convolution kernel groups (such as 3x3 and 5x5 parallel convolution) can be used to extract the spatial-spectral features of the fluorescence image, and the fluorescence intensity of each pixel point and the spatial distribution features of its 8-neighborhood can be encoded into a 128-dimensional feature vector. After integration by a spatial pyramid pooling layer, the third historical data with regularized dimensions are generated, wherein each data sample corresponds to a multi-modal feature matrix containing texture features, morphological parameters and fluorescence gradient changes of the lesion area. The multi-modal feature matrix can carry the concentration information of protoporphyrin IX.

[0048] The finally constructed CNN model can use a residual connection structure to establish cross-layer feature reuse channels between convolution layers. The network can gradually abstract multi-scale feature expressions from microscopic fluorescence spots (50 μm level) to macroscopic infection areas (cm level) by alternately stacking 3x3 convolution layers and maximum pooling layers.

[0049] Consistent with the above, please refer to Figure 2 , Figure 2 A structure diagram of a skin lesion treatment light source determination device based on a Wood's lamp is provided for the embodiments of the present application. As shown in Figure 2 , the device comprises: A preparation unit 0 for applying 5-aminolevulinic acid to a target skin area; An acquisition unit 1 for acquiring Wood's lamp irradiation image data of the target skin area; A first processing unit 2 for inputting the Wood's lamp irradiation image data into a pre-trained lesion degree recognition model to obtain a lesion recognition result; A second processing unit 3 for determining lesion skin partition region data according to the lesion recognition result; A determination unit 4 for determining light source output doses of different infection sub-areas according to the lesion skin partition region data.

[0050] In a possible implementation, in the aspect of determining the lesion skin division area data according to the lesion identification result, the second processing unit is configured to: perform three-dimensional space reconstruction on the lesion identification result to obtain a lesion skin heat map; decompose the heat map by using an adaptive grid division algorithm to obtain a plurality of independent detection units; perform fluorescence intensity calculation on the detection units to obtain a fluorescence intensity standard deviation and a fluorescence intensity mean value; determine a lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean value; collect the lesion skin division area types to obtain lesion skin division area data.

[0051] In a possible implementation, in the aspect of determining the lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean value, the second processing unit is configured to: when the fluorescence intensity standard deviation σ < 15 and the fluorescence intensity mean value μ < 50 a.u., mark the lesion skin division area type as an uninfected area; when the fluorescence intensity standard deviation σ < 30 and the fluorescence intensity mean value 50 ≤ μ < 150 a.u., mark the lesion skin division area type as a mild infection area; when the fluorescence intensity standard deviation σ < 50 and the fluorescence intensity mean value μ > 150 a.u., mark the lesion skin division area type as a severe infection area.

[0052] In a possible implementation, in the aspect of inputting the Wood's light image data into the pre-trained lesion degree identification model to obtain a lesion identification result, the training process of the pre-trained lesion degree identification model in the first processing unit includes: obtain first historical Wood's light image data, the first historical Wood's light image data containing lesion area images labeled with protoporphyrin IX fluorescence intensity grading; construct a first lesion degree identification model of a cloud data center according to the first historical Wood's light image data; distribute the first lesion degree identification model to each client node, and each client node respectively trains the first lesion degree identification model according to client local Wood's light image data to obtain first lesion degree identification model update parameters; upload the first lesion degree identification model update parameters to the cloud data center, and the cloud data center performs global model aggregation on the first lesion degree identification model according to the first intrusion detection model update parameters to obtain a second lesion degree identification model. The steps of repeatedly issuing the first lesion degree recognition model to each client node, each client node obtaining first lesion degree recognition model update parameters, uploading the first lesion degree recognition model update parameters to the cloud data center, and obtaining a second lesion degree recognition model are repeated k times until iteration to obtain a lesion degree recognition model.

[0053] In one possible implementation, in the aspect of constructing the first lesion degree recognition model of the cloud data center according to the first historical Wood's light image data, the first processing unit is configured to: perform data cleaning on the first historical Wood's light image data to obtain cleaned first historical Wood's light image data; perform format conversion on the cleaned first historical Wood's light image data to obtain second historical Wood's light image data; perform feature encoding on the second historical Wood's light image data to obtain third historical Wood's light image data; determine a network flow feature matrix according to the third historical Wood's light image data; construct the first lesion degree recognition model of the cloud data center by using a CNN neural network according to the network flow feature matrix.

[0054] For the above embodiments, please refer to Figure 3 , Figure 3 A terminal structure schematic diagram provided by the embodiments of the present application is shown in the figure, which includes a processor, an input device, an output device, and a memory, and the processor, the input device, the output device, and the memory are connected with each other, wherein the memory is used for storing a computer program, the computer program includes program instructions, the processor is configured to invoke the program instructions, and the above program includes instructions for executing the following steps: obtain Wood's light irradiation image data of a target skin area; input the Wood's light irradiation image data into a pre-trained lesion degree recognition model to obtain a lesion recognition result; determine lesion skin partition region data according to the lesion recognition result; determine light source output doses of different infection partitions according to the lesion skin partition region data.

[0055] In this example, by acquiring the Wood's light irradiation image data of the target skin area, using the pre-trained lesion degree recognition model to analyze the fluorescence features, generating the lesion recognition result, and combining the three-dimensional reconstruction and dynamic partitioning technology to establish the lesion skin partition area data, the differential light source dose regulation of the infection partition is finally realized. According to the Wood's light irradiation image data of the target skin area, the boundary range of the uninfected area, the mild infection area and the severe infection area can be distinguished, the light dose output strategy is dynamically matched according to the fluorescence intensity gradient of the partition, the over-irradiation of normal tissues is avoided, and the sterilization efficiency of the severe infection area is improved, so as to solve the technical problems of "insufficient treatment" and "excessive damage" in the traditional homogenization treatment, and the accuracy of light source treatment is not high.

[0056] The above mainly introduces the scheme of the embodiments of the present application from the perspective of the execution process of the method. It can be understood that the terminal includes hardware structure and / or software module corresponding to the execution of each function in order to realize the above functions. Those skilled in the art should easily realize that, in combination with the unit and algorithm steps of each example described in the embodiments provided herein, the present application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driven hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0057] The embodiments of the present application can divide the terminal into functional units according to the above method examples, for example, each functional unit can be divided according to each function, or two or more functions can be integrated into one processing unit. The integrated unit can be realized in the form of hardware or software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative, and is only a logical function division. There can be another division method in actual implementation.

[0058] The embodiments of the present application also provide a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program causes the computer to execute part or all steps of any one of the Wood's light based skin lesion treatment light source determination methods described in the above method embodiments.

[0059] The embodiments of the present application also provide a computer program product, which includes a non-transitory computer readable storage medium storing a computer program, and the computer program causes the computer to execute part or all steps of any one of the Wood's light based skin lesion treatment light source determination methods described in the above method embodiments.

[0060] It should be noted that, for the foregoing method embodiments, the sequences of the described actions are not necessarily required to achieve the objects of the application, and certain steps can be performed in other sequences or even concurrently. Additionally, the described embodiments are merely provided as examples, and not all of the actions described are necessarily required to achieve desired results.

[0061] In the above embodiments, the description of each embodiment is focused on different aspects, and the parts not described in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0062] In several embodiments provided in the present application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, another division manner can be adopted. For example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interfaces, apparatuses or units, and can be electrical or other forms.

[0063] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. can be located in one place or distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0064] In addition, the functional units in each embodiment of the application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware or software program modules.

[0065] If the integrated unit is realized in the form of a software program module and sold or used as an independent product, it can be stored in a computer readable memory. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a memory and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned memory includes: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0066] A person of ordinary skill in the art can understand that all or part of the steps of the various methods of the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer readable memory, and the memory can include a flash disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, etc.

[0067] The embodiments of the present application are described in detail above, and the principles and implementation manners of the present application are described by applying specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea; meanwhile, for those of ordinary skill in the art, according to the idea of the present application, the specific implementation manner and application range will be changed, and the above description of the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for determining a light source for treating skin lesions based on a Wood's lamp, characterized in that: The method comprises the following steps: applying 5-aminolevulinic acid to a target skin area; obtaining Wood's light irradiation image data of the target skin area; inputting the Wood's light irradiation image data into a pre-trained lesion degree recognition model to obtain a lesion recognition result; determining lesion skin division area data according to the lesion recognition result; determining the light source output dose of different infection subareas according to the lesion skin division area data.

2. The method of claim 1, wherein the Wood's light based skin lesion treatment light source determination method further comprises: The step of determining lesion skin division area data according to the lesion recognition result comprises the following steps: reconstructing the lesion recognition result in a three-dimensional space to obtain a lesion skin heat map; decomposing the heat map by using an adaptive grid division algorithm to obtain a plurality of independent detection units; calculating the fluorescence intensity of the detection units to obtain a fluorescence intensity standard deviation and a fluorescence intensity mean value; determining the type of lesion skin division area according to the fluorescence intensity standard deviation and the fluorescence intensity mean value; obtaining lesion skin division area data by collecting the types of lesion skin division area.

3. The method of claim 2, wherein the Wood's light based skin lesion treatment light source determination is based on a color of the skin lesion. The step of determining the type of lesion skin division area according to the fluorescence intensity standard deviation and the fluorescence intensity mean value comprises the following steps: when the fluorescence intensity standard deviation σ is less than 15 and the fluorescence intensity mean value μ is less than 50 a.u., marking the type of lesion skin division area as an uninfected area; when the fluorescence intensity standard deviation σ is less than 30 and the fluorescence intensity mean value 50≤μ<150 a.u., marking the type of lesion skin division area as a mild infection area; when the fluorescence intensity standard deviation σ is less than 50 and the fluorescence intensity mean value μ is greater than 150 a.u., marking the type of lesion skin division area as a severe infection area.

4. The method of claim 1, wherein the Wood's light based skin lesion treatment light source determination is based on a color of the skin lesion. The training process of the pre-trained lesion degree recognition model comprises the following steps: obtaining first historical Wood's light image data, wherein the first historical Wood's light image data contains lesion area images labeled with protoporphyrin IX fluorescence intensity grading; constructing a first lesion degree recognition model of a cloud data center according to the first historical Wood's light image data; downloading the first lesion degree recognition model to each client node, and each client node trains the first lesion degree recognition model according to local Wood's light image data to obtain first lesion degree recognition model update parameters; uploading the first lesion degree recognition model update parameters to the cloud data center, and the cloud data center aggregates the first lesion degree recognition model globally according to the first intrusion detection model update parameters to obtain a second lesion degree recognition model; repeating the steps of downloading the first lesion degree recognition model to each client node, each client node... obtaining first lesion degree recognition model update parameters to uploading the first lesion degree recognition model update parameters to the cloud data center... obtaining a second lesion degree recognition model until iteration k times, to obtain a lesion degree recognition model.

5. The method of claim 4, wherein the Wood's light based skin lesion treatment light source determination is based on a color of the skin lesion. The step of constructing a first lesion degree recognition model of a cloud data center according to the first historical Wood's light image data comprises the following steps: performing data cleaning on the first historical Wood's light image data to obtain cleaned first historical Wood's light image data; The first historical Wood's lamp image data after cleaning is format-converted to obtain second historical Wood's lamp image data; The second historical Wood's lamp image data is feature-encoded to obtain third historical Wood's lamp image data; The network traffic feature matrix is determined according to the third historical Wood's lamp image data; The first lesion degree recognition model of the cloud data center is constructed by using the CNN neural network according to the network traffic feature matrix.

6. A Wood's light based skin lesion treatment light source determination device, characterized in that, The device comprises: The preparation unit is configured to apply 5-aminolevulinic acid to the target skin area; The acquisition unit is configured to acquire Wood's lamp irradiation image data of the target skin area; The first processing unit is configured to input the Wood's lamp irradiation image data into the pre-trained lesion degree recognition model to obtain a lesion recognition result; The second processing unit is configured to determine lesion skin division area data according to the lesion recognition result; The determination unit is configured to determine the light source output dose of different infection subareas according to the lesion skin division area data.

7. The Wood's light based skin lesion treatment light source determination device of claim 6, wherein, In the aspect of determining lesion skin division area data according to the lesion recognition result, the second processing unit is configured to: Reconstruct the lesion recognition result in a three-dimensional space to obtain a lesion skin heat map; Decompose the heat map by using an adaptive grid division algorithm to obtain a plurality of independent detection units; Calculate the fluorescence intensity of the detection units to obtain a fluorescence intensity standard deviation and a fluorescence intensity mean value; Determine the lesion skin division area type according to the fluorescence intensity standard deviation and the fluorescence intensity mean value; Collect the lesion skin division area types to obtain lesion skin division area data.

8. The Wood's light based skin lesion treatment light source determination device of claim 6, wherein, In the aspect of inputting the Wood's lamp irradiation image data into the pre-trained lesion degree recognition model to obtain a lesion recognition result, the training process of the pre-trained lesion degree recognition model in the first processing unit comprises: Obtain first historical Wood's lamp image data, which contains lesion area images labeled with protoporphyrin IX fluorescence intensity grading; Construct a first lesion degree recognition model of the cloud data center according to the first historical Wood's lamp image data; Distribute the first lesion degree recognition model to each client node, and each client node trains the first lesion degree recognition model according to the client local Wood's lamp image data to obtain first lesion degree recognition model update parameters; Upload the first lesion degree recognition model update parameters to the cloud data center, and the cloud data center performs global model aggregation on the first lesion degree recognition model according to the first intrusion detection model update parameters to obtain a second lesion degree recognition model; Repeat the steps of distributing the first lesion degree recognition model to each client node, each client node... obtaining first lesion degree recognition model update parameters to the cloud data center... obtaining second lesion degree recognition model until iteration k times to obtain a lesion degree recognition model.

9. A terminal, characterized by comprising: The computer readable storage medium stores a computer program, and the computer program comprises program instructions. The program instructions, when executed by a processor, cause the processor to perform the Wood's lamp-based skin lesion treatment light source determination method according to any one of claims 1-5.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program comprises program instructions. The program instructions, when executed by a processor, cause the processor to perform the Wood's lamp-based skin lesion treatment light source determination method according to any one of claims 1-5.

Images