Laser hair removal method based on multi-parameter feedback

Abstract

The invention relates to the technical field of medical cosmetology, discloses a laser hair removal method based on multi-parameter feedback, and aims to solve the problems of unstable curative effect and insufficient safety caused by single parameter of traditional laser hair removal. The method comprises the following steps: synchronously acquiring multi-dimensional physiological parameters such as skin type, hair follicle depth and epidermis temperature through multispectral imaging, high-frequency ultrasound and temperature sensing; constructing a personalized treatment parameter model based on a deep neural network, and generating initial energy density, pulse width and cooling parameters; laser output is regulated and controlled in real time by using a fuzzy logic controller in the treatment process, and it is ensured that thermal injury accurately acts on the hair papilla; performing adaptive optimization on the treatment strategy in combination with reinforcement learning, and continuously updating the model through a closed-loop learning mechanism; meanwhile, a treatment history database is established, curative effect prediction and parameter backtracking are supported, precision, individuation and high safety of laser depilation are achieved, depilation efficiency is remarkably improved, and the risk of skin adverse reaction is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical cosmetology, in particular to a laser hair removal method based on multi-parameter feedback. BACKGROUND

[0002] With the popularization of laser cosmetic technology, laser hair removal has become the mainstream hair removal method due to its high efficiency and durability. The basic principle is to use laser of specific wavelength to penetrate the epidermis, and the black pigment in the hair follicle selectively absorbs and converts it into heat energy, thereby destroying the hair papilla and inhibiting hair regeneration. However, existing laser hair removal equipment mostly adopts fixed parameter mode, that is, a unified energy density, pulse width and cooling intensity are preset in a single treatment, which is difficult to adapt to the significant differences in skin and hair characteristics between individuals and different parts of the same human body.

[0003] Among them, the effect and safety of laser hair removal are highly dependent on the accurate adaptation of multi-dimensional physiological parameters such as skin type, hair color, hair follicle depth and local temperature. The existing technology usually only makes rough classification according to Fitzpatrick skin type, lacking dynamic perception and response mechanism for real-time physiological feedback. This leads to the risk of insufficient energy (ineffective hair removal) or excessive energy (burning, pigmentation) in complex scenarios such as dark skin or light hair caused by fixed parameters.

[0004] In the prior art, some high-end devices have introduced contact temperature sensors or simple skin color recognition modules, but their feedback dimensions are single and response is lagging, and they cannot achieve coordinated control under multi-parameter coupling. At the same time, the cumulative effect of hair follicle thermal damage and the dynamic changes of heat diffusion of surrounding tissues during laser action are not included in the control logic, resulting in a lack of individualization and precision in treatment. In addition, the existing system lacks closed-loop learning ability for historical data of the treatment area, and cannot dynamically optimize subsequent parameters according to the individual's response to previous treatment. Therefore, there is an urgent need for a laser hair removal method that can integrate multi-source physiological parameters and achieve real-time feedback and adaptive control, in order to balance hair removal efficiency and skin safety. SUMMARY

[0005] The purpose of the present application is to provide a laser hair removal method based on multi-parameter feedback, which can effectively solve the problems in the background art.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is: A laser hair removal method based on multi-parameter feedback, comprising the following specific steps: Step 1: Obtain multi-dimensional physiological parameters of the treatment area, collect skin surface reflectance spectrum data through multi-spectral imaging sensors, simultaneously use high-frequency ultrasonic probes to obtain hair follicle depth and structure information, and combine contact temperature sensors to monitor real-time epidermal temperature changes; Step 2: Construct a personalized treatment parameter model, based on the multi-dimensional physiological parameters, use deep neural network algorithm to analyze the non-linear relationship between skin type, hair pigment content, hair follicle morphological characteristics and thermal diffusion characteristics, generate recommended values of initial energy density , pulse width and dynamic cooling parameters; wherein the initial energy density is estimated by the following formula:

[0007] wherein is the concentration of hair melanin, is the absorption coefficient at the laser wavelength , is the depth of the hair follicle, is the initial temperature of the epidermis, and is the empirical coefficient.

[0008] Step 3: Implement real-time multi-parameter feedback control, continuously monitor the epidermal temperature, heat accumulation in the hair follicle area and skin impedance changes during laser emission, dynamically adjust the laser output energy and pulse duration through a fuzzy logic controller to ensure that the heat damage accurately acts on the hair papilla and avoids excessive heating of the surrounding tissue. The output correction amount of the fuzzy logic controller is calculated by the following formula:

[0009] wherein is the deviation of the epidermal temperature from the safety threshold (45°C), is the skin impedance, , , is the adjustable weight coefficient. Step 4: Perform adaptive treatment strategy optimization, update the personalized treatment parameter model using reinforcement learning algorithm based on the physiological response data collected during a single treatment, optimize the parameter combination of the subsequent pulse sequence; Step 5: Establish a treatment history database and closed-loop learning mechanism, record the parameter settings, real-time feedback data and clinical effect evaluation of each treatment, and continuously improve the accuracy and safety of parameter recommendations through periodic model retraining.

[0010] Preferably, the multi-spectral imaging sensor in step 1 contains 6 light source modules of different wavelengths, covering a wavelength range of 500 nanometers to 1000 nanometers, with a spectral resolution of 5 nanometers, capable of simultaneously obtaining the distribution information of hemoglobin, melanin and water content.

[0011] ​​Preferably, the high-frequency ultrasound probe has a center frequency of 20 MHz and an axial resolution of 50 microns, which can clearly distinguish the hair papilla, hair sheath and surrounding connective tissue structure of the hair follicle, and the measurement depth can reach 4 millimeters.

[0012] Preferably, the contact temperature sensor uses a thin-film thermocouple array, which contains 16 independent temperature measurement points, with a temperature measurement accuracy of plus or minus 0.1 degrees Celsius, a response time of less than 100 milliseconds, and a real-time generation of a two-dimensional temperature distribution map of the treatment area.

[0013] Preferably, the deep neural network algorithm in step 2 uses an 8-layer fully connected structure, with 512 hidden layer nodes, a rectified linear unit activation function, an input feature dimension of 28, and an output of a 3-dimensional treatment parameter vector.

[0014] Preferably, the training data of the personalized treatment parameter model comes from more than 10,000 clinical treatment records, covering 6 Fitzpatrick skin types and 4 hair color grades, and the consistency of the model prediction and expert recommended parameters reaches 95%.

[0015] Preferably, the input variables of the fuzzy logic controller in step 3 include the epidermal temperature deviation, the hair follicle temperature gradient, and the skin impedance change rate, and the output variables are the laser energy correction coefficient and the pulse width adjustment amount, with a control period of 50 milliseconds.

[0016] Preferably, in the real-time multi-parameter feedback regulation process, the epidermal temperature safety threshold is set to 45 degrees Celsius, and when the temperature of any temperature measurement point exceeds this threshold, the system automatically triggers the strong cooling mechanism, and the cooling duration can be adjusted within the range of 100 milliseconds to 500 milliseconds.

[0017] Preferably, the dynamic cooling system adopts a dual-mode design, including contact semiconductor cooling and non-contact cold spray, and the cooling intensity is regulated according to real-time temperature feedback, with a maximum cooling power of 200 watts.

[0018] Preferably, the reinforcement learning algorithm in step 4 uses the deep deterministic policy gradient method, the state space is the joint features of multi-dimensional physiological parameters and real-time feedback data, the action space is the continuous adjustment amount of the treatment parameters, and the reward function considers the hair removal efficiency and safety indicators.

[0019] Preferably, the adaptive treatment strategy optimization process updates the strategy network parameters after each pulse emission, uses an experience replay buffer to store the last 1000 treatment decision data, and ensures the stability and convergence of the learning process.

[0020] Preferably, the treatment history database in step 5 is stored in a distributed architecture, capable of processing more than 1000 treatment records per day, each record containing 15 categories of parameter and feedback data fields, with a data retention period of 10 years.

[0021] Preferably, the closed-loop learning mechanism includes incremental model training performed once a week, fine-tuning deep neural network parameters with new data accumulated in the past 7 days, while preserving historical knowledge to prevent catastrophic forgetting.

[0022] Preferably, the periodic model retraining uses a flexible weight consolidation algorithm to calculate the importance weight of network parameters for historical tasks, imposing regularization constraints on important parameters when training new data, maintaining the model's adaptability to diverse skin types.

[0023] Preferably, it also includes a treatment effect prediction module, based on treatment history data and real-time feedback information, using a long short-term memory neural network to predict the hair regrowth probability of a single treatment, with a prediction time span covering 1 month to 6 months.

[0024] Preferably, the input features of the treatment effect prediction module include the degree of hair follicle destruction, the intensity of surrounding tissue inflammatory response, and individual metabolic characteristics, and the output is the percentage of hair density reduction at different time points, with a prediction accuracy of more than 90%.

[0025] Preferably, the method is integrated into an intelligent laser hair removal device, displaying real-time treatment parameters, physiological feedback data and safety status indicators through a graphical user interface, supporting therapists to confirm or fine-tune according to system recommendations.

[0026] Preferably, the intelligent laser hair removal device is equipped with a safety interlock mechanism that automatically suspends laser emission and sends visual and audible alarms when sensor failure or parameter abnormalities are detected, ensuring the safety of the treatment process.

[0027] Preferably, the multi-parameter feedback laser hair removal method supports specific parameter presets for different body parts, including the face, underarms, limbs, and bikini areas, each containing at least 3 sets of differentiated parameter templates.

[0028] Preferably, the differentiated parameter templates are generated based on large-scale clinical data analysis, taking into account the differences in skin thickness, hair follicle density and nerve sensitivity of different parts, ensuring precise matching of treatment parameters with local anatomical features.

[0029] Preferably, it also includes a post-treatment skin state monitoring function, continuously observing skin erythema and edema reactions after treatment, quantitatively evaluating tissue recovery through multispectral imaging, and providing data support for subsequent treatment intervals.

[0030] Compared with the prior art, the beneficial effects achieved by the present invention are: By integrating multispectral imaging, high-frequency ultrasound, and temperature sensing technologies, this method achieves comprehensive acquisition and precise analysis of multidimensional physiological parameters in the treatment area, overcoming the limitations of traditional methods with their single parameter dimension. A personalized treatment parameter model based on deep neural networks accurately captures the complex relationship between individual physiological characteristics and optimal treatment parameters, significantly improving the targeting of parameter recommendations. A real-time multi-parameter feedback control mechanism achieves dynamic optimization of the treatment process through fuzzy logic control, ensuring that thermal damage is precisely applied to the target tissue while maximizing the protection of surrounding healthy tissue. Adaptive treatment strategy optimization and closed-loop learning mechanisms enable the system to continuously learn from historical treatment data, constantly improving the quality of parameter recommendations. The treatment effect prediction function provides a scientific basis for personalized adjustments to the treatment plan. Overall, this method significantly reduces the risk of adverse reactions such as skin burns and pigmentation while improving hair removal efficiency, achieving precise, personalized, and safe laser hair removal treatment. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall technical solution architecture of the method of the present invention; Figure 2 This is a schematic diagram of the core principle framework of the personalized treatment parameter model based on deep neural networks in this invention; Figure 3 This is a flowchart illustrating the logic flow of real-time multi-parameter feedback regulation and fuzzy logic control in this invention. Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow of the adaptive treatment strategy optimization and closed-loop learning mechanism in this invention; Figure 5 This is a logical flowchart of the treatment effect prediction module and the historical database working together in this invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] Example 1 To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0034] Currently, with the increasing popularity of laser cosmetic technology, laser hair removal has become the mainstream hair removal method due to its high efficiency and long-lasting effects. Its basic principle is to use a specific wavelength of laser light to penetrate the epidermis, where it is selectively absorbed by the melanin in the hair follicle and converted into heat energy, thereby destroying the hair papilla and inhibiting hair regrowth. However, most existing laser hair removal devices use fixed parameter modes, meaning that the energy density, pulse width, and cooling intensity are preset uniformly in a single treatment. This makes it difficult to adapt to the significant differences in skin and hair characteristics between individuals and in different parts of the same body. The effectiveness and safety of laser hair removal highly depend on the precise adaptation to multi-dimensional physiological parameters such as skin type, hair color, hair follicle depth, and local temperature. Current technologies typically only provide a rough classification based on the Fitzpatrick skin type, lacking a dynamic perception and response mechanism for real-time physiological feedback. This leads to risks of insufficient energy (ineffective hair removal) or energy overload (burns, pigmentation) in complex scenarios such as dark skin or light hair. While some high-end devices in existing technologies incorporate contact temperature sensors or simple skin color recognition modules, their feedback dimensions are singular and their response is lagging, failing to achieve coordinated control under multi-parameter coupling. Furthermore, the cumulative effect of thermal damage to hair follicles during laser treatment and the dynamic changes in thermal diffusion of surrounding tissues are not incorporated into the control logic, resulting in a lack of personalization and precision in treatment. In addition, existing systems lack the ability to learn from historical data of the treatment area in a closed loop, failing to dynamically optimize subsequent parameters based on individual responses to previous treatments. To address these technical problems, this invention proposes a laser hair removal method based on multi-parameter feedback. By integrating multispectral imaging, high-frequency ultrasound, and temperature sensing technologies, a personalized treatment parameter model is constructed, and real-time multi-parameter feedback control and adaptive strategy optimization are implemented to achieve precise, personalized, and safe laser hair removal treatment. This method is applied to a laser hair removal approach based on multi-parameter feedback.

[0035] refer to Figure 1The overall technical architecture of the laser hair removal method based on multi-parameter feedback proposed in this invention includes a multi-dimensional physiological parameter acquisition unit, a personalized treatment parameter model construction unit, a real-time multi-parameter feedback control unit, an adaptive treatment strategy optimization unit, and a treatment history database and closed-loop learning unit. Each unit achieves tight coupling and efficient collaboration through a high-speed data bus and event-driven mechanism. In the above-mentioned laser hair removal method based on multi-parameter feedback, step (1) acquires multi-dimensional physiological parameters of the treatment area. The multi-spectral imaging sensor collects the skin surface reflectance spectrum data, while a high-frequency ultrasonic probe is used to acquire hair follicle depth and structural information, and a contact temperature sensor is used to monitor epidermal temperature changes in real time. Specifically, the multi-spectral imaging sensor contains six light source modules with different wavelengths, covering a wavelength range of 500 nm to 1000 nm, and a spectral resolution of 5 nm. It can simultaneously acquire the distribution information of hemoglobin, melanin, and water content. The sensor adopts a ring LED array layout to ensure uniform illumination, and the acquisition frame rate can reach 30 frames per second. After the data is processed by the spectral inversion algorithm, a two-dimensional concentration distribution map of various skin components is output. The high-frequency ultrasound probe has a center frequency of 20 MHz and an axial resolution of 50 micrometers, enabling clear differentiation of the hair follicle's dermal papilla, hair sheath, and surrounding connective tissue structure. The measurement depth can reach 4 mm. The probe is integrated into the front end of the laser handpiece and coaxially aligned with the laser emission window, ensuring complete overlap between the ultrasound scanning area and the laser action area. After bandpass filtering and envelope detection, the ultrasound echo signal is used to reconstruct a B-mode ultrasound image of the hair follicle using a deconvolution algorithm. Edge detection and morphological segmentation techniques are then used to accurately extract structural parameters such as hair follicle depth (dh) and diameter. The contact temperature sensor employs a thin-film thermocouple array containing 16 independent temperature measurement points. It has a temperature measurement accuracy of ±0.1 degrees Celsius and a response time of less than 100 milliseconds, generating a real-time two-dimensional temperature distribution map of the treatment area. The sensor array is embedded in the cooling contact surface of the laser handpiece, directly contacting the skin. It continuously collects temperature data before, during, and after laser emission, providing a foundation for subsequent thermal safety control. The data from the three sensors are strictly synchronized in time, and the sampling clock is provided by the same main control unit, ensuring the precise alignment of multi-source data in the spatiotemporal dimension, laying the foundation for building a comprehensive physiological profile.

[0036] In the aforementioned laser hair removal method based on multi-parameter feedback, step (2) involves constructing a personalized treatment parameter model. Based on the multi-dimensional physiological parameters, a deep neural network algorithm is used to analyze the nonlinear relationship between skin type, hair pigment content, hair follicle morphology, and thermal diffusion characteristics, generating recommended values ​​for initial energy density E0, pulse width τ, and dynamic cooling parameters. Specifically, the deep neural network algorithm employs an 8-layer fully connected structure with 512 hidden layer nodes, a modified linear unit activation function, an input feature dimension of 28, and an output of a 3-dimensional treatment parameter vector. Input features include, but are not limited to: melanin concentration Cm, hemoglobin concentration, skin moisture content, hair follicle depth dh, hair follicle diameter, initial epidermal temperature Tepi, skin impedance baseline value, Fitzpatrick skin type encoding, hair color grade encoding, and treatment site encoding. The training data for this model comes from over 10,000 clinical treatment records, covering 6 Fitzpatrick skin types and 4 hair color grades. The model's predictions show a 95% consistency with expert-recommended parameters. The core of the model lies in estimating the initial energy density E0, which physically means providing the minimum effective energy sufficient to destroy the hair papilla without damaging the surrounding tissue. This estimation process is precisely quantified by the following formula:

[0037] in, and These are empirical coefficients, obtained through regression fitting of large-scale clinical data. The hair melanin concentration was derived from multispectral imaging data. wavelength The absorption coefficient is calculated based on the laser's preset wavelength (e.g., 810 nm) from a pre-stored optical parameter database. dh represents the hair follicle depth, measured by a high-frequency ultrasonic probe. Tepi is the initial epidermal temperature, an average value collected by a contact temperature sensor array within one second before laser emission. This formula comprehensively considers the target tissue's light absorption capacity, heat conduction path length, and initial thermal state, providing a solid physical basis for initial parameter setting. (Reference) Figure 2 The core principle framework of the personalized treatment parameter model based on deep neural networks is shown. The model first normalizes and performs feature engineering on the original multidimensional physiological parameters, and then inputs them into the deep neural network for nonlinear mapping. Finally, it outputs recommended values ​​for the initial energy density E0, pulse width τ, and dynamic cooling parameters, which serve as the benchmark for subsequent real-time regulation.

[0038] In the above-mentioned laser hair removal method based on multi-parameter feedback, step (3) implements real-time multi-parameter feedback control, continuously monitoring epidermal temperature, heat accumulation in the hair follicle area, and changes in skin impedance during laser emission. The laser output energy and pulse duration are dynamically adjusted through a fuzzy logic controller to ensure that thermal damage is precisely applied to the hair papilla while avoiding excessive heating of surrounding tissues. Specifically, the input variables of the fuzzy logic controller include epidermal temperature deviation, hair follicle temperature gradient, and skin impedance change rate. The output variables are the laser energy correction coefficient and pulse width adjustment. The control cycle is 50 milliseconds, meaning parameter correction is performed every 50 milliseconds. The epidermal temperature deviation eT is defined as the difference between the real-time temperature of any measurement point and the safety threshold of 45 degrees Celsius. When eT is greater than 0, it indicates an overheating risk. The hair follicle temperature gradient is indirectly estimated using a high-frequency ultrasound thermal imaging mode, reflecting the rate of heat accumulation inside the hair follicle. Skin impedance R is measured using a four-electrode method integrated into the handpiece, and its rate of change dR / dt reflects the change in cell membrane permeability caused by thermal damage. The fuzzy logic controller fuzzifies these continuous input variables into linguistic variables (such as "negative large", "zero", "positive small", etc.), performs inference based on a preset fuzzy rule base, and finally defuzzifies to obtain the precise output correction amount ΔE. The calculation process of this correction amount is precisely described by the following formula:

[0039] in This refers to the deviation of the skin temperature from the safe threshold (45℃). For skin impedance, , , These are adjustable weighting coefficients, which can be preset according to the treatment area and skin type. During real-time multi-parameter feedback control, a safe threshold for epidermal temperature is set at 45 degrees Celsius. When the temperature at any measuring point exceeds this threshold, the system automatically triggers a strong cooling mechanism, with the cooling duration adjustable from 100 milliseconds to 500 milliseconds. The dynamic cooling system employs a dual-mode design, including contact semiconductor cooling and non-contact cold spray. The cooling intensity is graded and adjusted based on real-time temperature feedback, with a maximum cooling power of 200 watts. (Reference) Figure 3 The system, as shown in the logic flow framework of real-time multi-parameter feedback regulation and fuzzy logic control, performs closed-loop control of "acquisition-calculation-correction-output" in a 50-millisecond cycle during laser pulse emission, ensuring that the treatment process is always in a safe and efficient dynamic equilibrium state.

[0040] In the above-mentioned laser hair removal method based on multi-parameter feedback, step (4) involves optimizing the adaptive treatment strategy. Based on the physiological response data collected during a single treatment, the personalized treatment parameter model is updated using a reinforcement learning algorithm to optimize the parameter combination of subsequent pulse sequences. Specifically, the reinforcement learning algorithm adopts a deep deterministic policy gradient method. The state space is the joint feature of multi-dimensional physiological parameters and real-time feedback data, the action space is the continuous adjustment amount of treatment parameters, and the reward function comprehensively considers hair removal efficiency and safety indicators. The hair removal efficiency indicator can be quantified by the thermal damage integral of the hair follicle area, while the safety indicator is jointly determined by the highest epidermal temperature and the degree of abnormal impedance change. The adaptive treatment strategy optimization process updates the strategy network parameters after each pulse emission and uses an experience replay buffer to store the treatment decision data of the most recent 1000 times to ensure the stability and convergence of the learning process. For example, when irradiating the same treatment area multiple times with pulses, the system will dynamically adjust the energy and pulse width of subsequent pulses based on the actual thermal response of the previous pulses (such as the rate of temperature rise and the magnitude of impedance drop) to avoid the risk of overheating caused by the accumulation of tissue heat and to achieve fine arrangement of pulse sequences.

[0041] In the aforementioned laser hair removal method based on multi-parameter feedback, step (5) establishes a treatment history database and a closed-loop learning mechanism to record the parameter settings, real-time feedback data, and clinical effect evaluation for each treatment. The accuracy and safety of parameter recommendations are continuously improved through periodic model retraining. Specifically, the treatment history database uses a distributed architecture for storage, capable of processing over 1000 treatment records daily. Each record contains 15 types of parameter and feedback data fields, with a data retention period of 10 years. Fields include patient ID, treatment date, location, initial physiological parameters, recommended parameters, actual execution parameters, real-time feedback data stream, safety event logs, therapist evaluations, and subsequent follow-up effects. The closed-loop learning mechanism includes weekly incremental model training, using new data accumulated over the past 7 days to fine-tune the deep neural network parameters while retaining historical knowledge to prevent catastrophic forgetting. The periodic model retraining employs an elastic weight consolidation algorithm to calculate the importance weights of network parameters to historical tasks. Regularization constraints are applied to important parameters when training new data to maintain the model's adaptability to diverse skin types. (Reference) Figure 4 The diagram shows the multi-level interaction relationship and data flow of the adaptive treatment strategy optimization and closed-loop learning mechanism. This mechanism forms a complete learning closed loop from real-time optimization of a single treatment to long-term model evolution.

[0042] To further enhance the scientific rigor and forward-looking nature of the treatment, this invention also includes a treatment effect prediction module. Based on historical treatment data and real-time feedback information, it uses a long short-term memory neural network to predict the probability of hair regrowth after a single treatment, with a prediction time span covering 1 to 6 months. The input features of the treatment effect prediction module include the degree of hair follicle damage (estimated by ultrasound thermography), the intensity of the surrounding tissue inflammatory response (quantified by changes in hemoglobin concentration from multispectral imaging), and individual metabolic characteristics (inferred from historical treatment intervals and recovery speed). The output is the percentage reduction in hair density at different time points, with a prediction accuracy exceeding 90%. (Reference) Figure 5 The diagram illustrates the logical workflow of the treatment effect prediction module working in conjunction with the historical database. This module provides data-driven decision support for therapists to develop long-term treatment plans, such as determining the optimal time interval for the next treatment. Furthermore, the method is integrated into a smart laser hair removal device, displaying real-time treatment parameters, physiological feedback data, and safety status indicators through a graphical user interface, allowing therapists to confirm or fine-tune operations based on system recommendations. The smart laser hair removal device is equipped with a safety interlock mechanism that automatically pauses laser emission and issues visual and auditory alarms when sensor malfunctions or abnormal parameters are detected, ensuring the safety of the treatment process. The multi-parameter feedback laser hair removal method supports specific parameter presets for different body parts, including the face, armpits, limbs, and bikini area. Each part includes at least three sets of differentiated parameter templates, generated based on large-scale clinical data analysis, taking into account differences in skin thickness, hair follicle density, and nerve sensitivity in different areas, ensuring precise matching of treatment parameters with local anatomical features. It also includes a post-treatment skin condition monitoring function, continuously observing skin erythema and edema reactions after treatment, and quantitatively assessing tissue recovery through multispectral imaging, providing data support for subsequent treatment intervals.

[0043] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A laser hair removal method based on multi-parameter feedback, characterized in that: The specific steps include the following: Step 1: Obtain multi-dimensional physiological parameters of the treatment area. Collect skin surface reflectance spectrum data through a multispectral imaging sensor, and at the same time use a high-frequency ultrasound probe to obtain hair follicle depth and structural information, and combine it with a contact temperature sensor to monitor epidermal temperature changes in real time. Step 2: Construct a personalized treatment parameter model. Based on the multi-dimensional physiological parameters, use a deep neural network algorithm to analyze the nonlinear relationship between skin type, hair pigment content, hair follicle morphology characteristics and heat diffusion characteristics, and generate recommended values ​​for initial energy density, pulse width and dynamic cooling parameters. Step 3: Implement real-time multi-parameter feedback control, continuously monitor epidermal temperature, heat accumulation in the hair follicle area and changes in skin impedance during laser emission, and dynamically adjust the laser output energy and pulse duration through a fuzzy logic controller to ensure that thermal damage is precisely applied to the hair papilla and avoids excessive heating of surrounding tissues. Step 4: Execute adaptive treatment strategy optimization. Based on the physiological response data collected during a single treatment, update the personalized treatment parameter model using a reinforcement learning algorithm to optimize the parameter combination of subsequent pulse sequences. Step 5: Establish a treatment history database and closed-loop learning mechanism to record the parameter settings, real-time feedback data, and clinical effect evaluation for each treatment. Continuously improve the accuracy and safety of parameter recommendations through periodic model retraining.

2. The laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: The multispectral imaging sensor includes six light source modules with different wavelengths, covering a wavelength range of 500 nm to 1000 nm, and a spectral resolution of 5 nm; the high-frequency ultrasonic probe has a center frequency of 20 MHz, an axial resolution of 50 μm, and a measurement depth of up to 4 mm; the contact temperature sensor uses a thin-film thermocouple array, containing 16 independent temperature measurement points, with a temperature measurement accuracy of ±0.1 degrees Celsius and a response time of less than 100 milliseconds.

3. The laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: The deep neural network algorithm adopts an 8-layer fully connected structure with 512 hidden layer nodes, a modified linear unit activation function, an input feature dimension of 28, and an output of a 3-dimensional treatment parameter vector. The training data of the personalized treatment parameter model comes from more than 10,000 clinical treatment records, covering 6 Fitzpatrick skin types and 4 hair color grades.

4. The laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: The initial energy density From the formula The calculation shows that, among which The concentration of melanin in hair. laser wavelength The absorption coefficient below, For hair follicle depth, The initial temperature of the epidermis. and This is an empirical coefficient.

5. The laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: The input variables of the fuzzy logic controller include epidermal temperature deviation, hair follicle temperature gradient and skin impedance change rate, and the output variables are laser energy correction coefficient and pulse width adjustment amount. The control period is 50 milliseconds. The epidermal temperature safety threshold is set to 45 degrees Celsius. When the temperature of any temperature measurement point exceeds this threshold, the system automatically triggers a strong cooling mechanism.

6. The laser hair removal method based on multi-parameter feedback according to claim 5, characterized in that: The laser energy correction amount From the formula The calculation shows that, among which The deviation between skin temperature and the safety threshold, For skin impedance, , , These are adjustable weighting coefficients.

7. The laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: The dynamic cooling system adopts a dual-mode design, including contact semiconductor cooling and non-contact cold spray. The cooling intensity is adjusted in stages according to real-time temperature feedback, and the maximum cooling power reaches 200 watts.

8. The laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: The reinforcement learning algorithm employs a deep deterministic policy gradient method. The state space consists of the joint features of multi-dimensional physiological parameters and real-time feedback data, the action space consists of the continuous adjustment of treatment parameters, and the reward function comprehensively considers hair removal efficiency and safety indicators. The experience replay buffer stores the treatment decision data of the most recent 1000 times.

9. The laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: The periodic model retraining employs an elastic weight consolidation algorithm to calculate the importance weights of network parameters to historical tasks and apply regularization constraints to important parameters when training on new data. The closed-loop learning mechanism performs incremental training once a week, using new data accumulated over the past 7 days to fine-tune the model parameters.

10. A laser hair removal method based on multi-parameter feedback according to claim 1, characterized in that: It also includes a treatment effect prediction module, which uses a long short-term memory neural network to predict the percentage reduction in hair density within 1 to 6 months based on historical treatment data and real-time feedback information. The input features of the prediction module include the degree of hair follicle damage, the intensity of inflammation in the surrounding tissue, and individual metabolic characteristics, with a prediction accuracy of over 90%.

Images