Near-infrared light controllable irradiator for stellate ganglion stimulation

Abstract

The invention discloses a near-infrared light controllable irradiator for astroganglion stimulation, which comprises a physiological signal acquisition module used for acquiring a near-infrared light sensing signal of a carotid artery area in real time; the signal processing module is used for performing variational mode decomposition on the near-infrared sensing signal and extracting a heart rate component; the autonomic nerve analysis module is used for receiving the heart rate component and calculating the excitement ratio of sympathetic nerves to parasympathetic nerves; a deep learning model is arranged in the control signal generation module, and an optical parameter control signal is generated according to the heart rate cycle characteristics and the excitement ratio; and the light adjusting module is used for receiving the light parameter control signal so as to control the emission of near-infrared light. According to the invention, the output of the infrared light can be adjusted according to the physiological signal of the human body, intelligent adjustment is realized, the near-infrared light stimulation better fits the individual condition, and the pertinence and effectiveness of the stimulation are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical equipment, and in particular to a near-infrared light controllable irradiator for stellate ganglion stimulation. Background Art

[0002] Near-infrared light (810-850 nm) has been used in the fields of neural stimulation and disease treatment due to its strong tissue penetration and photobiomodulation (PBM) effect. However, existing technologies for using near-infrared light to stimulate human tissue, particularly deep tissues like the stellate ganglion, have significant limitations, limiting their development and application: 1. Lack of feedback control: Existing near-infrared light stimulation technologies often rely on physical signals or imaging guidance for feedback control, aiming to achieve physical / optical goals while ignoring real-time feedback from the human body's physiological and emotional state. The inability to implement closed-loop feedback control based on biosignals such as heart rate variability (HRV) and breathing patterns makes it difficult to regulate the user's emotions and autonomic nervous system.

[0003] 2. Single control dimensions and methods: Existing technologies primarily control the physical parameters of light, such as beam pointing, focusing, basic parameter adjustment, and simple intensity adjustment. These limited control dimensions ignore the real-time responses of organisms and lack the ability to dynamically and intelligently adjust light stimulation parameters across multiple dimensions using advanced signal processing and artificial intelligence algorithms, making refined and personalized adjustments impossible.

[0004] 3. Limited application scenarios and poor portability: Many related patents involve complex systems that integrate multiple technologies, are bulky, and cumbersome to operate, making them unsuitable for portability or continuous daily use. Research has primarily focused on the brain, with a lack of non-invasive photobiomodulation studies targeting the stellate ganglion. Some deep-seated stimulation techniques require invasive assistance or biomodulation, increasing risk and reducing universality and ease of use. The market lacks portable, non-invasive modulation devices specifically designed for the stellate ganglion, making them inadequate for daily use. Summary of the Invention

[0005] The present invention aims to provide a controllable near-infrared light illuminator for stellate ganglion stimulation. This device can adjust the infrared light output based on human physiological signals, achieving intelligent regulation and making near-infrared light stimulation more tailored to individual conditions, effectively improving the targetedness and effectiveness of the stimulation.

[0006] The technical solution of the present invention is a near-infrared light controllable illuminator for stellate ganglion stimulation, comprising: Physiological signal acquisition module, used to obtain near-infrared light sensing signals from the carotid artery area in real time; a signal processing module, connected to the physiological signal acquisition module, configured to perform variational mode decomposition on the near-infrared sensing signal and extract a heart rate component; an autonomic nerve analysis module, communicating and interacting with the signal processing module, receiving the heart rate component and calculating the excitability ratio of the sympathetic nerves to the parasympathetic nerves; A control signal generation module is communicatively connected to the autonomic nerve analysis module and the signal processing module; the control signal generation module has a built-in deep learning model, and generates a light parameter control signal according to the heart rate cycle characteristics and the excitability ratio; a light adjustment module connected to the control signal generation module and receiving the light parameter control signal to control the emission of near-infrared light. The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation, wherein the physiological signal acquisition module includes a contact CPPG sensor having a pressure sensitive element on its optical receiving surface. When the sensor pressure reaches a preset threshold, a signal acquisition circuit is activated to acquire signals. The protection monitoring module is connected to the signal processing module, the autonomic nerve analysis module and the control signal generation module.

[0007] The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation, the signal processing module includes: A pre-processing unit, which performs baseline correction and band-pass filtering on the raw near-infrared light sensing signal; The variational mode decomposition unit performs multimodal decomposition on the signal processed by the preprocessing unit and extracts the heart rate component.

[0008] The aforementioned near-infrared light controllable irradiator for stellate ganglion stimulation, the autonomic nerve analysis module uses a double elliptical filter group to separate low-frequency and high-frequency band energy; the double elliptical filter group includes: The first elliptical filter: passband 0.04-0.15Hz, used to extract low-frequency energy components; Second elliptical filter: passband 0.15-0.4Hz, used to extract high-frequency energy components; The autonomic nerve analysis module calculates the energy ratio of the low-frequency energy component to the high-frequency energy component to obtain the excitement ratio.

[0009] The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation, the control signal generating module includes: Dynamic matched filter unit, which uses matched filter to track the heartbeat pulse waveform and generate heart rate cycle characteristics; The neural network inference unit deploys an LSTM neural network, which receives the heart rate cycle characteristics and excitement ratio and outputs PWM modulated light parameter control instructions.

[0010] The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation, wherein the matched filter tracks the heartbeat pulse waveform and generates the heart rate cycle characteristics is as follows: Use the peak detection method to find the first few pulse waves of the heartbeat pulse waveform, average them as the initial template, and construct the impulse response of the matched filter based on the currently selected initial template: : Where: is the impulse response of the matched filter, For the initial template, For time, The duration of the initial template; The current input signal segment Impulse response of the matched filter Perform convolution to get the output : ; The convolution operation enhances the main component of the pulse wave, suppresses noise and non-pulse components, and makes the peak of the pulse wave more prominent; Search for the local maximum within the expected heartbeat interval, which represents the position in the input signal that has the highest match with the template, and then extract the local maximum from the signal segment. In the example, a time point corresponding to the template duration is extracted with the detected peak time point as the center. For the same signal segment, the original template is updated with the extracted signal segment to obtain a new template, thereby adapting to the change of the pulse wave shape and improving the matched filtering effect; Finally, the above steps of template construction, convolution calculation, peak search, and template update are repeated to process the input signal segment by segment or continuously. Each iteration is based on the previous processing result, and the template of the matched filter is continuously adjusted to accurately track the heartbeat pulse waveform, thereby generating accurate heart rate cycle characteristics.

[0011] The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation, the LSTM neural network is deployed by the following steps: a) Train four basic LSTM models corresponding to excitatory, soothing, daily, and intense stimulation modes; b) Using temperature scaling distillation to transfer the knowledge of the four models into a single lightweight network; c) Deployment on embedded devices is achieved through 4-bit quantization compression and the TF-Lite framework.

[0012] The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation, the light adjustment module comprises: A focusing mechanism drives the lens displacement according to the focusing parameter in the light parameter control signal; The near-infrared LED adjusts the near-infrared light output according to the intensity parameter in the light parameter control signal.

[0013] The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation, the protection monitoring module comprises: an abnormality detection unit for monitoring heart rate components and / or arousal ratios in real time; The hierarchical response unit is configured with a three-level response mechanism. When the monitored heart rate component and / or excitement ratio exceeds the first threshold, the stimulation mode switch is triggered; when it exceeds the second threshold, the sound and light alarm is activated; when it exceeds the third threshold, the power is cut off.

[0014] The aforementioned near-infrared light controllable illuminator for stellate ganglion stimulation adopts a wearable neck support structure.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. Achieve closed-loop feedback control of physiological signals: The near-infrared light controllable illuminator of the present invention is equipped with multiple functional modules, including a physiological signal acquisition module, a signal processing module, and an autonomic nerve analysis module. The physiological signal acquisition module acquires near-infrared light sensing signals from the carotid artery region in real time. The signal processing module extracts the heart rate component. The autonomic nerve analysis module calculates the excitability ratio of the sympathetic and parasympathetic nerves. The control signal generation module generates a light parameter control signal based on this information. As a result, the present invention can perform closed-loop feedback control based on biological signals such as heart rate variability, making up for the lack of feedback control in the existing technology and effectively regulating the user's mood and autonomic nervous state.

[0016] 2. Multi-dimensional dynamic intelligent adjustment of light stimulation parameters: The control signal generation module incorporates a built-in deep learning model, which uses a dynamic matched filter unit to track the heartbeat pulse waveform to generate heart rate cycle characteristics. The LSTM neural network deployed by the neural network inference unit then combines these heart rate cycle characteristics with the excitability ratio to output PWM-modulated light parameter control instructions. The light adjustment module adjusts the near-infrared light pulse waveform and beam width based on these instructions. It also adjusts the focus and light intensity through a focusing mechanism and multi-channel LED array. This invention leverages advanced signal processing and artificial intelligence algorithms to dynamically and intelligently adjust light stimulation parameters from multiple dimensions, achieving refined and personalized adjustment.

[0017] 3. Portable and suitable for daily continuous use: The irradiator adopts a wearable neck support structure. Compared with many related devices in the prior art that have complex systems, large volumes, and cumbersome operations, the present invention is more convenient to carry. It is also designed specifically for the stellate ganglion and is a portable non-invasive adjustment device that meets the market demand for daily continuous use devices. At the same time, the setting of the protection monitoring module, through the abnormality detection unit and the graded response unit, monitors the heart rate components and excitement ratios in real time and takes corresponding measures, which improves the safety of the device and further enhances its applicability in daily use scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of module connection of the present invention; Figure 2 It is a schematic diagram of the signal processing module; Figure 3 is a schematic diagram of a control signal generation module; Figure 4 is a schematic diagram of the light adjustment module; Figure 5 It is a schematic diagram of the protection monitoring module; Figure 6 This is a schematic diagram of the construction and deployment of the LSTM model; Figure 7 It is a functional structure topology diagram of each module of the present invention; Figure 8 It is a structural schematic diagram of a near-infrared light controllable illuminator.

[0019] Figure numerals: 100, physiological signal acquisition module; 200, signal processing module; 300, autonomic nerve analysis module; 400, control signal generation module; 500, light adjustment module; 600, protection monitoring module; 210, preprocessing unit; 220, variational mode decomposition unit; 410, dynamic matched filtering unit; 420, neural network reasoning unit; 510, focusing mechanism; 520, near-infrared LED; 610, abnormality detection unit; 620, graded response unit. DETAILED DESCRIPTION

[0020] The present invention will be further described below with reference to the accompanying drawings and examples, but they are not intended to limit the present invention.

[0021] Example: A near-infrared light controllable illuminator for stellate ganglion stimulation, such as Figure 1 and Figure 7 As shown, including: The physiological signal acquisition module 100 is used to obtain near-infrared light sensing signals of the carotid artery area in real time; The signal processing module 200 is connected to the physiological signal acquisition module 100 and is used to perform variational mode decomposition on the near-infrared sensing signal and extract the heart rate component; The autonomic nerve analysis module 300 communicates and interacts with the signal processing module 200, receives the heart rate component and calculates the excitability ratio of the sympathetic nerves to the parasympathetic nerves; A control signal generation module 400 is in communication with the autonomic nerve analysis module 300 and the signal processing module 200; the control signal generation module 400 has a built-in deep learning model to generate a light parameter control signal based on the heart rate cycle characteristics and the excitability ratio; The light adjustment module 500 is connected to the control signal generation module 400 and receives the light parameter control signal to control the emission of near-infrared light. Furthermore, the physiological signal acquisition module includes a contact photoplethysmography (CPPG) sensor (also known as a contact photoplethysmography sensor). Its optical receiving surface is equipped with a pressure-sensitive element, which is applied to the carotid artery and pressed firmly. When the sensor's pressure reaches a preset threshold, the signal acquisition circuit is activated, which can then collect the body's respiration and pulse signals.

[0022] Furthermore, if Figure 2 As shown, the signal processing module 200 includes: Preprocessing unit 210 performs baseline correction and bandpass filtering on the raw near-infrared light sensing signal. The purpose of the preprocessing unit is to remove baseline drift (DC component and low-frequency interference) and filter out high-frequency noise and motion artifacts. This is achieved using a bandpass filter (Butterworth filter). For heart rate, the appropriate frequency band is typically between 0.5 Hz and 3-4 Hz (corresponding to 30 BPM to 180-240 BPM). More strictly, to account for respiratory interference (typically 0.1-0.4 Hz) and high-frequency noise, a bandpass filter of 0.5 Hz to 10 Hz or higher can be used, preserving the heartbeat and respiratory components before separation by VMD. Removing the DC component can be achieved by high-pass filtering or by subtracting the mean / trend term.

[0023] The variational mode decomposition unit 220 performs multimodal decomposition on the signal processed by the preprocessing unit (210) and extracts the heart rate component. The variational mode decomposition (VMD) method adopted by the variational mode decomposition unit is an adaptive, non-recursive signal decomposition method that decomposes a signal into a series of modal components (Intrinsic Mode Functions, IMFs) with specific sparse characteristics. Unlike empirical mode decomposition (EMD), VMD is based on the solution of variational problems and realizes signal decomposition by constructing and solving a constrained optimization model. VMD assumes that the original signal f(t) can be decomposed into the sum of K modal components uk(t). Each mode uk(t) is assumed to be an amplitude-frequency modulation (AM-FM) signal with a limited bandwidth. The goal of VMD is to minimize the sum of the bandwidths of all modal components, while requiring that the sum of these modal components can accurately reconstruct the original signal.

[0024] The signal (the photoplethysmographic signal, acquired by a photoplethysmographic sensor, is a biological signal detected optically that primarily reflects changes in blood volume during the cardiac cycle) typically contains the superposition of multiple physiological components, including: 1. Direct current (DC) component: related to tissue optical properties, venous blood volume, etc.

[0025] 2. Alternating current (AC) component: Heartbeat component: The main component is the periodic change in blood volume caused by each heartbeat, and its frequency corresponds to the heart rate.

[0026] Respiratory component: Changes in thoracic pressure caused by breathing affect venous return and arterial blood volume, and the frequency corresponds to the respiratory rate.

[0027] Low-frequency components: related to vascular regulation (such as vasoconstriction).

[0028] It is suitable for separating components in these different frequency ranges. The number of modes, K, in VMD decomposition is typically chosen empirically. For PPG signals, considering heart rate, respiration, low-frequency components, and noise, K is typically chosen between 3 and 6. Choosing K requires careful consideration. A too small K may cause different physiological components to alias within the same mode; a too large K may decompose a single physiological component into multiple modes or produce noise modes. The bandwidth factor, α, affects the compactness of the modes. For PPG signals, a certain degree of frequency variation (AM-FM characteristics) must be tolerated, and this factor typically requires adjustment, typically choosing a value between 0.1 and 0.3.

[0029] Mode selection in decomposition includes the following methods: First, check the center frequency: Iterate over the center frequency of each mode. Heart rate usually falls within the normal physiological range of the human body (for example, 0.8 Hz - 3 Hz, corresponding to 48 - 180 BPM), and find the mode whose center frequency falls within this target range.

[0030] Energy / Power Analysis: If the center frequencies of multiple modes fall within the target range, or if you want to find the "strongest" heartbeat mode, you can calculate the energy or power of each mode and select the mode with the highest energy / power as the heart rate mode. Alternatively, you can view the power spectrum of each mode and select the mode with the most significant peak within the heart rate frequency range.

[0031] Comprehensive judgment: Make a judgment based on the waveform characteristics of the center frequency and mode (whether there is an obvious pulse waveform).

[0032] In this embodiment, the multiple intrinsic mode function (IMF) components decomposed by the variational mode decomposition unit contain signals with different frequency components. Because the heart rate signal has a specific frequency range, by analyzing the frequency characteristics of each IMF component, IMF components with frequencies within the heart rate-related range are screened out. These components are the heart rate components. For example, the normal adult heart rate range is generally 60-100 beats / minute, corresponding to a frequency range of approximately 1-1.67 Hz. Selecting IMF components within this frequency range for subsequent analysis can obtain characteristic information related to the heart rate, which can be used for subsequent autonomic nervous system analysis and control signal generation.

[0033] Furthermore, the autonomic nervous system (ANS) regulates heart rhythm through the sympathetic nervous system (activating the "fight or flight" response) and the parasympathetic nervous system (activating the "rest and digest" response). The dynamic balance between these two is reflected in subtle variations in the heartbeat interval (RR interval), known as heart rate variability (HRV). Spectral analysis can decompose the HRV signal into different frequency bands, corresponding to different physiological mechanisms: High frequency (HF, 0.15–0.4 Hz): Strongly associated with parasympathetic nervous system activity, reflecting respiratory sinus arrhythmia (RSA), the effect of respiration on heart rate.

[0034] Low frequency (LF, 0.04–0.15 Hz): This is associated with the co-regulation of the sympathetic and parasympathetic nervous systems and is often associated with blood pressure regulation (e.g., the baroreflex). However, the physiological significance of LF is controversial, with some studies suggesting that it reflects sympathetic nervous system activity.

[0035] Very low frequency (VLF, 0.003–0.04 Hz): Associated with long-term physiological processes such as temperature regulation and hormone cycles, it is less commonly used in autonomic nervous system assessment.

[0036] Therefore, in this embodiment, the autonomic nerve analysis module 300 uses a double elliptic filter group to separate the low-frequency and high-frequency band energies; the double elliptic filter group includes: The first elliptical filter: passband 0.04-0.15Hz, used to extract low-frequency energy components; Second elliptical filter: passband 0.15-0.4Hz, used to extract high-frequency energy components; The autonomic nerve analysis module 300 calculates the energy ratio of the low-frequency energy component to the high-frequency energy component to obtain the sympathetic nerve excitability ratio. The parasympathetic nerve excitability is directly related to the HF power.

[0037] In this example, a Python program is used to extract HRV frequency domain features (LF / HF ratio) from the heart rate component signal. This program includes the complete process of R wave detection, RR interval calculation, interpolation, trend removal, and power spectrum analysis: The R-wave peaks are detected from the heart rate component signal, and the time intervals between adjacent R waves (RR intervals) are calculated to form a non-uniformly sampled time series. The non-uniform RR interval series is converted to a uniformly sampled signal (typically at a 4 Hz sampling rate) through interpolation (e.g., cubic spline interpolation). Linear or nonlinear trends are removed (e.g., using polynomial fitting). The power spectral density is calculated using a Fourier transform (FFT). The power spectrum is integrated over the LF (0.04-0.15 Hz) and HF (0.15-0.4 Hz) bands to calculate the power and derive the LF / HF ratio.

[0038] Furthermore, if Figure 3 As shown, the control signal generating module 400 includes: A dynamic matched filter unit 410 uses a matched filter to track the heartbeat pulse waveform and generate a heart rate cycle feature; Since the heartbeat pulse has a relatively stable shape (rapid rise, peak, slow fall, and may contain dicrotic waves), this shape can be used as a template for matched filtering. Matched filtering can enhance the main components of the pulse wave and suppress noise and non-pulse components, making the peak of the pulse wave more prominent and easier to detect. However, since the shape of the pulse wave is not completely fixed, it will be affected by various factors such as respiration, vascular tension, and changes in body position and undergo slight changes. If a fixed template is used, the effect of matched filtering will decrease as the signal shape changes. Therefore, in this embodiment, a matched filter is used to dynamically track the heartbeat pulse waveform, and the template of the matched filter is updated through dynamic tracking. Specifically, the process of the matched filter tracking the heartbeat pulse waveform and generating the heart rate cycle characteristics is: Use peak detection methods (such as threshold method, derivative-based method) to find the first few pulse waves of the heartbeat pulse waveform and average them as the initial template , based on the currently selected initial template, construct the impulse response of the matched filter : : Where: is the impulse response of the matched filter, For the initial template, For time, is the duration of the initial template (should be roughly equal to one heartbeat cycle or slightly shorter, enough to include the rising edge and peak); The current input signal segment (or the entire signal) and the impulse response of the matched filter Perform convolution to get the output : ; The convolution operation enhances the main component of the pulse wave, suppresses noise and non-pulse components, and makes the peak of the pulse wave more prominent; Search for the local maximum within the expected heartbeat interval, which represents the position in the input signal that has the highest match with the template, and then extract the local maximum from the signal segment. The peak time point detected As the center, extract a template duration For the same signal segment, the original template is updated with the extracted signal segment to obtain a new template, thereby adapting to the change of the pulse wave shape and improving the matched filtering effect; Finally, the above steps of template construction, convolution calculation, peak search, and template update are repeated to process the input signal segment by segment or continuously. Each iteration is based on the previous processing result, and the template of the matched filter is continuously adjusted to accurately track the heartbeat pulse waveform, thereby generating accurate heart rate cycle characteristics.

[0039] The neural network inference unit 420 deploys an LSTM neural network, which receives the heart rate cycle characteristics and the excitement ratio and outputs a PWM modulated light parameter control instruction.

[0040] In this embodiment, Figure 6 As shown, the LSTM neural network is deployed through the following steps: a) Train LSTM models for four types of stimulation: an excitatory stimulus model with a slow rise and rapid fall, a soothing stimulus model with a fast rise and slow fall, a sinusoidal daily stimulus model, and an intense impulse stimulus model with a sharp rise and fall. LSTM models are used to predict the stimulus signal at time t+1 based on the input heart rate cycle characteristics and the excitability ratio. The models are trained until convergence. The hidden layer size of the LSTM model is typically 64-256, which needs to be adjusted based on data complexity. The number of layers ranges from 1 to 3; deeper networks can capture more complex temporal dependencies.

[0041] b) Using knowledge distillation, the knowledge from the four models is compressed into a single lightweight network (this example uses a relatively simple small LSTM model). The process is as follows: First, set the distillation temperature T. A higher temperature can smooth the probability distribution, facilitating knowledge transfer. For each training sample, forward propagation is performed through the four base models and the small LSTM model. The softmax probability distributions (temperature-scaled) of the four base model outputs are calculated and used as "soft labels." A distillation loss function is defined. For example, the KL divergence (Kullback-Leibler Divergence) can be used to measure the difference between the output of the small LSTM model (the temperature-scaled softmax probability distribution) and the output of the base model (the soft labels). The overall loss function is also combined with the prediction loss of the small LSTM model for the true labels (such as mean squared error loss for regression tasks or cross entropy loss for classification tasks). During training, the parameters of the small LSTM model are continuously adjusted through backpropagation, so that it fits the true labels while also maximizing the knowledge of the base models. Use the test dataset to evaluate the small LSTM model after knowledge distillation training, and calculate relevant evaluation indicators (such as accuracy, mean square error, etc., determined according to the specific task) to verify the performance of the model.

[0042] In practical applications, real-time input data containing 4-bit one-hot encoding is input into a trained small LSTM model. The model outputs the stimulation signal prediction result at time t+1, providing a basis for generating control signals for the near-infrared light controllable illuminator for stellate ganglion stimulation.

[0043] c) Deployment on embedded devices is achieved through 4-bit quantization compression and the TF-Lite framework. 4-bit quantization is the process of reducing the precision of data in the model (such as weights, activation values) from the usual 32-bit floating point numbers or 8-bit integers to 4 bits for representation. This can significantly reduce the storage size of the model and the memory usage during the calculation process. TF-Lite is a lightweight deep learning inference framework developed by Google, designed for running in resource-constrained environments such as mobile devices, embedded devices, and IoT devices. It has the advantages of small size, fast speed, and good compatibility. TF-Lite provides a series of tools and APIs to facilitate the conversion of trained TensorFlow models into a format suitable for running on embedded devices and perform efficient inference calculations. Through 4-bit quantization compression and the TF-Lite framework, deep learning models can be effectively deployed on embedded devices. Furthermore, if Figure 4 As shown, the light adjustment module 500 includes: Focusing mechanism 510 drives lens displacement based on the focus parameter in the light parameter control signal. In this embodiment, focusing mechanism 510 utilizes a macro-focusing stepper motor connected to the lens. Upon receiving a control command corresponding to the focus parameter, the controller drives the motor. If the focus parameter requires an increase in focus, i.e., a more concentrated illumination of the stellate ganglion region by the near-infrared light, the controller sends a pulse signal to cause the stepper motor to rotate forward a certain number of steps. Conversely, if the focus parameter requires a decrease in focus, the motor rotates backward a corresponding number of steps. As the lens shifts, the focus of the near-infrared light changes. As the lens moves closer to the light source, the light converges closer to the irradiated area; as the lens moves further away from the light source, the light converges further away. In this way, by precisely adjusting the lens displacement based on the focus parameter, the focusing effect of the near-infrared light can be controlled, enabling more precise targeting of the stellate ganglion by the near-infrared light, improving the stimulation effect while minimizing the impact on surrounding tissue.

[0044] Near-infrared LED 520 adjusts near-infrared light output based on the intensity parameter in the light parameter control signal. The near-infrared LED comprises a plurality of LEDs in an array, with each LED or group of LEDs having an independent driver circuit. These driver circuits operate based on the processed intensity parameter signal. This regulation utilizes pulse-width modulation (PWM) technology, which modulates the LED's average current by varying the duty cycle of the pulse signal, thereby controlling its luminous intensity. If the intensity parameter requires enhanced near-infrared light output, the driver circuit increases the duty cycle of the PWM signal, causing the LED to illuminate longer per unit time and emit more intense light. Conversely, if the intensity parameter requires reduced light output, the duty cycle is reduced, shortening the LED's illumination time and reducing light intensity. Independently regulating each LED or group of LEDs enables precise control of near-infrared light output intensity. This not only allows for real-time adjustment of light stimulation intensity based on the user's physiological state to meet personalized needs, but also allows for flexible control of light output in different application scenarios, improving the device's applicability and safety. When the user's autonomic nervous state changes, the intensity parameters will change accordingly. The LED array can respond quickly and adjust the near-infrared light output intensity to ensure that the stimulation of the stellate ganglion is always in the best state.

[0045] Furthermore, the near-infrared light controllable illuminator further includes a protection monitoring module 600, such as Figure 5 As shown, the protection monitoring module 600 includes: The abnormality detection unit 610 is used to monitor heart rate components and / or arousal ratios in real time. The abnormality detection unit 610 is connected to the signal processing module 200 and the autonomic nervous system analysis module 300 to acquire processed heart rate component data and the arousal ratio data of the sympathetic and parasympathetic nervous systems in real time. This data is an important reflection of the human body's physiological state, and the abnormality detection unit uses this data as a basis for analysis and judgment. Heart rate component data reflects the heart's beating, while the arousal ratio reflects the balance of the autonomic nervous system. The abnormality detection unit 610 pre-sets normal range thresholds for heart rate components and arousal ratios. A normal heart rate component fluctuates within a specific frequency range, and the arousal ratio also has a corresponding reasonable range. The abnormality detection unit 610 continuously compares the acquired real-time data with these thresholds. If a heart rate component exceeds the frequency range corresponding to a normal heart rate, or if the arousal ratio deviates from the normal autonomic nervous system balance range, the corresponding abnormality detection mechanism is triggered. When the heart rate component is too high or too low, it may mean that the user's heart state is abnormal; if the excitement ratio is abnormally high or low, it indicates that the balance of the autonomic nervous system is broken, which may affect the safety and effectiveness of the device.

[0046] The hierarchical response unit 620 has a three-level response mechanism, initiating different levels of response measures based on the severity of the abnormality. If the heart rate component or arousal ratio only slightly exceeds the first threshold, it may trigger a stimulation mode switch, adjusting the near-infrared light stimulation mode to prevent further abnormalities. If it exceeds the second threshold, an audible and visual alarm is activated to alert the user and relevant personnel. If it exceeds the third threshold, the power is immediately cut off and the device is stopped to ensure user safety.

[0047] Furthermore, if Figure 8 As shown, the near-infrared light controllable illuminator in this embodiment is a wearable structure that surrounds the neck as a whole. A PCB control board is provided at the rear of the neck, wherein the signal processing module 200, the autonomic nerve analysis module 300 and the control signal generation module are all integrated into the PCB control board. Near-infrared light emitting components, such as near-infrared LEDs 520, are arranged around the neck to generate near-infrared light and irradiate the target area (such as the stellate ganglion). In addition, the near-infrared light controllable illuminator also includes a power supply module, such as a built-in battery, to provide power to the PCB control board and the near-infrared light emitting components.

[0048] In summary, the present invention realizes intelligent regulation of near-infrared light based on the near-infrared light sensing signals collected from the carotid artery area, through a series of module processing and analysis, which makes up for the problems of lack of feedback control and single control dimension in the prior art, and can perform closed-loop feedback control according to the physiological and emotional state of the human body, thereby realizing refined and personalized regulation. The present invention adopts a wearable neck support structure to improve portability and is suitable for daily continuous use. The setting of the protection monitoring module of the present invention can monitor key indicators in real time and respond in a graded manner to ensure safe use, thereby providing a more effective, safer and convenient solution for non-invasive photobiomodulation of the stellate ganglion.

Claims

1. A near-infrared light controllable illuminator for stellate ganglion stimulation, characterized in that: include: A physiological signal acquisition module (100) for acquiring near-infrared light sensing signals from the carotid artery region in real time; a signal processing module (200), connected to the physiological signal acquisition module (100), for performing variational modal decomposition on the near-infrared sensing signal and extracting a heart rate component; An autonomic nerve analysis module (300) communicates and interacts with the signal processing module (200), receives the heart rate component and calculates the excitability ratio of the sympathetic nerves to the parasympathetic nerves; A control signal generation module (400) is communicatively connected to the autonomic nerve analysis module (300) and the signal processing module (200); the control signal generation module (400) has a built-in deep learning model, and generates a light parameter control signal according to the heart rate cycle characteristics and the excitability ratio; A light adjustment module (500), connected to the control signal generation module (400), receives the light parameter control signal to control the emission of near-infrared light; The protection monitoring module (600) is connected to the signal processing module (200), the autonomic nerve analysis module (300), and the control signal generation module (400).

2. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 1, characterized in that: The physiological signal acquisition module includes a contact CPPG sensor, whose optical receiving surface is provided with a pressure sensitive element. When the pressure of the sensor reaches a preset threshold, the signal acquisition circuit is activated to perform signal acquisition.

3. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 1, characterized in that: The signal processing module (200) comprises: A pre-processing unit (210) performs baseline correction and band-pass filtering on the original near-infrared light sensing signal; The variational mode decomposition unit (220) performs multi-modal decomposition on the signal processed by the pre-processing unit (210) and extracts the heart rate component.

4. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 1, characterized in that: The autonomic nerve analysis module (300) uses a double elliptical filter bank to separate low-frequency and high-frequency band energy; The double elliptic filter group comprises: The first elliptical filter: passband 0.04-0.15Hz, used to extract low-frequency energy components; Second elliptical filter: passband 0.15-0.4Hz, used to extract high-frequency energy components; The autonomic nerve analysis module (300) calculates the energy ratio of the low-frequency energy component to the high-frequency energy component to obtain the excitement ratio.

5. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 1, characterized in that: The control signal generating module (400) comprises: A dynamic matched filter unit (410) uses a matched filter to track the heartbeat pulse waveform and generate a heart rate cycle feature; The neural network inference unit (420) deploys an LSTM neural network, which receives the heart rate cycle characteristics and the excitability ratio and outputs a PWM modulated light parameter control instruction.

6. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 5, characterized in that: The process of the matched filter tracking the heartbeat pulse waveform and generating the heart rate cycle characteristics is: Use the peak detection method to find the first few pulse waves of the heartbeat pulse waveform, average them as the initial template, and construct the impulse response of the matched filter based on the currently selected initial template: ： Where: is the impulse response of the matched filter, For the initial template, For time, The duration of the initial template; The current input signal segment Impulse response of the matched filter Perform convolution to get the output : ； The convolution operation enhances the main component of the pulse wave, suppresses noise and non-pulse components, and makes the peak of the pulse wave more prominent; Search for the local maximum within the expected heartbeat interval, which represents the position in the input signal that has the highest match with the template, and then extract the local maximum from the signal segment. In the example, a time point corresponding to the template duration is extracted with the detected peak time point as the center. For the same signal segment, the original template is updated with the extracted signal segment to obtain a new template, thereby adapting to the change of the pulse wave shape and improving the matched filtering effect; Finally, the above steps of template construction, convolution calculation, peak search, and template update are repeated to process the input signal segment by segment or continuously. Each iteration is based on the previous processing result, and the template of the matched filter is continuously adjusted to accurately track the heartbeat pulse waveform, thereby generating accurate heart rate cycle characteristics.

7. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 5, characterized in that: The LSTM neural network is deployed through the following steps: a) Train four basic LSTM models corresponding to excitatory, soothing, daily, and intense stimulation modes; b) Using temperature scaling distillation to transfer the knowledge of the four models into a single lightweight network; c) Deployment on embedded devices is achieved through 4-bit quantization compression and the TF-Lite framework.

8. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 1, characterized in that: The light adjustment module (500) comprises: A focusing mechanism (510) drives the lens to move according to the focus parameter in the light parameter control signal; A near-infrared LED (520) adjusts near-infrared light output according to the intensity parameter in the light parameter control signal.

9. The near-infrared light controllable irradiator for stellate ganglion stimulation according to claim 1, characterized in that: The protection monitoring module (600) comprises: An abnormality detection unit (610), for monitoring heart rate components and / or excitement ratios in real time; The hierarchical response unit (620) is configured with a three-level response mechanism, which triggers stimulation mode switching when the monitored heart rate component and / or excitement ratio exceeds a first threshold, activates an audible and visual alarm when it exceeds a second threshold, and cuts off power when it exceeds a third threshold.

10. The near-infrared light controllable illuminator for stellate ganglion stimulation according to any one of claims 1 to 9, characterized in that: The irradiator adopts a wearable neck support structure.

Images