Real-time feedback and adaptive phototherapy parameter adjustment method based on AI

Abstract

The invention discloses a real-time feedback and self-adaptive phototherapy parameter adjustment method based on AI, and the method comprises the steps: S1, collecting multi-source data of a patient in a phototherapy process, and forming a multi-source biological feedback data set; s2, constructing an AI algorithm model, inputting the multi-source biological feedback data set into the AI algorithm model, and outputting the change trend of the physiological state of the human body; s3, setting a phototherapy effect evaluation index system, comparing the change trend of the physiological state of the human body with a preset phototherapy effect evaluation index, and calculating a deviation value between a current phototherapy parameter and a parameter required for realizing an optimal treatment effect; s4, dynamically adjusting the phototherapy parameters through a reinforcement learning algorithm according to the deviation value, and determining a parameter combination of phototherapy in the next stage; and S5, the optimized phototherapy parameters are sent to phototherapy equipment for adaptive dynamic adjustment. The problems that phototherapy parameters cannot adapt to individual differences and physiological state changes of patients, so that the treatment effect is poor, and safety and comfort are insufficient are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of phototherapy adjustment, and in particular to an AI-based real-time feedback and adaptive phototherapy parameter adjustment method. Background Art

[0002] In modern medicine, phototherapy, as an important treatment method, is widely used to treat skin diseases, promote wound healing, and alleviate inflammation. However, traditional phototherapy often lacks precision and dynamic adjustment capabilities in parameter settings, making it difficult to meet the needs of individual patients and the dynamic changes in their condition during treatment. For example, when treating skin diseases, different patients have different skin types, disease severity, and light sensitivity. If uniform phototherapy parameters are used, some patients may not receive adequate treatment, or even experience adverse reactions, delaying treatment. This not only reduces the effectiveness of phototherapy but also causes additional pain and financial burden for patients.

[0003] Traditional phototherapy technology typically sets phototherapy parameters, such as fixed wavelength, power, and irradiation time, based on the doctor's experience and clinical routine standards. Although this method is simple to operate and has a certain degree of therapeutic effect on some patients, it has many drawbacks. On the one hand, the doctor's experience is subjective and limited, and the judgments of different doctors vary greatly, making it difficult to guarantee that each treatment will achieve the best results. On the other hand, it cannot monitor the patient's physiological changes during phototherapy in real time and cannot adjust parameters based on the patient's immediate response. For example, during treatment, the patient may experience changes in pain threshold or abnormal skin temperature, but traditional solutions cannot respond in time, which may cause discomfort to the patient and even damage skin tissue. In addition, traditional phototherapy does not fully consider the differences in the mechanisms of action of different phototherapy parameters on human tissues, making it difficult to achieve precise treatment.

[0004] With the advancement of science and technology, existing phototherapy technology has improved to a certain extent. Some phototherapy devices are beginning to be equipped with simple sensors that can monitor some physiological indicators, such as skin temperature. However, the monitoring data is single and cannot fully reflect the patient's physiological state. At the same time, some studies have attempted to use algorithms to optimize phototherapy parameters, but these algorithms are often too simple to effectively handle complex human physiological data and treatment needs. For example, some algorithms adjust parameters based only on a single physiological indicator, ignoring the interrelationships between different indicators, resulting in the adjusted parameters failing to achieve the optimal treatment effect. Moreover, existing technologies still have shortcomings in addressing individual patient differences and cannot provide personalized treatment plans for each patient.

[0005] Therefore, this application proposes an AI-based real-time feedback and adaptive phototherapy parameter adjustment method to achieve personalized and precise phototherapy to meet the diverse needs of clinical treatment. Summary of the Invention

[0006] Based on the above, this application discloses an AI-based real-time feedback and adaptive phototherapy parameter adjustment method, including:

[0007] S1. Deploy a biosensor group to collect multi-source data of patients during phototherapy and form a multi-source biofeedback dataset after preprocessing;

[0008] S2. Build an AI algorithm model, input multi-source biofeedback data sets into the AI algorithm model, and output the changing trend of the human body's physiological state;

[0009] S3. Preset the expected treatment goals for phototherapy. Combined with the mechanisms of action of different phototherapy wavelengths, powers, and irradiation times on human tissues, establish a phototherapy effect evaluation index system. Compare the trend of changes in human physiological status obtained through AI algorithm model analysis with the preset phototherapy effect evaluation indexes. Calculate the deviation between the current phototherapy parameters and the parameters required to achieve the optimal treatment effect through the error back propagation algorithm.

[0010] S4. Dynamically adjust the phototherapy parameters using a reinforcement learning algorithm based on the deviation value to determine the parameter combination for the next stage of phototherapy;

[0011] S5. Send the optimized phototherapy parameters to the phototherapy device, repeat the operation at the set time interval, collect feedback data in real time to optimize the phototherapy parameters, and perform adaptive dynamic adjustment.

[0012] Preferably, the biosensor group includes a thermistor, a blood flow sensor and a skin galvanic response sensor; the thermistor is used to monitor changes in skin temperature; the blood flow sensor monitors local blood flow velocity and changes; the skin galvanic response sensor is used to sense skin electrical activity to reflect changes in pain threshold, and the acquired raw sensor data values are preprocessed to form a multi-source biofeedback data set.

[0013] Preferably, the preprocessing of the collected raw data is specifically as follows:

[0014] The raw data of skin temperature, blood flow, and skin electrical response type are obtained and filtered through the constructed composite filter model; the raw data is formed into a raw data sequence T(n), and wavelet decomposition is performed to decompose it into sub-band coefficients W of different frequencies j (k), where j is the number of layers of wavelet decomposition, k is the index of the subband coefficient, and the threshold T is set. th , perform threshold processing on the high-frequency sub-band coefficients to remove the high-frequency noise components caused by sensor noise, and obtain the processed high-frequency sub-band coefficients W j ′(k), the formula is Through wavelet reconstruction, the target data sequence T′(n) after preliminary denoising is obtained, and T′(n) is filtered using Kalman filtering. Through prediction and updating, a multi-source biofeedback data set is formed.

[0015] Preferably, the AI algorithm model in S2 is constructed by the fusion of convolutional neural networks and long short-term memory networks. By performing modal separation on the preprocessed multi-source biofeedback data set, skin temperature, blood flow data and skin electrical response data reflecting the pain threshold are mapped to different subspaces respectively; in each subspace, local spatial features are extracted using a convolutional neural network, and the local spatial features extracted by the convolutional neural network are input into the long short-term memory network. The memory unit and the gating mechanism are used to analyze the dependency between features at different times, learn the long-term trends and short-term fluctuations in the time series, integrate the time series characteristics of each subspace, and output the trend of changes in the human physiological state.

[0016] Preferably, in each subspace, when using a convolutional neural network to extract local spatial features, for the input subspace data X, a convolution operation is performed at the lth convolution layer, and the formula is: in Represents the value of the output feature map of the lth convolutional layer at the (i, j) position, is the weight of the l-th layer convolution kernel at position (a, b), b l is the bias term, p and q are the convolution kernel sizes; the convolution kernel weight is dynamically adjusted according to the characteristics of different subspace data, and the formula is Where σ is the adaptive adjustment value of the fluctuation range, and the importance score of each position feature is calculated through the attention mechanism. The formula is: where ω l It is a learnable parameter that weights the convolution output according to the score to highlight key local spatial features.

[0017] Preferably, in each subspace, the local spatial features extracted by the convolutional neural network are sequentially input into the long short-term memory network in time series order. Input gate i of the long short-term memory network t Control the current input features The part that needs to be added to the memory unit, the forget gate f t Determine the memory unit C at the previous moment t-1 The information that needs to be retained in the output gate o t The information in the control memory unit is used to generate the hidden state h at the current moment t , the formula is: h t =o t ⊙tanh(C t ); For the time series features of multiple subspaces, the hidden state H is integrated by weighted fusiont , the formula is: k is the number of subspaces, α s The weight of the subspace importance is dynamically adjusted by H t The analysis outputs the changing trend of human physiological state.

[0018] Preferably, in said S3, the human physiological state change trend data T obtained by the AI algorithm model analysis is td Compared with the preset phototherapy effect evaluation index data T tr Compare and calculate the degree of difference D. The formula is: Where n is the dimension of the evaluation index in the subspace, w i Dynamically assign weights to the importance of the i-th dimension in the evaluation of phototherapy effects. The formula is: ρ i is the correlation coefficient between the features in the subspace and the phototherapy effect. The difference degree D is used to measure the gap between the current physiological state change trend and the ideal phototherapy effect. The error back propagation algorithm is used to calculate the deviation between the current phototherapy parameters and the parameters required to achieve the best treatment effect. The formula is: in is the kth phototherapy parameter Through back propagation calculation, the deviation value corresponding to each phototherapy parameter is obtained.

[0019] Preferably, in S4, the phototherapy parameters are dynamically adjusted using a reinforcement learning algorithm based on the deviation values corresponding to each phototherapy parameter. By obtaining the penetration depth values of different wavelengths of light on different tissues, the changes in power and irradiation time, multiple sets of constraints are set, with maximizing the treatment effect and ensuring the safety and comfort of the patient as the reward function. The optimal solution is searched in the parameter space that meets the constraints, and the wavelength combination, power and irradiation time of the next stage of phototherapy are determined.

[0020] Preferably, the reward function R is to maximize the treatment effect and ensure the safety and comfort of the patient, and guide the search for the optimal solution. The formula is: Among them E t is a comprehensive treatment index, S t is the skin tolerance value, C s is the subjective comfort value, δ t is the time attenuation factor, P is the current phototherapy parameter deviation value, P r is the comparison parameter deviation value, P max and P minare the maximum deviation value and the minimum deviation value of the parameter deviation value respectively. The optimal solution is searched by the particle swarm optimization algorithm in the parameter space that meets the constraints. Each particle represents a set of phototherapy parameters. In each iteration, the reward function value R corresponding to each particle is calculated, and the phototherapy parameter combination that maximizes the reward function R is gradually searched to achieve the maximum treatment effect.

[0021] Preferably, the optimized phototherapy parameters are sent to the phototherapy device in S5, and the light-emitting components of the phototherapy device are adjusted so that the device outputs phototherapy according to the new wavelength, power and irradiation time. The adjusted phototherapy parameter information is fed back to the patient terminal and medical staff monitoring platform, so that patients and medical staff can understand the progress of phototherapy in real time.

[0022] Compared with the prior art, the technical solution of this application has the following technical effects:

[0023] The present invention deploys a biosensor group consisting of a thermal sensor, a blood flow sensor, and a galvanic skin response sensor during the phototherapy process to collect multi-source data such as skin temperature changes, local blood flow velocity and changes, and pain threshold changes in real time. This data is then preprocessed to form a multi-source biofeedback dataset. This data is then input into an AI algorithm model constructed by fusing a convolutional neural network and a long-short-term memory network. The model then performs modal separation, feature extraction, and time series feature analysis to output the trend of changes in the human body's physiological state. This trend is compared with preset phototherapy effect evaluation indicators, and the error backpropagation algorithm is used to calculate the deviation value of the phototherapy parameters. The phototherapy parameters are then dynamically adjusted based on a reinforcement learning algorithm. This achieves precise dynamic adjustment of the phototherapy parameters, allowing the phototherapy process to match the patient's physiological state changes in real time, improving the accuracy and effectiveness of treatment.

[0024] This method uses an AI algorithm model to output the changing trend of the human body's physiological state, compares the trend with a preset phototherapy effect evaluation index, and uses the error back propagation algorithm to calculate the deviation between the current phototherapy parameters and the optimal parameters. Then, with maximizing the treatment effect and ensuring patient safety and comfort as the reward function, the method searches for the optimal solution within the parameter space that meets the constraints to dynamically adjust the phototherapy parameters. This achieves precise and efficient adjustment of the phototherapy parameters, can adapt in real time to the patient's physiological state, and significantly improves the treatment effect. For example, when treating skin diseases, the wavelength, power, and irradiation time of the phototherapy can be adjusted quickly and accurately to accelerate the improvement of the condition.

[0025] This method utilizes an AI algorithm model to deeply mine and analyze multi-source biofeedback datasets. Taking into account the differences in physiological states and disease characteristics of different patients, combined with a reward function, it searches for the optimal solution within the parameter space, enabling customized phototherapy plans for each patient. This approach fully considers the individual characteristics of each patient, improving the targeted nature of phototherapy and making treatment plans more tailored to their actual needs. For example, using skin rehabilitation treatment for patients with different physical conditions as an example, phototherapy parameters can be precisely adjusted based on their individual physiological responses and recovery status, improving treatment effectiveness and cure rates.

[0026] The present invention dynamically adjusts phototherapy parameters by providing real-time feedback on the changing trends of the human body's physiological state, ensuring that the phototherapy process is always in the optimal treatment state, thus avoiding the prolonged treatment cycle and waste of resources caused by traditional phototherapy due to fixed parameters or untimely adjustment. The reward function searches for the optimal solution with the goal of maximizing the treatment effect, which can quickly find the most suitable phototherapy parameter combination for the patient and improve treatment efficiency. At the same time, the rational use of phototherapy resources reduces unnecessary treatment time and energy consumption, thereby improving the overall utilization rate of medical resources. For example, in the treatment of batches of patients, the use of phototherapy equipment can be arranged more efficiently, improving overall treatment efficiency.

[0027] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application so that it can be implemented in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following is a detailed description of the preferred embodiment of the present application in conjunction with the accompanying drawings.

[0028] Based on the detailed description of the specific embodiments of the present application in conjunction with the accompanying drawings below, those skilled in the art will become more aware of the above and other objects, advantages and features of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without inventive work. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0030] Figure 1 This is a structural diagram of an AI-based real-time feedback and adaptive phototherapy parameter adjustment method of the present invention;

[0031] Figure 2 This is a schematic diagram of the patient's initial facial inflammation scan;

[0032] Figure 3 Schematic diagram of a scan of a patient's face during the inflammatory phase of treatment;

[0033] Figure 4 A diagram of the wavelength, power, and exposure time of the phototherapy device during the third treatment;

[0034] Figure 5 This is a comparison chart of changes in skin inflammation area. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. In the following description, specific details such as specific configurations and components are provided only to help fully understand the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, for clarity and brevity, the description of known functions and structures has been omitted in the embodiments.

[0036] It should be understood that references throughout this specification to "one embodiment" or "this embodiment" mean that a particular feature, structure, or characteristic associated with the embodiment is included in at least one embodiment of the present application. Therefore, the appearance of "one embodiment" or "this embodiment" throughout this specification does not necessarily refer to the same embodiment. Furthermore, these particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0037] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity and does not in itself indicate the relationship between the various embodiments and / or settings discussed.

[0038] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" in this article describes another type of association object relationship, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the previous and subsequent associated objects are in an "or" relationship.

[0039] The term "at least one" in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, at least one of A and B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0040] It should also be noted that, in this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include," "comprises," or any other variations thereof are intended to cover non-exclusive inclusion.

[0041] Example 1

[0042] This embodiment mainly describes a method for real-time feedback and adaptive phototherapy parameter adjustment based on AI, such as Figure 1 As shown, including:

[0043] S1. Deploy a biosensor group to collect multi-source data of patients during phototherapy and form a multi-source biofeedback dataset after preprocessing;

[0044] S2. Build an AI algorithm model, input multi-source biofeedback data sets into the AI algorithm model, and output the changing trend of the human body's physiological state;

[0045] S3. Preset the expected treatment goals for phototherapy. Combined with the mechanisms of action of different phototherapy wavelengths, powers, and irradiation times on human tissues, establish a phototherapy effect evaluation index system. Compare the trend of changes in human physiological status obtained through AI algorithm model analysis with the preset phototherapy effect evaluation indexes. Calculate the deviation between the current phototherapy parameters and the parameters required to achieve the optimal treatment effect through the error back propagation algorithm.

[0046] S4. Dynamically adjust the phototherapy parameters using a reinforcement learning algorithm based on the deviation value to determine the parameter combination for the next stage of phototherapy;

[0047] S5. Send the optimized phototherapy parameters to the phototherapy device, repeat the operation at the set time interval, collect feedback data in real time to optimize the phototherapy parameters, and perform adaptive dynamic adjustment.

[0048] Furthermore, the biosensor group includes a thermistor, a blood flow sensor and a skin galvanic response sensor; the thermistor is used to monitor changes in skin temperature; the blood flow sensor monitors local blood flow velocity and changes; the skin galvanic response sensor is used to sense skin electrical activity to reflect changes in pain threshold, and the acquired raw sensor data values are preprocessed to form a multi-source biofeedback data set.

[0049] Furthermore, the collected raw data is preprocessed, specifically:

[0050] The raw data of skin temperature, blood flow, and skin electrical response type are obtained and filtered through the constructed composite filter model; the raw data is formed into a raw data sequence T(n), and wavelet decomposition is performed to decompose it into sub-band coefficients W of different frequenciesj (k), where j is the number of layers of wavelet decomposition, k is the index of the subband coefficient, and the threshold T is set. th , perform threshold processing on the high-frequency sub-band coefficients to remove the high-frequency noise components caused by sensor noise, and obtain the processed high-frequency sub-band coefficients W j ′(k), the formula is Through wavelet reconstruction, the target data sequence T′(n) after preliminary denoising is obtained, and T′(n) is filtered using Kalman filtering. Through prediction and updating, a multi-source biofeedback data set is formed.

[0051] Furthermore, the AI algorithm model in S2 is constructed by the fusion of convolutional neural networks and long short-term memory networks. By performing modal separation on the preprocessed multi-source biofeedback data set, skin temperature, blood flow data and skin electrical response data reflecting the pain threshold are mapped to different subspaces respectively; in each subspace, the convolutional neural network is used to extract local spatial features, and the local spatial features extracted by the convolutional neural network are input into the long short-term memory network. The memory unit and gating mechanism are used to analyze the dependency between features at different times, learn the long-term trends and short-term fluctuations in the time series, and integrate the time series characteristics of each subspace to output the trend of changes in the human physiological state.

[0052] Furthermore, in each subspace, when using the convolutional neural network to extract local spatial features, for the input subspace data X, the convolution operation is performed at the lth convolution layer, and the formula is: in Represents the value of the output feature map of the lth convolutional layer at the (i, j) position, is the weight of the l-th layer convolution kernel at position (a, b), b l is the bias term, p and q are the convolution kernel sizes; the convolution kernel weight is dynamically adjusted according to the characteristics of different subspace data, and the formula is Where σ is the adaptive adjustment value of the fluctuation range, and the importance score of each position feature is calculated through the attention mechanism. The formula is: where ω l It is a learnable parameter that weights the convolution output according to the score to highlight key local spatial features.

[0053] Furthermore, in each subspace, the local spatial features extracted by the convolutional neural network are sequentially input into the long short-term memory network in time series order. Input gate i of the long short-term memory network t Control the current input features The part that needs to be added to the memory unit, the forget gate f t Determine the memory unit C at the previous moment t-1 The information that needs to be retained in the output gate ot The information in the control memory unit is used to generate the hidden state h at the current moment t , the formula is: h t =o t ⊙tanh(C t ); For the time series features of multiple subspaces, the hidden state H is integrated by weighted fusion t , the formula is: k is the number of subspaces, α s The weight of the subspace importance is dynamically adjusted by H t The analysis outputs the changing trend of human physiological state.

[0054] Furthermore, the human physiological state change trend data T obtained by the AI algorithm model analysis in S3 td Compared with the preset phototherapy effect evaluation index data T tr Compare and calculate the degree of difference D. The formula is: Where n is the dimension of the evaluation index in the subspace, w i Dynamically assign weights to the importance of the i-th dimension in the evaluation of phototherapy effects. The formula is: ρ i is the correlation coefficient between the features in the subspace and the phototherapy effect. The difference degree D is used to measure the gap between the current physiological state change trend and the ideal phototherapy effect. The error back propagation algorithm is used to calculate the deviation between the current phototherapy parameters and the parameters required to achieve the best treatment effect. The formula is: in is the kth phototherapy parameter Through back propagation calculation, the deviation value corresponding to each phototherapy parameter is obtained.

[0055] Furthermore, in S4, the phototherapy parameters are dynamically adjusted using a reinforcement learning algorithm based on the deviation values corresponding to each phototherapy parameter. By obtaining the penetration depth values of different wavelengths of light on different tissues, the changes in power and irradiation time, multiple sets of constraints are set, with maximizing the treatment effect and ensuring the safety and comfort of the patient as the reward function. The optimal solution is searched in the parameter space that meets the constraints, and the wavelength combination, power and irradiation time of the next stage of phototherapy are determined.

[0056] Furthermore, the reward function R is used to maximize the treatment effect and ensure the safety and comfort of the patient, which guides the search for the optimal solution. The formula is: Among them E t is a comprehensive treatment index, S t is the skin tolerance value, C s is the subjective comfort value, δ tis the time attenuation factor, P is the current phototherapy parameter deviation value, P r is the comparison parameter deviation value, P max and P min are the maximum deviation value and the minimum deviation value of the parameter deviation value respectively. The optimal solution is searched by the particle swarm optimization algorithm in the parameter space that meets the constraints. Each particle represents a set of phototherapy parameters. In each iteration, the reward function value R corresponding to each particle is calculated, and the phototherapy parameter combination that maximizes the reward function R is gradually searched to achieve the maximum treatment effect.

[0057] Furthermore, in S5, the optimized phototherapy parameters are sent to the phototherapy equipment. By adjusting the light-emitting components of the phototherapy equipment, the equipment outputs phototherapy according to the new wavelength, power and irradiation time. The adjusted phototherapy parameter information is fed back to the patient terminal and medical staff monitoring platform, making it convenient for patients and medical staff to understand the progress of phototherapy in real time.

[0058] This embodiment describes in detail that the present application collects multi-source physiological data through biosensors, and through in-depth analysis by AI algorithms, realizes precise dynamic adjustment of phototherapy parameters, effectively improving the pertinence and effectiveness of phototherapy, and can adapt the optimal parameters according to the patient's real-time status, accelerate the disappearance of inflammation, and improve skin condition; at the same time, it ensures the safety and comfort of treatment and avoids excessive stimulation; it also reduces the risk of recurrence of inflammation, provides patients with long-lasting and stable therapeutic effects, innovates the traditional phototherapy model, and brings a better treatment experience.

[0059] Based on Example 1, this example describes the implementation process of this application in detail, specifically:

[0060] When a patient's face was treated with the phototherapy technology of this application, the initial examination revealed that the patient had obvious inflammatory symptoms on the face, with a large area of redness on the skin and multiple visible papules and other inflammatory manifestations, such as Figure 2 As shown;

[0061] like Figure 3 As shown, as the first treatment progressed, the multi-source biofeedback data monitored in real time in the technical solution, such as skin temperature, blood flow velocity, and skin electrical response information, were deeply analyzed through the AI algorithm model to accurately adjust the phototherapy parameters. At this time, it can be observed that the area of redness on the patient's face has decreased, and the redness and swelling of some papules have also been alleviated; by the second treatment, based on the model's continuous analysis of the trend of changes in physiological status, the phototherapy parameters were further optimized, the inflammation area was significantly reduced, the overall redness of the skin was greatly reduced, and the number of papules was also significantly reduced. This shows that the technical solution can dynamically and accurately adjust the phototherapy parameters according to the patient's physiological state at each treatment, thereby gradually alleviating the symptoms of skin inflammation and achieving effective control and improvement of skin inflammation.

[0062] At the beginning of the treatment, the patient's facial skin was not only obviously inflamed, but also had a dark and uneven color and a rough texture. This technical solution relies on the comprehensive capture of skin physiological information by the biosensor group and the efficient processing of this information by the AI algorithm model, so that the parameters during the phototherapy process are continuously adapted to the real-time status of the patient's skin. During the third treatment, compared with before, the skin color had changed significantly, the dark areas gradually became brighter, and the overall color tended to be uniform. This is because the appropriate phototherapy parameters promoted the metabolism of the skin and improved local blood circulation. At the same time, the texture of the skin was also improved and was no longer as rough as in the early stages. This is due to the precise adjustment of the phototherapy parameters, which can stimulate the activity of skin cells and enhance the connection and repair between cells. By the fourth treatment, the skin color was brighter and more even, and the texture was more delicate and smoother, which fully demonstrated the excellent effect of this technical solution in improving skin color and texture. Through continuous and precise adjustment of phototherapy parameters, a good repair environment was created for the skin.

[0063] like Figure 4 As shown in the figure, during the third treatment, the wavelength, power and irradiation time were accurately and dynamically adjusted according to the patient's physiological state; the initial wavelength was set to 410nm. Considering the initial state of the patient's skin inflammation, this wavelength has a certain inhibitory effect on the microorganisms in the inflammatory area; the power was set to 32mW / cm 2 , to provide moderate energy to activate the phototherapy effect; the irradiation time is 2 minutes to avoid excessive stimulation to the skin; based on the data collected by the biosensor group and analyzed by the AI algorithm model, it was found that the local blood flow velocity of the patient's skin increased, the skin temperature rose slightly, and the pain threshold decreased slightly. Based on these physiological changes, in order to avoid excessive stimulation of the skin, the wavelength was adjusted to 405nm to reduce the thermal effect on the skin; the power was reduced to 30mW / cm 2 To prevent excessive energy; the irradiation time was shortened to 2 minutes to give the skin appropriate recovery time; the model analysis data showed that the blood flow velocity in the inflamed area of the patient's skin tended to be stable and the temperature was close to the normal range, but some inflammatory cells were still active. Therefore, the wavelength was adjusted to 415nm to enhance the effect on inflammatory cells; the power was increased to 34mW / cm 2 To more effectively promote the absorption of inflammation; the irradiation time is extended to 4 minutes to ensure that the phototherapy has a full effect on the inflamed area; by this time, the patient's skin inflammation has been significantly reduced and the overall physiological state is more stable. To consolidate the treatment effect, the wavelength is fine-tuned to 420nm to further promote skin cell repair; the power is maintained at 34mW / cm 2 , maintain stable treatment energy; adjust the irradiation time back to 6 minutes, balance the treatment effect and skin tolerance; maintain gentle stimulation to skin cells, adjust the wavelength to 418nm, and slightly reduce the power to 30mW / cm 2, to avoid overtreatment, the irradiation time was shortened to 4 minutes for the final consolidation treatment.

[0064] This embodiment describes in detail how to collect physiological data in real time with the help of sensors, and after analysis by AI models, accurately and dynamically adjust the wavelength, power and irradiation time. Each treatment adapts the parameters according to the physiological state, from inhibiting microorganisms, regulating energy to promoting cell repair, gradually reducing inflammation, optimizing skin condition, reducing the risk of recurrence, achieving efficient treatment, ensuring skin safety and comfort, and demonstrating the advantages of personalized and intelligent excellent phototherapy; through the continuous dynamic adjustment of wavelength, power and irradiation time in these four treatments, the advantage of this phototherapy technology solution in accurately regulating phototherapy parameters according to the patient's physiological state is fully demonstrated, achieving effective treatment of skin inflammation, while ensuring the safety and comfort of the skin, reducing the risk of inflammation recurrence, and making the phototherapy process more scientific and efficient.

[0065] Based on Example 1, this example describes in detail the comparative experiment between the present application and the prior art, specifically:

[0066] 150 patients with skin inflammation were selected and randomly divided into three groups, with 50 patients in each group; Group A used the phototherapy parameter adjustment method of the present application, Group B used the empirical fixed parameter phototherapy method, and Group C used the physiological indicator threshold-triggered phototherapy parameter adjustment method;

[0067] Each patient is equipped with a phototherapy device that integrates a thermal sensor, a blood flow sensor, and a skin galvanic response sensor. The thermal sensor has an accuracy of 0.1°C and can monitor skin temperature changes in real time. The blood flow sensor can accurately measure local blood flow velocity with a resolution of 0.01cm / s. The skin galvanic response sensor can sense skin electrical activity with an accuracy of 0.001μS. The phototherapy device has multiple wavelength outputs (380-760nm) and a power adjustment range of 10-100mW / cm 2 , the irradiation time can be set within 1-60 minutes.

[0068] During phototherapy, the sensor continuously collected data on the patient's skin temperature, blood flow velocity, and pain threshold. During a single 5-minute data collection cycle, the skin temperature of Patient A in Group A was 32.56°C, 32.63°C, 32.71°C, 32.68°C, and 32.74°C; the blood flow velocity was 2.13 cm / s, 2.15 cm / s, 2.17 cm / s, 2.16 cm / s, and 2.18 cm / s; and the galvanic skin response was 0.032 μS, 0.033 μS, 0.034 μS, 0.033 μS, and 0.035 μS. These raw data were processed using a composite filtering model to remove noise and outliers.

[0069] The preprocessed data is fed into an AI algorithm model constructed by integrating a convolutional neural network and a long short-term memory network. The model performs modal separation on the multi-source data, analyzing the characteristics of skin temperature, blood flow, and pain threshold data. Taking skin temperature data as an example, the model extracts local spatial features through a convolutional neural network, then analyzes time series features through a long short-term memory network to predict trends in human physiological states.

[0070] Group A compared the prediction results of the AI algorithm model with the preset phototherapy effect evaluation indicators, used the error back propagation algorithm to calculate the parameter deviation, and dynamically adjusted the phototherapy parameters through the reinforcement learning algorithm. At a certain moment, the model predicted that the current treatment effect did not meet the expectations, and the wavelength was adjusted to 450.5nm and the power to 35.5mW / cm 2 , the irradiation time was extended to 25.5 minutes; Group B used a fixed wavelength (420nm), power (30mW / cm 2 ) and irradiation time (20 minutes) for phototherapy; Group C adjusted the parameters based on the single indicator of skin temperature. If the temperature increased by 0.5℃, the power was reduced by 2mW / cm 2 .

[0071] During the 15-day phototherapy, patients were evaluated every 24 hours, including measurement of the skin inflammation area, as shown in Table 1 , and inflammation severity score, as shown in Table 2 (0-10 points, with higher scores indicating more severe inflammation);

[0072] Table 1 Changes in skin inflammation area

[0073]

[0074] from Figure 5 It can be seen that the downward trend of inflammation area in Group A is the most obvious, indicating that the phototherapy parameter adjustment method of the present application can more effectively promote the disappearance of inflammation.

[0075] Table 2 Inflammation degree score

[0076]

[0077] As can be seen from the above table, the inflammation score of Group A decreased the fastest, and the effect of reducing the inflammation level was significant.

[0078] Through a group comparison experiment on 150 patients with skin inflammation, the effectiveness of the AI-based real-time feedback and adaptive phototherapy parameter adjustment method was fully verified. Compared with the traditional fixed parameter phototherapy method and the existing simple parameter adjustment phototherapy method, the method of this application performs better in promoting inflammation regression, reducing the degree of inflammation and improving patient comfort.

[0079] This embodiment describes in detail the parameter adjustment experiment of adaptive phototherapy of this application and traditional and existing technologies. The above data show that this application can accurately adjust phototherapy parameters according to the patient's real-time physiological state, realize personalized and efficient phototherapy, and provide new direction and strong support for the development of phototherapy technology.

[0080] The above are only preferred embodiments of the present invention, which do not limit the scope of protection of the present invention. For those skilled in the art, the present invention can be modified and varied in various ways. Any changes, modifications, replacements, integrations and parameter changes to these embodiments through conventional substitutions or that can achieve the same functions without departing from the principles and spirit of the present invention fall within the scope of protection of the present invention.

Claims

1. A method for real-time feedback and adaptive phototherapy parameter adjustment based on AI, characterized in that: include: S1. Deploy a biosensor group to collect multi-source data of patients during phototherapy and form a multi-source biofeedback dataset after preprocessing; S2. Build an AI algorithm model, input multi-source biofeedback data sets into the AI algorithm model, and output the changing trend of the human body's physiological state; S3. Preset the expected treatment goals for phototherapy. Combined with the mechanisms of action of different phototherapy wavelengths, powers, and irradiation times on human tissues, establish a phototherapy effect evaluation index system. Compare the trend of changes in human physiological status obtained through AI algorithm model analysis with the preset phototherapy effect evaluation indexes. Calculate the deviation between the current phototherapy parameters and the parameters required to achieve the optimal treatment effect through the error back propagation algorithm. S4. Dynamically adjust the phototherapy parameters using a reinforcement learning algorithm based on the deviation value to determine the parameter combination for the next stage of phototherapy; S5. Send the optimized phototherapy parameters to the phototherapy device, repeat the operation at the set time interval, collect feedback data in real time to optimize the phototherapy parameters, and perform adaptive dynamic adjustment.

2. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 1, characterized in that: The biosensor group includes a thermistor, a blood flow sensor and a skin galvanic response sensor; the thermistor is used to monitor changes in skin temperature; the blood flow sensor monitors local blood flow velocity and changes; the skin galvanic response sensor is used to sense skin electrical activity to reflect changes in pain thresholds, and preprocesses the acquired raw sensor data values to form a multi-source biofeedback dataset.

3. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 2, characterized in that: The preprocessing of the collected raw data is specifically as follows: The raw data of skin temperature, blood flow, and skin electrical response type are obtained and filtered through the constructed composite filter model; the raw data is formed into a raw data sequence T(n), and wavelet decomposition is performed to decompose it into sub-band coefficients W of different frequencies j (k), where j is the number of layers of wavelet decomposition, k is the index of the subband coefficient, and the threshold T is set. th , perform threshold processing on the high-frequency sub-band coefficients to remove the high-frequency noise components caused by sensor noise, and obtain the processed high-frequency sub-band coefficients W′ j (k), the formula is Through wavelet reconstruction, the target data sequence T′(n) after preliminary denoising is obtained, and T′(n) is filtered using Kalman filtering. Through prediction and updating, a multi-source biofeedback data set is formed.

4. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 1, characterized in that: The AI algorithm model in S2 is constructed by the fusion of convolutional neural networks and long short-term memory networks. By performing modal separation on the preprocessed multi-source biofeedback data set, skin temperature, blood flow data, and skin electrical response data reflecting the pain threshold are mapped to different subspaces respectively; in each subspace, the convolutional neural network is used to extract local spatial features, and the local spatial features extracted by the convolutional neural network are input into the long short-term memory network. The memory unit and gating mechanism are used to analyze the dependency between features at different times, learn the long-term trends and short-term fluctuations in the time series, and integrate the time series characteristics of each subspace to output the trend of changes in the human physiological state.

5. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 4, characterized in that: In each subspace, when using convolutional neural networks to extract local spatial features, for the input subspace data X, a convolution operation is performed at the lth convolution layer, and the formula is: in Represents the value of the output feature map of the lth convolutional layer at the (i, j) position, is the weight of the l-th layer convolution kernel at position (a, b), b l is the bias term, p and q are the convolution kernel sizes; the convolution kernel weight is dynamically adjusted according to the characteristics of different subspace data, and the formula is: Where σ is the adaptive adjustment value of the fluctuation range, and the importance score of each position feature is calculated through the attention mechanism. The formula is: where ω l It is a learnable parameter that weights the convolution output according to the score to highlight key local spatial features.

6. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 4, characterized in that: In each subspace, the local spatial features extracted by the convolutional neural network are input into the long short-term memory network in time series order. Input gate i of the long short-term memory network t Control the current input features The part that needs to be added to the memory unit, the forget gate f t Determine the memory unit C at the previous moment t-1 The information that needs to be retained in the output gate o t The information in the control memory unit is used to generate the hidden state h at the current moment t , the formula is: h t =o t ⊙tanh(C t ); For the time series features of multiple subspaces, the hidden state H is integrated by weighted fusion t , the formula is: k is the number of subspaces, α s The weight of the subspace importance is dynamically adjusted by H t The analysis outputs the changing trend of human physiological state.

7. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 1 or 4, characterized in that: In S3, the human physiological state change trend data T obtained by the AI algorithm model analysis is td Compared with the preset phototherapy effect evaluation index data T tr Compare and calculate the degree of difference D. The formula is: Where n is the dimension of the evaluation index in the subspace, w i Dynamically assign weights to the importance of the i-th dimension in the evaluation of phototherapy effects. The formula is: ρ i is the correlation coefficient between the features in the subspace and the phototherapy effect. The difference degree D is used to measure the gap between the current physiological state change trend and the ideal phototherapy effect. The error back propagation algorithm is used to calculate the deviation between the current phototherapy parameters and the parameters required to achieve the best treatment effect. The formula is: in is the kth phototherapy parameter Through back propagation calculation, the deviation value corresponding to each phototherapy parameter is obtained.

8. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 1, characterized in that: In S4, the phototherapy parameters are dynamically adjusted using a reinforcement learning algorithm based on the deviation values corresponding to each phototherapy parameter. By obtaining the penetration depth values of different wavelengths of light on different tissues, the changes in power and irradiation time, multiple sets of constraints are set, with maximizing the treatment effect and ensuring the safety and comfort of the patient as the reward function. The optimal solution is searched within the parameter space that meets the constraints to determine the wavelength combination, power and irradiation time for the next stage of phototherapy.

9. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 8, characterized in that: The reward function R is used to maximize the treatment effect and ensure the safety and comfort of the patient, and guides the search for the optimal solution. The formula is: Among them E t is a comprehensive treatment index, S t is the skin tolerance value, C s is the subjective comfort value, δ t is the time attenuation factor, P is the current phototherapy parameter deviation value, P r is the comparison parameter deviation value, P max and P min are the maximum deviation value and the minimum deviation value of the parameter deviation value respectively. The optimal solution is searched by the particle swarm optimization algorithm in the parameter space that meets the constraints. Each particle represents a set of phototherapy parameters. In each iteration, the reward function value R corresponding to each particle is calculated, and the phototherapy parameter combination that maximizes the reward function R is gradually searched to achieve the maximum treatment effect.

10. The AI-based real-time feedback and adaptive phototherapy parameter adjustment method according to claim 1, characterized in that: In S5, the optimized phototherapy parameters are sent to the phototherapy equipment. By adjusting the light-emitting components of the phototherapy equipment, the equipment outputs phototherapy according to the new wavelength, power and irradiation time. The adjusted phototherapy parameter information is fed back to the patient terminal and medical staff monitoring platform, making it convenient for patients and medical staff to understand the progress of phototherapy in real time.

Images