Sleep-awakening circadian rhythm disorder adjusting method and device based on ultrasonic stimulation

Abstract

The invention relates to the technical field of data analysis and processing, and discloses a sleep-awakening circadian rhythm disorder adjusting method and device based on ultrasonic stimulation. The method comprises the following steps: collecting sleep physiological data of a user to construct a feature matrix; analyzing a sleep stage to obtain a disorder type; generating an ultrasonic stimulation parameter scheme; determining brain region stimulation coordinates to generate spatial positioning data; predicting user feedback through the convolutional neural network; and comparing the prediction and actual feedback adjustment schemes to obtain an optimized target stimulation scheme. According to the method and the device, accurate recognition, personalized stimulation parameter generation, accurate brain region positioning, predictive evaluation and dynamic scheme optimization of sleep-awakening circadian rhythm disorders are realized, so that the regulation effect of ultrasonic stimulation on various sleep-awakening circadian rhythm disorders is improved.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis and processing, and in particular to a method and device for regulating sleep-wake circadian rhythm disorders based on ultrasonic stimulation. Background Art

[0002] Circadian rhythm sleep-wake disorder (CRSWD) is a common health problem, which is mainly manifested by the sleep-wake cycle being out of sync with the normal circadian rhythm, including phase delay syndrome, phase advance syndrome, non-24-hour sleep-wake syndrome and irregular sleep-wake rhythm. Common symptoms of circadian rhythm disorders are difficulty falling asleep, difficulty maintaining sleep and excessive sleepiness. In severe cases, they can further affect health, impair social function, work, life, study and safety. At present, the main methods used in clinical intervention are drug therapy, light therapy, melatonin supplementation and cognitive behavioral therapy. Although drug therapy is effective quickly, it is easy to become dependent and has many side effects; light therapy requires strict control of time and intensity and is inconvenient to use; the efficacy of melatonin supplementation varies from person to person, and it is difficult to ensure regular use; cognitive behavioral therapy requires long-term persistence and poor patient compliance. In recent years, non-invasive brain stimulation technologies such as transcranial magnetic stimulation (r-TMS) have been used in the treatment of sleep disorders and have shown good therapeutic effects. However, the large size of the equipment, complex operation and high cost have limited its widespread application.

[0003] As an emerging non-invasive brain regulation method, the existing transcranial ultrasound stimulation (TUS) technology has been gradually applied to the field of neuroregulation due to its advantages such as moderate penetration depth, high spatial accuracy, and portable equipment. However, the current TUS technology still has obvious shortcomings in the treatment of sleep-wake circadian rhythm disorders: first, the setting of stimulation parameters lacks scientific basis and is mostly determined by experience, resulting in large differences in individualized treatment effects; second, there is a lack of accurate mechanism for locating target brain areas, making it difficult to achieve precise stimulation of specific sleep regulation brain areas; third, there is a lack of a closed-loop feedback system, and the treatment plan cannot be dynamically adjusted according to the patient's actual response; fourth, the treatment effect prediction mechanism is imperfect, making it difficult to evaluate the possible treatment effect before treatment. These problems seriously restrict the application effect and promotion value of ultrasound stimulation technology in the regulation of sleep-wake circadian rhythm disorders. Summary of the invention

[0004] The present application provides a method and device for regulating sleep-wake circadian rhythm disorders based on ultrasonic stimulation, which is used to achieve accurate identification of sleep-wake circadian rhythm disorders, generation of personalized stimulation parameters, precise brain area positioning, predictive evaluation and dynamic scheme optimization, thereby improving the regulating effect of ultrasonic stimulation on various sleep-wake circadian rhythm disorders, and overcoming technical defects in the prior art such as arbitrary parameter setting, inaccurate positioning, and lack of feedback mechanism.

[0005] In the first aspect, the present application provides a method for regulating sleep-wake circadian rhythm disorders based on ultrasonic stimulation, and the method for regulating sleep-wake circadian rhythm disorders based on ultrasonic stimulation includes: collecting the user's breathing rate data, heart rate data, blood oxygen concentration data and electromyography data during sleep, and obtaining the user's facial image, recording brain wave data, and constructing a sleep cycle feature matrix; according to the sleep cycle feature matrix, analyzing the user's sleep stage distribution, measuring the sleep latency and the number of awakenings at night, and generating a sleep-wake circadian rhythm disorder type determination result; according to the sleep-wake circadian rhythm disorder type determination result, generating an ultrasonic stimulation parameter combination, including stimulation frequency, The ultrasonic stimulation scheme is obtained by analyzing the stimulation intensity, stimulation duration, stimulation waveform and stimulation focal position; according to the ultrasonic stimulation scheme, the stimulation coordinates of key brain areas are determined, including the coordinates of the pineal region, the hypothalamus region, the prefrontal region and the preoptic region, and spatial positioning data are generated; through the spatial positioning data, according to the ultrasonic stimulation scheme, user feedback prediction based on a convolutional neural network is performed to obtain predicted feedback data; the actual feedback data of the user stimulated by the ultrasonic stimulation scheme is obtained, feedback difference data is generated according to the predicted feedback data and the actual feedback data, and the ultrasonic stimulation scheme is corrected according to the feedback difference data to obtain a target stimulation scheme.

[0006] In a second aspect, the present application provides a sleep-wake circadian rhythm disorder regulating device based on ultrasound stimulation, wherein the sleep-wake circadian rhythm disorder regulating device based on ultrasound stimulation comprises: The acquisition module is used to collect the user's breathing rate data, heart rate data, blood oxygen concentration data and electromyography data during sleep, obtain the user's facial image, record brain wave data, and construct a sleep cycle feature matrix; An analysis module, used to analyze the user's sleep stage distribution according to the sleep cycle feature matrix, measure the sleep latency and the number of awakenings at night, and generate a sleep-wake circadian rhythm disorder type determination result; A generation module, configured to generate an ultrasound stimulation parameter combination, including stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform, and stimulation focus position, according to the sleep-wake circadian rhythm disorder type determination result, to obtain an ultrasound stimulation plan; A stimulation module, used to determine the stimulation coordinates of key brain regions according to the ultrasound stimulation scheme, including the coordinates of the pineal region, the hypothalamus region, the prefrontal region and the preoptic region, and generate spatial positioning data; A prediction module, configured to perform user feedback prediction based on a convolutional neural network according to the spatial positioning data and the ultrasound stimulation scheme to obtain predicted feedback data; The correction module is used to obtain actual feedback data of the user stimulated by the ultrasonic stimulation scheme, generate feedback difference data according to the predicted feedback data and the actual feedback data, and correct the ultrasonic stimulation scheme according to the feedback difference data to obtain a target stimulation scheme.

[0007] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned sleep-wake circadian rhythm disorder regulation method based on ultrasonic stimulation.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, which stores instructions, and when the computer-readable storage medium is run on a computer, it enables the computer to execute the above-mentioned sleep-wake circadian rhythm disorder regulation method based on ultrasonic stimulation.

[0009] In the technical solution provided by this application, the sleep cycle feature matrix constructed by multi-dimensional physiological data collection and integration greatly improves the accuracy of sleep disorder judgment. Compared with the analysis of a single data source, the matrix contains six heterogeneous data including respiratory rate, heart rate, blood oxygen concentration, electromyographic data, facial image and brain wave data, providing a more comprehensive representation of sleep state and making sleep stage identification more accurate. Secondly, the segmented setting method of ultrasonic stimulation parameters based on the results of the determination of the type of sleep-wake circadian rhythm disorder selects a specific frequency range for different types of disorders, effectively solving the problem of inaccurate traditional empirical parameter setting, especially through targeted frequency selection (phase delay type 200-500kHz, phase advance type 500-800kHz, irregular type 800-1200kHz), significantly improving the accuracy of ultrasonic energy regulation of the target brain area and enhancing the targeted treatment. Furthermore, the safe stimulation intensity calculation method combined with the skull attenuation factor effectively balances the treatment effect and safety, ensuring that the ultrasonic energy can penetrate the skull to reach the target brain area without exceeding the tissue damage threshold, thereby reducing the incidence of adverse reactions. In addition, the precise positioning technology of key brain areas has improved the stimulation positioning accuracy to the millimeter level, which is a qualitative leap compared to the centimeter-level accuracy of traditional methods. It directly improves the efficiency of ultrasonic energy transmission by forming spatial positioning data. It is particularly worth emphasizing that the technical feature of this scheme using convolutional neural networks for user feedback prediction has made an outstanding contribution. The model realizes the mapping from stimulation parameters to expected effects through a three-layer convolution structure and a fully connected layer, reducing the cost of trial and error in treatment. The dynamic scheme adjustment mechanism based on feedback difference data forms a complete closed-loop control system. By comparing the predicted feedback with the actual feedback in real time, the parameter deviation is accurately calculated, and the ultrasonic stimulation scheme is optimized accordingly, which realizes the personalization, precision and efficiency of the analysis of sleep-wake circadian rhythm disorder regulation strategies as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 This is a schematic diagram of an embodiment of a method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation in an embodiment of the present application; Figure 2 This is a schematic diagram of an embodiment of a sleep-wake circadian rhythm disorder regulating device based on ultrasound stimulation in an embodiment of the present application; Figure 3 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a method and device for regulating sleep-wake circadian rhythm disorders based on ultrasonic stimulation. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiments of the present application, an embodiment of the sleep-wake circadian rhythm disorder regulation method based on ultrasound stimulation includes: Step S101, collecting the user's breathing rate data, heart rate data, blood oxygen concentration data and electromyography data during sleep, obtaining the user's facial image, recording brain wave data, and constructing a sleep cycle feature matrix; Step S102: Analyze the sleep stage distribution of the user according to the sleep cycle characteristic matrix, measure the sleep latency and the number of awakenings at night, and generate a sleep-wake circadian rhythm disorder type determination result; Step S103, generating an ultrasound stimulation parameter combination according to the sleep-wake circadian rhythm disorder type determination result, including stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform and stimulation focus position, to obtain an ultrasound stimulation plan; Step S104, according to the ultrasound stimulation scheme, determining the stimulation coordinates of key brain regions, including the coordinates of the pineal region, the hypothalamus region, the prefrontal region and the preoptic region, and generating spatial positioning data; Step S105, performing user feedback prediction based on a convolutional neural network according to the spatial positioning data and the ultrasound stimulation scheme to obtain predicted feedback data; Step S106, obtaining actual feedback data of the user stimulated by the ultrasound stimulation scheme, generating feedback difference data according to the predicted feedback data and the actual feedback data, and modifying the ultrasound stimulation scheme according to the feedback difference data to obtain a target stimulation scheme.

[0014] It is understandable that the execution subject of the present application may be a sleep-wake circadian rhythm disorder regulating device based on ultrasound stimulation, or a terminal or a server, which is not limited here. The present application embodiment is described by taking the server as the execution subject as an example.

[0015] Specifically, when collecting multi-dimensional physiological data during the user's sleep, a multi-channel biosensor array is used to record respiratory rate, heart rate, blood oxygen concentration and electromyographic signals. These sensors are usually placed under the user's mattress or in a wearable device, without disturbing the user's normal sleep. At the same time, the infrared imaging device collects the user's facial image every hour, and the portable electroencephalogram device continuously records the brain wave data. After these raw data are noise-removed and smoothed, time domain features such as heart rate variability and frequency domain features such as respiratory frequency power spectrum density are extracted. The expression change features in the facial image and the energy distribution of the four bands (δ waves, θ waves, α waves and β waves) in the brain wave are integrated by time to form a sleep cycle feature matrix. This matrix contains the changes in the user's physiological state throughout the sleep process. When analyzing the sleep stage distribution according to the sleep cycle feature matrix, the system extracts key time points from the brain wave feature sequence and divides the sleep stage boundaries by the feature threshold segmentation method. For example, when the proportion of δ waves in the brain wave exceeds 40% and the electromyographic activity decreases, it is determined to be a deep sleep stage; when rapid eye movement occurs and the brain wave presents a similar awake state, it is determined to be REM sleep. The system calculates the ratio of REM sleep to non-REM sleep, measures the time from waking up to the first entry into light sleep (sleep latency), and identifies the number of awakenings at night. These parameters are compared with standard sleep patterns to generate a structural abnormality index, which is then weighted and fused to form a sleep disorder score. Ultimately, the system uses these data to determine the type of sleep-wake circadian rhythm disorder the user has, such as phase delay syndrome, phase advance syndrome, or irregular sleep-wake rhythm.

[0016] When generating a combination of ultrasonic stimulation parameters based on the results of the sleep-wake circadian rhythm disorder type determination, the system first extracts the parameter range of the corresponding type from the ultrasonic parameter basic library. For phase-delay type disorders, select a frequency range of 200-500kHz; for phase-advance type, select 500-800kHz; for irregular type, select 800-1200kHz. The system calculates the safe stimulation intensity range based on the frequency parameters, taking into account the skull attenuation factor to ensure that the ultrasonic energy can effectively penetrate the skull without causing tissue damage. The stimulation duration is set for different sleep cycle regulation targets, such as a longer time for deep sleep promotion and a shorter time for rapid eye movement sleep regulation. The waveform type selection is also targeted: continuous waves are used for slow regulation, pulse waves are used for fast regulation, and modulated waves are used for mixed regulation. These parameters are combined to form an ultrasonic stimulation plan.

[0017] When determining the stimulation coordinates of key brain areas, first obtain the user's head tomography image, extract the skull structure features and brain tissue boundaries, and form a three-dimensional brain structure model. In this model, mark the boundaries of key brain areas such as the pineal region (responsible for melatonin secretion), hypothalamus region (biological clock center), prefrontal region (executive function regulation), and preoptic region (sleep initiation), and calculate the geometric center point of each area as the initial stimulation point. According to the requirements of the stimulation focus position in the ultrasound stimulation scheme, the initial coordinates are precisely adjusted. The system also measures the thickness and density distribution of the user's skull at each stimulation point, calculates the attenuation coefficient of the ultrasound penetration path, and forms a skull sound transmission map. Finally, combined with the precise stimulation point coordinates and the skull sound transmission map, the phase delay matrix and energy distribution function of the multi-element ultrasonic transducer are calculated to generate spatial positioning data.

[0018] When predicting user feedback, the system combines spatial positioning data with the ultrasound stimulation scheme to generate a stimulation target description matrix. Similar cases are extracted from the historical user database to build a feedback feature vector library. The stimulation target description matrix is ​​processed by a three-layer convolutional neural network. The first layer uses 32 3×3 convolution kernels to extract low-level features (such as spatial position features), the second layer uses 64 3×3 convolution kernels to extract intermediate features (such as parameter combination features), and the third layer uses 128 3×3 convolution kernels to extract high-level features (such as stimulation effect features). After each layer of convolution, maximum pooling and batch normalization are performed to generate a multi-dimensional stimulation feature map. High-dimensional features are mapped to the sleep parameter evaluation space through two fully connected layers (256 and 128 neurons) to form a stimulation effect map. Finally, the correlation between the stimulation effect map and the feedback feature vector library is calculated to generate predicted feedback data.

[0019] The system obtains actual user feedback data and makes program corrections. After the user receives ultrasound stimulation, the system monitors changes in sleep quality, including the frequency of sleep stage transitions, the proportion of deep sleep, and the duration of rapid eye movement sleep. At the same time, it collects changes in physiological data such as heart rate variability, respiratory rhythm, and body temperature curve, as well as users' subjective reports on the difficulty of falling asleep, nighttime awakening, and morning mental state. These data are integrated into actual feedback data, compared with the predicted feedback data, and the feedback difference matrix is ​​calculated. According to the deviation direction and amplitude of each parameter in the difference matrix, the ultrasound stimulation program is precisely adjusted to form a target stimulation program.

[0020] For example, after a user was tested by the system, the sleep cycle characteristic matrix showed that his δ wave energy distribution was abnormal, the sleep latency reached 45 minutes (normal value <30 minutes), he woke up 6 times at night (normal value <3 times), and the rapid eye movement sleep ratio was only 15% (normal value 20-25%). The system determined that it was a phase-delayed sleep-wake circadian rhythm disorder and generated initial ultrasound stimulation parameters: frequency 350kHz, intensity 0.5W / cm², duration 5 minutes, continuous wave form, focusing on the suprachiasmatic nucleus of the hypothalamus. The convolutional neural network predicted that the program would reduce the sleep latency to 32 minutes and the night awakening to 4 times. After the actual treatment, the user reported that the sleep latency was 29 minutes and he woke up 3 times at night. The system calculated the feedback difference data and adjusted the ultrasound parameters to: frequency 380kHz, intensity 0.6W / cm², duration 6 minutes, forming a more accurate target stimulation program, and achieving effective regulation of sleep-wake circadian rhythm disorders.

[0021] In the embodiment of the present application, the sleep cycle feature matrix constructed by multi-dimensional physiological data collection and integration greatly improves the accuracy of sleep disorder judgment. Compared with the analysis of a single data source, the matrix contains six heterogeneous data including respiratory rate, heart rate, blood oxygen concentration, electromyographic data, facial image and brain wave data, providing a more comprehensive representation of sleep state and making sleep stage identification more accurate. Secondly, the ultrasonic stimulation parameter segmentation setting method based on the results of the determination of the type of sleep-wake circadian rhythm disorder selects a specific frequency range for different types of disorders, effectively solving the problem of inaccurate traditional empirical parameter setting, especially through targeted frequency selection (phase delay type 200-500kHz, phase advance type 500-800kHz, irregular type 800-1200kHz), significantly improving the accuracy of ultrasonic energy regulation of the target brain area and enhancing the targeted treatment. Furthermore, the safe stimulation intensity calculation method combined with the skull attenuation factor effectively balances the treatment effect and safety, ensuring that the ultrasonic energy can penetrate the skull to reach the target brain area without exceeding the tissue damage threshold, thereby reducing the incidence of adverse reactions. In addition, the precise positioning technology of key brain areas has improved the stimulation positioning accuracy to the millimeter level, which is a qualitative leap compared to the centimeter-level accuracy of traditional methods. It directly improves the efficiency of ultrasonic energy transmission by forming spatial positioning data. It is particularly worth emphasizing that the technical feature of this scheme using convolutional neural networks for user feedback prediction has made an outstanding contribution. The model realizes the mapping from stimulation parameters to expected effects through a three-layer convolution structure and a fully connected layer, reducing the cost of trial and error in treatment. The dynamic scheme adjustment mechanism based on feedback difference data forms a complete closed-loop control system. By comparing the predicted feedback with the actual feedback in real time, the parameter deviation is accurately calculated, and the ultrasonic stimulation scheme is optimized accordingly, which realizes the personalization, precision and efficiency of the analysis of sleep-wake circadian rhythm disorder regulation strategies as a whole.

[0022] In a specific embodiment, the process of executing step S101 may specifically include the following steps: (1) Collect the user's breathing rate data, heart rate data, blood oxygen concentration data, and electromyography data during seven days of continuous sleep through a multi-channel biosensor array to obtain raw physiological data; (2) Using an infrared imaging device to collect the user's facial image once every hour during the user's sleep to obtain a facial image sequence; (3) Recording the user's brain wave data during sleep through a portable brain wave detection device to obtain brain wave time series data; (4) performing noise elimination and data smoothing processing on the original physiological data to obtain processed physiological data; (5) extracting time domain features and frequency domain features from the processed physiological data to obtain physiological feature values; (6) extracting facial expression change features from the facial image sequence to obtain facial expression feature values; (7) Separating the energy distribution of four wavebands, namely, delta wave, theta wave, alpha wave and beta wave, from the brain wave time series data to obtain brain wave characteristic values; (8) Integrate the physiological characteristic values, the facial expression characteristic values, and the brain wave characteristic values ​​in time series to obtain a sleep cycle characteristic matrix.

[0023] Specifically, the user's physiological data during seven days of continuous sleep is collected through a multi-channel biosensor array, which contains four special sensors: a respiratory monitoring belt, an ECG monitoring electrode, a pulse oximeter, and an electromyography electrode. These sensors are integrated into a portable device that users can wear at home without going to a hospital or sleep center. Respiratory rate data is collected through a chest and abdomen respiratory belt to record the number of breaths per minute and changes in breathing depth; heartbeat data is collected through ECG electrodes attached to the chest to record heart rate and its variability; blood oxygen concentration data is collected through a finger-clip pulse oximeter to monitor changes in oxygen saturation in the blood; and electromyography data is collected through surface electrodes attached to the mandible and legs to record muscle activity status. These data are collected once a second to form a high-temporal resolution raw physiological data stream. At the same time, an infrared imaging device is installed at the head of the bed, which triggers automatic collection every hour during the user's sleep to capture the user's facial image. Infrared technology enables the device to work without interference in a completely dark environment without affecting the user's normal sleep. 8-10 facial images are collected every night, and about 60 images are accumulated in seven days to form a facial image sequence. These images capture the user's facial expressions during different sleep stages, such as rapid eye movements and micro-expressions such as frowning or relaxing the brows.

[0024] The portable EEG detection device uses dry electrode technology. Users only need to wear a headband-like device to record EEG activity in the frontal, temporal and parietal regions. The device collects data every 0.5 seconds to form continuous EEG time series data. These data contain information about the brain's electrical activity in different sleep stages and are the key basis for judging the depth and quality of sleep. The original physiological data often contains various interference signals, such as body motion artifacts, environmental electromagnetic interference and equipment noise. The data is de-noised by digital filters. Specifically, bandpass filtering technology is used to apply 0.1-0.5Hz filtering to respiratory data, 0.5-40Hz filtering to ECG data, and 10-500Hz filtering to EMG data. The filtered data is then smoothed by the moving average method, and the mean of 5 consecutive data points is calculated to replace the center point, thereby suppressing random fluctuations. These processes make the data clearer and more stable, which is convenient for subsequent feature extraction.

[0025] Time domain features and frequency domain features are extracted from the processed physiological data. Time domain features include the mean, standard deviation, maximum and minimum values ​​of heart rate, mean and coefficient of variation of respiratory rate, average level and fluctuation amplitude of blood oxygen saturation, intensity of myoelectric activity and number of bursts. Frequency domain features are obtained through fast Fourier transform (FFT), which converts time domain signals into frequency domain and analyzes the energy distribution of different frequency bands. For example, in the analysis of heart rate variability, the power of very low frequency components (VLF, <0.04Hz), low frequency components (LF, 0.04-0.15Hz) and high frequency components (HF, 0.15-0.4Hz) is extracted, and the LF / HF ratio is calculated to reflect the balance state of sympathetic-parasympathetic nerves. These characteristic values ​​together constitute a multidimensional physiological feature vector.

[0026] The facial image sequence uses computer vision technology to extract the characteristics of facial expression changes. First, the facial key points are located, and 68 feature points such as eyes, eyebrows, and mouth corners are marked. Then the displacement of these points between consecutive images is calculated to capture subtle changes in expression. The facial area grid is defined and the degree of grid deformation is calculated to quantify the intensity of expression. In addition, eye movement is detected through brightness analysis to determine whether it is in the rapid eye movement period. These data form facial expression feature values, which reflect the changes in the subconscious state during sleep.

[0027] The EEG time series data is decomposed into different frequency bands through wavelet transform: delta waves (0.5-4Hz) mainly appear in deep sleep, theta waves (4-8Hz) are common in light sleep, alpha waves (8-12Hz) mostly appear in a relaxed and awake state, and beta waves (12-30Hz) are related to mental activity. The proportion of each waveform in the total energy and the time variation law are calculated to form the EEG characteristic values. These characteristic values ​​are the core indicators for judging the sleep stage and depth.

[0028] Finally, the physiological feature values, facial expression feature values, and brain wave feature values ​​are synchronized and integrated according to the timestamp to create a multidimensional data structure. Each time point has corresponding physiological state, facial expression, and brain wave activity data, forming a sleep cycle feature matrix that fully describes the sleep process. This matrix is ​​the data basis for subsequent analysis of sleep stage distribution and judgment of sleep disorder types.

[0029] In a specific embodiment, the process of executing step S102 may specifically include the following steps: (1) extracting the brain wave feature sequence from the sleep cycle feature matrix, dividing the sleep stage boundaries by a feature threshold segmentation method, and obtaining a sleep stage time distribution map; (2) calculating the time ratio of rapid eye movement sleep to non-rapid eye movement sleep from the sleep stage time distribution diagram to obtain a sleep structure ratio value; (3) Determine the time required from the wakefulness state to the first entry into light sleep according to the sleep stage time distribution diagram to obtain a sleep latency value; (4) Identifying the number of transitions from deep sleep or light sleep to wakefulness during sleep according to the sleep stage time distribution diagram, and obtaining the number of awakenings at night; (5) comparing the sleep structure ratio value with the standard sleep ratio range to obtain a structure abnormality index; (6) Obtaining a sleep disorder score by weighted fusion of the sleep latency value, the number of nighttime awakenings, and the structural abnormality index; (7) Generating a sleep-wake circadian rhythm disorder type determination result based on the sleep disorder score and the temporal pattern of the sleep stage time distribution diagram and in comparison with the sleep disorder classification standard.

[0030] Specifically, the brain wave feature sequence is extracted from the sleep cycle feature matrix. This sequence contains data on the energy distribution of four bands: delta wave, theta wave, alpha wave and beta wave. The feature threshold segmentation method is a technology that divides sleep stages by setting specific thresholds based on the energy ratio of these bands. In specific operations, when the proportion of delta waves exceeds 40%, it is judged as deep sleep (N3 stage), when the proportion of theta waves exceeds 50% and the proportion of delta waves is less than 20%, it is judged as light sleep (N2 stage), when both alpha waves and theta waves have obvious activity and beta waves increase, it is judged as N1 stage light sleep, when the electroencephalogram shows a desynchronized pattern and the facial electromyogram shows decreased muscle tension and rapid eye movement is detected, it is judged as REM sleep, and when beta waves dominate and myoelectric activity increases, it is judged as awake. According to these standards, the whole night sleep data is judged period by period, and a sleep stage time distribution diagram showing the changes of different sleep stages over time is drawn.

[0031] When calculating the time ratio of rapid eye movement (REM) sleep to non-rapid eye movement (NREM, including N1, N2 and N3 stages) sleep from the sleep stage time distribution diagram, firstly, the time periods determined as REM stages throughout the night are accumulated to obtain the total REM duration, and then the time periods determined as N1, N2 and N3 stages are accumulated respectively to obtain the total NREM duration, and the two are divided to obtain the REM / NREM ratio, which should be between 0.2-0.25 in normal sleep. In addition, the proportion of deep sleep (N3) in the total sleep time is calculated, which should usually account for 15-25%. These values ​​together constitute the sleep structure ratio value, which reflects the quality and structural integrity of sleep.

[0032] When determining the sleep latency value based on the sleep stage time distribution diagram, the time interval from the start of the recording to the first time it is judged as any sleep stage (usually stage N1) is quantified. In the specific operation, find the time point when the sleep recording starts, and then move forward along the time axis until the first time point judged as stage N1, calculate the time difference between the two, and get the sleep latency value, which should be less than 30 minutes under normal circumstances. If it exceeds, it indicates difficulty falling asleep. When identifying the number of awakenings at night, look for any sleep stage (N1, N2, N3 or REM) in the sleep stage time distribution diagram to transition to a wakeful state, and count these transition points. Only awakenings lasting more than 30 seconds are counted as an awakening to exclude the influence of minor short-term awakenings. Count the total number of such transitions throughout the night to get the number of awakenings at night, which usually does not exceed 3-5 times in normal sleep.

[0033] When comparing the sleep structure ratio with the standard sleep ratio, the standardized difference scoring method is used. Taking the REM ratio as an example, if the measured value is 15%, the standard range is 20-25%, the deviation is 5-10%, and the standardized score is -1 to -2 points; if the deep sleep ratio is 8%, the standard range is 15-25%, the deviation is 7-17%, and the standardized score is -2 to -3 points. The standardized scores of various sleep structure indicators are added together to form a structural abnormality index. The higher the index, the farther the sleep structure deviates from the normal standard.

[0034] When weighted fusion is performed on the sleep latency value, number of nighttime awakenings, and structural abnormality index, weights are assigned according to their different degrees of impact on sleep quality. Usually, abnormal sleep structure has the greatest impact on health, with a weight of 0.5; frequent awakenings at night are second, with a weight of 0.3; and sleep latency has a relatively small impact, with a weight of 0.2. Multiply the three by the corresponding weights and add them together to obtain a comprehensive sleep disorder score. The score range is usually set to 0-100, with the higher the score, the more severe the sleep disorder. Based on the sleep disorder score and the temporal pattern of the sleep stage time distribution diagram, and in comparison with the sleep disorder classification criteria, the sleep-wake circadian rhythm disorder type determination result is generated. In the specific judgment, not only the size of the disorder score is considered, but also the time pattern characteristics of the sleep stage distribution are analyzed. If difficulty in falling asleep is accompanied by normal late sleep and moderate disorder scores, and the start time of sleep is more than 2 hours later than normal, it is judged as phase delay syndrome; if the start and end time of sleep are too early, accompanied by difficulty in maintaining sleep, it is judged as phase advancement syndrome; if the length of the sleep-wake cycle deviates significantly from 24 hours and is not fixed, it is judged as non-24-hour sleep-wake syndrome; if sleep fragmentation is severe, and the sleep stage transitions are frequent and irregular, it is judged as irregular sleep-wake rhythm.

[0035] In a specific embodiment, the process of executing step S103 may specifically include the following steps: (1) matching the sleep-wake circadian rhythm disorder type determination result with the ultrasound parameter basic library, extracting the parameter range table of the corresponding type, and obtaining the initial parameter range; (2) According to the initial parameter range, the stimulation frequency is segmented and set to a range of 200-500 kHz for phase delay type disorder, a range of 500-800 kHz for phase advance type disorder, and a range of 800-1200 kHz for irregular type disorder, to obtain a frequency parameter value; (3) Based on the frequency parameter value, a safe stimulation intensity range is calculated while taking into account the skull attenuation factor to obtain a stimulation intensity parameter value; (4) According to different sleep cycle regulation targets, set the stimulation duration gradient, select a longer time parameter for deep sleep promotion, and select a shorter time parameter for rapid eye movement sleep regulation, and obtain the stimulation duration parameter value; (5) According to the frequency parameter value and the stimulation intensity parameter value, select a suitable waveform type, select a continuous wave for slow adjustment, select a pulse wave for fast adjustment, and select a modulated wave for mixed adjustment, to obtain a stimulation waveform parameter value; (6) The frequency parameter value, the stimulation intensity parameter value, the stimulation duration parameter value, and the stimulation waveform parameter value are integrated to form an ultrasound stimulation parameter combination, and the stimulation focus position is set according to the rhythm control priority to obtain an ultrasound stimulation plan.

[0036] Specifically, the sleep-wake circadian rhythm disorder type determination result is matched with the ultrasound parameter basic library, which is a data set containing parameter ranges corresponding to various types of sleep disorders. The disorder type is associated with the parameter range table by searching and matching. In specific operations, when the determination result is phase delay syndrome, the phase delay type parameter table is called; when it is phase advance syndrome, the phase advance type parameter table is called; when it is an irregular sleep-wake rhythm, the irregular type parameter table is called. Each parameter table contains a frequency range, an intensity range, a duration range, waveform options, and a focus position priority list, which constitute the initial parameter range.

[0037] According to the initial parameter range, the stimulation frequency is set in segments based on the response characteristics of different types of disorders to ultrasonic frequencies. The 200-500kHz range is selected for phase-delay disorders because lower-frequency ultrasound can more effectively activate the suprachiasmatic nucleus of the hypothalamus, promote melatonin secretion, and help advance the sleep phase; the 500-800kHz range is selected for phase-advance disorders, and medium-frequency ultrasound can moderately inhibit pineal gland activity and delay the peak time of melatonin secretion; the 800-1200kHz range is selected for irregular disorders, and higher-frequency ultrasound helps stabilize the rhythmic activity of the sleep-wake center. Within the selected frequency range, further precise selection is made based on the severity of the patient's sleep disorder. The higher the disorder score, the closer the frequency setting is to the center value of the range, thereby obtaining a specific frequency parameter value.

[0038] When calculating the safe stimulation intensity range based on the frequency parameter value, the skull attenuation factor needs to be considered, which involves a complex physical calculation process. The calculation formula for the safe stimulation intensity is: , in, Indicates the safe stimulation intensity parameter value (W / cm²), Indicates the ultrasound bioeffect threshold intensity (W / cm²), represents the skull ultrasound attenuation coefficient ( ), which is frequency dependent, Indicates the thickness of the skull (cm), and SF indicates the safety factor (usually 1.5-2). With frequency The relationship can be approximately expressed as: , in, and are constants, 0.2 and 0.1 respectively, and p is the frequency exponent, usually 1.1. Obtained from head CT scan. For example, when the frequency parameter is 350kHz and the skull thickness is 0.6cm, the calculation is Approx. ,like 3W / cm², SF is 1.5, then the safe stimulation intensity About 1.03W / cm². Ensure that the stimulation intensity can effectively penetrate the skull to reach the target brain area, but will not exceed the safety threshold to cause tissue damage.

[0039] When setting the stimulation duration gradient for different sleep cycle regulation targets, it is necessary to consider the matching of the regulation speed of ultrasonic stimulation with the target effect. For the promotion of deep sleep, a longer time parameter is selected, usually in the range of 5-10 minutes. Longer continuous stimulation can more thoroughly regulate the activity of the ventrolateral preoptic area of ​​the anterior hypothalamus and enhance the production of slow wave sleep; for rapid eye movement sleep regulation, a shorter time parameter is selected, usually in the range of 1-3 minutes. Short-term precise stimulation can avoid excessive interference with the natural process of REM sleep, and is sufficient to adjust its timing. In addition, in response to the problem of prolonged sleep latency, a medium-length stimulation parameter of 3-5 minutes is set. The determination of the specific duration value also needs to be fine-tuned in combination with individual factors such as the patient's age and weight to finally obtain the stimulation duration parameter value.

[0040] The selection of appropriate waveform types based on the frequency parameter value and the stimulation intensity parameter value is based on the regulatory characteristics of different waveforms on neuronal activity. Continuous waves are selected for slow regulation. Continuous waves provide continuous and stable energy input, which is suitable for neural networks that need to be gradually adjusted, such as the hypothalamic clock center; pulse waves are selected for rapid regulation. Pulse waves provide high peak energy in a short period of time, which is suitable for neuronal groups that need to respond quickly, such as the awakening center; modulated waves are selected for mixed regulation. Modulated waves combine continuity and periodic changes, which are suitable for situations that require stability and rhythmic adjustment at the same time, such as the conversion regulation between multiple sleep stages. Waveform selection also needs to consider frequency and intensity parameters. High-frequency and low-intensity schemes are usually combined with pulse waves, and low-frequency and high-intensity schemes are usually combined with continuous waves to obtain stimulation waveform parameter values. The frequency parameter value, stimulation intensity parameter value, stimulation duration parameter value and stimulation waveform parameter value are combined to form an ultrasonic stimulation parameter combination. The stimulation focus position is set according to the rhythm regulation priority, and the priority is determined by the main symptoms of sleep disorders: difficulty in falling asleep prioritizes the pineal gland area, difficulty in maintaining sleep prioritizes the hypothalamus area, decreased sleep quality prioritizes the prefrontal lobe area, and abnormal wake-sleep transition prioritizes the preoptic area. All parameters are combined into a structured data packet containing stimulation protocol information to obtain an ultrasound stimulation protocol.

[0041] In a specific embodiment, the process of executing step S104 may specifically include the following steps: (1) Obtain the user's head tomography image, extract the skull structure features and brain tissue boundaries, and form a three-dimensional brain structure model; (2) marking the pineal region, hypothalamus region, prefrontal lobe region, and preoptic region boundaries in the three-dimensional brain structure model to obtain a contour map of key brain regions; (3) Based on the outline map of the key brain regions, the geometric center point of each region is calculated to generate a set of initial stimulation point coordinates; (4) optimizing and adjusting the initial stimulation point coordinate set according to the stimulation focus position in the ultrasound stimulation scheme to generate an accurate stimulation point coordinate set; (5) Measure the thickness and density distribution of the user's skull at each stimulation point, calculate the attenuation coefficient of the ultrasonic penetration path, and form a skull sound transmission map; (6) Combining the precise stimulation point coordinate set and the skull acoustic transmission map, calculating the phase delay matrix and energy distribution function of the multi-element ultrasonic transducer to generate spatial positioning data.

[0042] Specifically, a head tomography image of the user is obtained, usually using magnetic resonance imaging (MRI) or computed tomography (CT) technology. The original image data obtained is usually in DICOM format, containing a series of two-dimensional slice images, each with a resolution of 512×512 pixels. These slice images are processed by an image segmentation algorithm to extract the skull structure and brain tissue boundaries. In the specific implementation, an automatic segmentation method based on region growing is used. First, the grayscale threshold intervals of the skull, cerebrospinal fluid, gray matter, and white matter are determined, and these thresholds are used for preliminary segmentation. Then, morphological operations are applied to improve the segmentation results, including opening and closing operations to eliminate noise and fill holes. Finally, these segmentation results are reconstructed in three-dimensional space to form an accurate three-dimensional brain structure model, which is represented by voxels and has a resolution of usually 1×1×1mm³.

[0043] Import the Talairach or MNI standard brain atlas, and then use the elastic deformation algorithm to align the standard atlas with the user's actual brain model. The mutual information maximization criterion is used in the alignment process, and the spatial correspondence between the two is optimized through iterative optimization. After the alignment is completed, the pineal region (located above the posterior end of the third ventricle), the hypothalamus region (located at the bottom of the third ventricle), the prefrontal region (located on the medial side of the frontal bone), and the preoptic area (located in front of the hypothalamus) can be accurately marked on the user model. Each region is defined by a set of boundary points to form a closed surface, which together constitutes the contour map of the key brain regions. When calculating the geometric center point based on the contour map of the key brain regions, the centroid algorithm is applied to each marked region. First, the region is regarded as a homogeneous body, and the coordinates of its geometric center point can be calculated by the weighted average of the coordinates of all voxels in the region. The calculation formula is: , in, Represents the three-dimensional coordinates of the geometric center point of the region, represents all voxels in region R, represents the coordinates of voxel v, The weight value of voxel v is usually set to 1 or assigned different values ​​according to tissue density. In this way, each key brain area obtains a geometric center point, and the collection of these points forms the initial stimulation point coordinate set.

[0044] When optimizing and adjusting the initial stimulation point coordinate set according to the stimulation focus position in the ultrasound stimulation scheme, the matching degree between the regional functional characteristics and the ultrasound parameters needs to be considered. For example, the pineal gland stimulation point is usually slightly offset to the ventral side to more accurately stimulate the melatonin secretion area, and the hypothalamus stimulation point needs to be accurately located to the suprachiasmatic nucleus according to the circadian rhythm regulation target. The optimization adjustment uses a correction algorithm based on the functional MRI activation map, and uses the known functional connection pattern to fine-tune the original geometric center so that the stimulation point corresponds more accurately to the key functional area. In addition, the angle and path accessibility of stimulation from the outside of the skull must be considered, avoiding blood vessels and high-density bone areas, and selecting the path with the least resistance to ultrasound penetration. After these optimization adjustments, an accurate stimulation point coordinate set is formed.

[0045] Measuring the thickness and density distribution of the user's skull at each stimulation point is the basis for calculating ultrasound transmission efficiency. The attenuation coefficient (Hounsfield unit) in CT data can be directly converted to bone density, while MRI data needs to be estimated through the conversion relationship between T1 image signal intensity and bone density. On the determined ultrasound incident path, a voxel-by-voxel scan is performed from the outer surface to the inner surface of the skull, and the density value and path length of each voxel are recorded. The calculation formula for the attenuation coefficient of the skull ultrasound penetration path is: ; in, represents the total attenuation coefficient of the path (dB), Q represents the total number of voxels on the path, represents the material attenuation constant of the qth voxel , represents the density of the qth voxel (g / cm³), represents the propagation distance of ultrasound in the qth voxel (cm), represents the ultrasound stimulation frequency (MHz), Represents the frequency dependence index, usually taking a value of 1.1-1.3. Calculate the attenuation contribution of all voxels on the path, summarize the path attenuation coefficients of each stimulation point, and form a skull acoustic transmission map covering the entire stimulation area. Combined with the precise stimulation point coordinate set and the skull acoustic transmission map, calculate the phase delay matrix and energy distribution function of the multi-element ultrasonic transducer. Multi-element ultrasonic transducers are usually composed of dozens of independently controlled piezoelectric elements, each of which can independently adjust the phase and amplitude. The focusing principle is to control the phase difference of the waves emitted by each element so that the waves are coherently superimposed at the target position to form an energy focus point. The phase delay matrix calculation formula is: , in, represents the phase delay (radians) of the gth element of the transducer relative to the reference element, represents the ultrasonic stimulation frequency (Hz), represents the propagation path length from the gth element to the target stimulation point e (m), represents the propagation path length from the reference element to the target point (m), Represents the speed of sound in tissue (m / s), usually 1540m / s. When calculating the actual phase delay, it is also necessary to make corrections in combination with the skull acoustic transmission spectrum to consider the effect of the skull on the speed of sound. The energy distribution function describes the spatial distribution characteristics of the focused ultrasound field, which is usually approximated by a Gaussian model: , in, represents the energy density at a spatial point (x, y, z), represents the peak energy density, represents the coordinates of the target stimulus point, It represents the focused beam width parameter, which is related to the transducer aperture, frequency and tissue acoustic impedance. The phase delay matrix and energy distribution function together constitute the spatial positioning data, providing precise spatial navigation for ultrasound stimulation.

[0046] In a specific embodiment, the process of executing step S105 may specifically include the following steps: (1) combining the spatial positioning data with the ultrasound stimulation scheme to generate a stimulation target description matrix; (2) Extracting user records with similar sleep-wake circadian rhythm disorders to the current user from the historical user database to form a reference case set; (3) Extracting features of the stimulation feedback results in the reference case set, including sleep structure change features, physiological signal change features, and subjective feeling description features, and constructing a feedback feature vector library; (4) Processing the stimulus target description matrix through a three-layer convolution structure, wherein the first layer uses 32 3×3 convolution kernels to extract low-level features, the second layer uses 64 3×3 convolution kernels to extract intermediate features, and the third layer uses 128 3×3 convolution kernels to extract high-level features. Each convolution layer is followed by a maximum pooling layer and a batch normalization layer to obtain a multi-dimensional stimulus feature map; (5) reducing the dimension of the multi-dimensional stimulation feature map through two fully connected layers, where the first fully connected layer has 256 neurons and the second fully connected layer has 128 neurons, and using the ReLU activation function to map the high-dimensional features to the sleep parameter evaluation space to form a stimulation effect map; (6) Calculating the correlation weight of the stimulation effect mapping and the feedback feature vector library through cosine similarity, weighted fusion is performed to generate sleep improvement prediction indicators, physiological response prediction values ​​and subjective experience expectation scores, and predictive feedback data is obtained.

[0047] Specifically, the spatial positioning data is combined with the ultrasound stimulation scheme to form a stimulation description, and the possible feedback results are predicted through deep learning technology. First, the spatial positioning data is combined with the ultrasound stimulation scheme to construct a stimulation target description matrix, which is a multidimensional data structure containing stimulation parameters, spatial positions and key brain area information. The specific construction process is to integrate the frequency, intensity, duration, waveform and other parameters in the ultrasound stimulation scheme with the coordinate points, phase delay and energy distribution in the spatial positioning data. The rows of the matrix represent different stimulation target points, and the columns represent the various parameter attributes of each point, forming a two-dimensional table usually n×m, where n is the number of stimulation points (usually 4-8) and m is the parameter dimension of each point (usually 15-20). Extracting similar cases from the historical user database is a process based on user similarity matching. The historical user database contains complete records of users who have received ultrasound stimulation treatment in the past, including basic characteristics, types of sleep disorders, treatment plans and effect evaluation. Similarity matching uses a weighted nearest neighbor algorithm. First, the distance between the current user and the historical user on key features is calculated. The key features include sleep-wake circadian rhythm disorder type (weight 0.4), age (weight 0.2), disorder severity (weight 0.3) and basic physiological parameters (weight 0.1). The distance calculation uses normalized Euclidean distance, and the distance value in multidimensional space is calculated after each feature is standardized. From the calculation results, 10-15 user records with the smallest distance are selected to form a reference case set.

[0048] Feature extraction of the stimulation feedback results in the reference case set is the process of converting the original feedback data into a structured feature vector. The characteristics of sleep structure changes include changes in the proportion of deep sleep, changes in the proportion of REM sleep, changes in sleep latency, changes in the number of awakenings at night, and other indicators; the characteristics of physiological signal changes include changes in heart rate variability, changes in breathing regularity, changes in body temperature curves, and other indicators; the subjective feeling description features extract keywords and emotional tendencies from user feedback texts through natural language processing technology and quantify them into numerical features. The extracted feature vector usually contains 20-30 dimensions, each of which represents a feedback indicator. After these feature vectors are standardized, a feedback feature vector library is formed, which provides a data basis for subsequent similarity calculations.

[0049] Processing the stimulus target description matrix through a convolutional neural network is a deep feature extraction process. First, the two-dimensional matrix is ​​converted into a three-dimensional tensor suitable for convolution operations, and the sample diversity is increased through data augmentation. The first layer of convolution uses 32 convolution kernels of size 3×3 to perform sliding window operations on the input data to extract low-level features such as local spatial relationships and basic parameter patterns. After convolution, the ReLU activation function is applied to introduce nonlinearity, and then the feature map size is reduced through a 2×2 maximum pooling layer to extract significant features. At the same time, a batch normalization layer is applied to standardize the feature distribution and accelerate the training process. The second layer of convolution uses 64 3×3 convolution kernels to capture more complex feature combinations and interactions between parameters. It is also activated, pooled and normalized. The third layer of convolution uses 128 3×3 convolution kernels to extract highly abstract feature representations, including the pattern characteristics and potential effects of the overall stimulus scheme. These three layers of convolution gradually extract multi-level feature representations from the raw data, and finally form a multi-dimensional stimulus feature map.

[0050] The process of reducing the dimension of the multi-dimensional stimulus feature map through a fully connected layer is a process of feature mapping and dimensional conversion. The fully connected layer connects all neurons in the previous layer to each neuron in the current layer to achieve the mapping conversion from feature space to target space. The first fully connected layer contains 256 neurons, which receive the flattened convolution feature map input, perform a fully connected transformation with the input features through the weight matrix, and then apply the ReLU activation function to introduce nonlinearity. The second fully connected layer contains 128 neurons, which further compresses the feature dimension, and also applies ReLU activation to output the final low-dimensional feature representation. These two layers of fully connected structures map the high-dimensional convolution feature space to the 128-dimensional sleep parameter evaluation space to form a stimulation effect mapping, which contains the potential relationship encoding from the original stimulation scheme to the expected sleep effect.

[0051] Comparing the stimulus effect mapping with the feedback feature vector library is the core step in generating prediction results. First, the cosine similarity of the stimulus effect mapping and each vector in the feedback feature vector library is calculated. The cosine similarity measures the closeness of the directions of two vectors. The closer the value is to 1, the more similar they are. Then, weights are assigned according to the similarity values, and cases with high similarity receive greater weights. By weighted fusion, the weighted average is calculated to generate the final prediction results, including sleep improvement prediction indicators (such as the percentage of increase in deep sleep, the reduction in sleep latency), physiological response prediction values ​​(such as the degree of improvement in heart rate variability), and subjective experience expectation scores (such as sleep quality satisfaction).

[0052] In a specific embodiment, the process of executing step S106 may specifically include the following steps: (1) After the user receives ultrasound stimulation, the user's sleep cycle data is monitored, and the frequency of sleep stage transitions, the proportion of deep sleep, and the duration of rapid eye movement sleep are extracted to form a sleep quality assessment index; (2) Collect the user's physiological data changes, including heart rate variability, respiratory rhythm, and body temperature curve, and construct a physiological response record; (3) Obtain the user's subjective report of sleep experience, including the difficulty of falling asleep, nighttime awakening, and morning mental state, and generate a subjective feedback form; (4) integrating the sleep quality assessment index, the physiological response record and the subjective feedback form into actual feedback data; (5) calculating a feedback difference matrix by comparing the deviation values ​​of the predicted feedback data and the actual feedback data in each dimension, and generating feedback difference data; (6) According to the deviation direction and amplitude of each parameter in the feedback difference data, the stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform and stimulation focus position in the ultrasonic stimulation scheme are accurately adjusted to obtain a target stimulation scheme.

[0053] Specifically, the personalized plan is continuously improved by accurately comparing the predicted feedback with the actual effect. After the user receives ultrasound stimulation, comprehensive sleep monitoring is required to evaluate the treatment effect. First, the user's sleep cycle data for 3-7 consecutive days is recorded through a portable polysomnography device. These raw data are processed by the sleep staging algorithm. According to the characteristics of brain waves, eye movements and electromyography, sleep is divided into the awake period, N1 period, N2 period, N3 period and REM period. The frequency of sleep stage conversion (that is, the number of sleep stage changes per hour) is calculated. The normal value is about 4-6 times / hour. Too high indicates unstable sleep. At the same time, the proportion of deep sleep (N3 stage) to total sleep time is calculated, which is about 15-25% for normal adults, as well as the duration of rapid eye movement sleep (REM period), which usually accounts for 20-25% of the total sleep time. These three indicators are combined to form a sleep quality assessment index, which objectively reflects the improvement of sleep structure. Collecting the changes in the user's physiological data requires the use of a variety of biosensors. Heart rate variability is an important indicator for evaluating the function of the autonomic nervous system. The heart rate changes for 24 consecutive hours are recorded through an ECG monitoring device, and the standard deviation (SDNN) and low-frequency / high-frequency power ratio (LF / HF) of adjacent heartbeat intervals are calculated. The respiratory rhythm is recorded through a chest and abdomen respiratory belt to analyze the stability of the respiratory frequency and the law of changes in depth. During normal sleep, breathing should be stable and fluctuate regularly. The body temperature curve records the changes in the user's body temperature throughout the day through an attached skin temperature sensor, paying special attention to the slope of the temperature drop before going to bed and the time when the lowest body temperature occurs during sleep. These are important indicators of the operating status of the biological clock. After digital filtering and artifact removal, these physiological data form standardized physiological response records, which are used to intuitively reflect the regulatory effect of ultrasound stimulation on physiological rhythms.

[0054] Obtaining the user's subjective report of sleep experience is another important dimension for evaluating the effectiveness of treatment. User feedback is collected by designing structured sleep diaries and questionnaires. The ease of falling asleep is obtained through the user's reported sleep time and subjective feeling score (1-10 points). The night awakening experience includes the number of awakenings, duration and degree of wakefulness. The mental state in the morning is evaluated through multiple dimensions such as refreshment, concentration and daytime sleepiness. These subjective data are converted into numerical features through text analysis and quantitative scoring to form a subjective feedback table, which reflects the user's perceived improvement in sleep quality.

[0055] When integrating sleep quality assessment indicators, physiological response records, and subjective feedback tables into actual feedback data, data standardization and structural unification are required. First, convert various types of data into a unified numerical range (0-100 points), and then organize them into three categories: physiological indicators, sleep structure indicators, and subjective feeling indicators. Each type of indicator contains multiple specific parameters to form a structured data table. This integration ensures that feedback data of different dimensions can be uniformly compared and processed, laying the foundation for subsequent comparative analysis with predicted feedback.

[0056] By comparing the deviation values ​​of the predicted feedback data and the actual feedback data in each dimension, calculating the feedback difference matrix is ​​the core link of the solution adjustment. The calculation formula of the feedback difference matrix is: , Where D represents the feedback difference matrix, and its dimension is , b is the number of indicator categories, r is the number of parameters for each indicator category; A represents the actual feedback data matrix, which has the same dimension; P represents the predicted feedback data matrix; W represents the indicator weight matrix, which reflects the importance of different parameters; Represents Hadamard product (element-wise multiplication). Furthermore, normalization is introduced to obtain standardized difference values ​​to eliminate the dimension effect; finally, through matrix transformation and feature extraction, the difference matrix is ​​converted into structured feedback difference data, which clearly shows the deviation direction and degree of each parameter. Accurate adjustment of the ultrasound stimulation scheme according to the feedback difference data is a key step in personalized optimization. The adjustment process follows the gradient response principle, that is, the corresponding parameters are adjusted proportionally according to the deviation direction and amplitude. Specifically, for sleep structure deviation, the stimulation frequency and focal position are mainly adjusted; for physiological rhythm deviation, the stimulation intensity and waveform are mainly adjusted; for subjective perception deviation, the stimulation duration and stimulation time are mainly adjusted. For example, when the proportion of deep sleep is lower than expected, the stimulation frequency is reduced by 5-10% and the focal position is more accurately positioned to the ventrolateral preoptic area; when the heart rate variability is not improved enough, the stimulation intensity is appropriately increased by 10-15% and replaced with a pulse waveform; when the subjective difficulty of falling asleep does not improve as expected, the stimulation duration is increased by 2-3 minutes. These adjustments are converted into specific scheme modifications through the parameter mapping matrix to form a more optimized target stimulation scheme.

[0057] The sleep-wake circadian rhythm disorder regulation method based on ultrasound stimulation in the embodiment of the present application is described above. The sleep-wake circadian rhythm disorder regulation device based on ultrasound stimulation in the embodiment of the present application is described below. Figure 2 In the embodiments of the present application, an embodiment of a sleep-wake circadian rhythm disorder regulating device based on ultrasound stimulation includes: The acquisition module is used to collect the user's breathing rate data, heart rate data, blood oxygen concentration data and electromyography data during sleep, obtain the user's facial image, record brain wave data, and construct a sleep cycle feature matrix; An analysis module, used to analyze the user's sleep stage distribution according to the sleep cycle feature matrix, measure the sleep latency and the number of awakenings at night, and generate a sleep-wake circadian rhythm disorder type determination result; A generation module, configured to generate an ultrasound stimulation parameter combination, including stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform, and stimulation focus position, according to the sleep-wake circadian rhythm disorder type determination result, to obtain an ultrasound stimulation plan; A stimulation module, used to determine the stimulation coordinates of key brain regions according to the ultrasound stimulation scheme, including the coordinates of the pineal region, the hypothalamus region, the prefrontal region and the preoptic region, and generate spatial positioning data; A prediction module, configured to perform user feedback prediction based on a convolutional neural network according to the spatial positioning data and the ultrasound stimulation scheme to obtain predicted feedback data; The correction module is used to obtain actual feedback data of the user stimulated by the ultrasonic stimulation scheme, generate feedback difference data according to the predicted feedback data and the actual feedback data, and correct the ultrasonic stimulation scheme according to the feedback difference data to obtain a target stimulation scheme.

[0058] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0059] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0060] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0061] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0062] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0063] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0064] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation, characterized in that: The sleep-wake circadian rhythm disorder regulation method based on ultrasound stimulation comprises: Collect the user's breathing rate data, heart rate data, blood oxygen concentration data, and electromyography data during sleep, obtain the user's facial image, record brain wave data, and build a sleep cycle feature matrix; Analyze the sleep stage distribution of the user according to the sleep cycle characteristic matrix, measure the sleep latency and the number of awakenings at night, and generate a sleep-wake circadian rhythm disorder type determination result; According to the sleep-wake circadian rhythm disorder type determination result, an ultrasound stimulation parameter combination is generated, including stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform and stimulation focal position, to obtain an ultrasound stimulation plan; According to the ultrasound stimulation scheme, determining the stimulation coordinates of key brain regions, including the coordinates of the pineal region, the hypothalamus region, the prefrontal region, and the preoptic region, and generating spatial positioning data; Using the spatial positioning data, performing user feedback prediction based on a convolutional neural network according to the ultrasound stimulation scheme to obtain predicted feedback data; The actual feedback data of the user stimulated by the ultrasound stimulation scheme is obtained, feedback difference data is generated according to the predicted feedback data and the actual feedback data, and the ultrasound stimulation scheme is corrected according to the feedback difference data to obtain a target stimulation scheme.

2. The method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation according to claim 1, characterized in that: The method collects the user's breathing rate data, heart rate data, blood oxygen concentration data and electromyography data during sleep, obtains the user's facial image, records brain wave data, and constructs a sleep cycle feature matrix, including: The multi-channel biosensor array collects the user's breathing rate data, heart rate data, blood oxygen concentration data, and electromyography data during seven days of continuous sleep to obtain raw physiological data; The infrared imaging device collects the user's facial image once every hour during the user's sleep to obtain a facial image sequence; The user's brain wave data during sleep is recorded by a portable brain wave detection device to obtain brain wave time series data; Performing noise elimination and data smoothing processing on the original physiological data to obtain processed physiological data; Extracting time domain features and frequency domain features from the processed physiological data to obtain physiological feature values; Extract facial expression change features from the facial image sequence to obtain facial expression feature values; Separate the energy distribution of four wavebands, namely, delta wave, theta wave, alpha wave and beta wave, from the brain wave time series data to obtain brain wave characteristic values; The physiological characteristic values, the facial expression characteristic values ​​and the brain wave characteristic values ​​are integrated in time series to obtain a sleep cycle characteristic matrix.

3. The method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation according to claim 1, characterized in that: The method of analyzing the sleep stage distribution of the user according to the sleep cycle characteristic matrix, measuring the sleep latency and the number of awakenings at night, and generating a sleep-wake circadian rhythm disorder type determination result includes: Extracting brain wave feature sequences from the sleep cycle feature matrix, dividing the sleep stage boundaries by a feature threshold segmentation method, and obtaining a sleep stage time distribution map; Calculating the time ratio of rapid eye movement sleep to non-rapid eye movement sleep from the sleep stage time distribution diagram to obtain a sleep structure ratio value; The time required from the awake state to the first entry into light sleep is measured according to the sleep stage time distribution diagram to obtain a sleep latency value; Identify the number of transitions from deep sleep or light sleep to wakefulness during sleep according to the sleep stage time distribution diagram to obtain the number of awakenings at night; Comparing the sleep structure ratio value with the standard sleep ratio range to obtain a structure abnormality index; Obtaining a sleep disorder score by weighted fusion of the sleep latency value, the number of nighttime awakenings and the structural abnormality index; According to the sleep disorder score and the time series pattern of the sleep stage time distribution diagram, and in comparison with the sleep disorder classification standard, a sleep-wake circadian rhythm disorder type determination result is generated.

4. The method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation according to claim 1, characterized in that: The method of generating an ultrasound stimulation parameter combination based on the sleep-wake circadian rhythm disorder type determination result, including stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform and stimulation focal position, and obtaining an ultrasound stimulation scheme includes: Matching the sleep-wake circadian rhythm disorder type determination result with the ultrasound parameter basic library, extracting the parameter range table of the corresponding type, and obtaining the initial parameter range; According to the initial parameter range, the stimulation frequency is segmented and set to a range of 200-500kHz for phase delay type disorder, a range of 500-800kHz for phase advance type disorder, and a range of 800-1200kHz for irregular type disorder, to obtain a frequency parameter value; Based on the frequency parameter value, a safe stimulation intensity interval is calculated, and a stimulation intensity parameter value is obtained by taking into account the skull attenuation factor; According to different sleep cycle regulation targets, a stimulation duration gradient is set, a longer time parameter is selected for deep sleep promotion, and a shorter time parameter is selected for rapid eye movement sleep regulation, and the stimulation duration parameter value is obtained; According to the frequency parameter value and the stimulation intensity parameter value, a suitable waveform type is selected, such as a continuous wave for slow adjustment, a pulse wave for fast adjustment, or a modulated wave for mixed adjustment, to obtain a stimulation waveform parameter value; The frequency parameter value, the stimulation intensity parameter value, the stimulation duration parameter value, and the stimulation waveform parameter value are integrated to form an ultrasound stimulation parameter combination, and the stimulation focus position is set according to the rhythm control priority to obtain an ultrasound stimulation plan.

5. The method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation according to claim 1, characterized in that: The method of determining the stimulation coordinates of key brain regions according to the ultrasound stimulation scheme, including the coordinates of the pineal region, the hypothalamus region, the prefrontal region and the preoptic region, and generating spatial positioning data, includes: Obtain the user's head tomography image, extract the skull structure features and brain tissue boundaries, and form a three-dimensional brain structure model; Marking the pineal region, hypothalamus region, prefrontal lobe region and preoptic area boundaries in the three-dimensional brain structure model to obtain a contour map of key brain regions; Based on the outline map of the key brain regions, the geometric center point of each region is calculated to generate a set of initial stimulation point coordinates; According to the stimulation focus position in the ultrasound stimulation scheme, the initial stimulation point coordinate set is optimized and adjusted to generate an accurate stimulation point coordinate set; Measure the thickness and density distribution of the user's skull at each stimulation point, calculate the attenuation coefficient of the ultrasonic penetration path, and form a skull sound transmission map; The precise stimulation point coordinate set and the skull acoustic transmission spectrum are combined to calculate the phase delay matrix and energy distribution function of the multi-element ultrasonic transducer to generate spatial positioning data.

6. The method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation according to claim 1, characterized in that: The method of performing user feedback prediction based on a convolutional neural network according to the ultrasound stimulation scheme using the spatial positioning data to obtain predicted feedback data includes: combining the spatial positioning data with the ultrasound stimulation scheme to generate a stimulation target description matrix; Extracting user records with similar sleep-wake circadian rhythm disorder types as the current user from a historical user database to form a reference case set; Extracting features of the stimulation feedback results in the reference case set, including sleep structure change features, physiological signal change features, and subjective feeling description features, and constructing a feedback feature vector library; The stimulus target description matrix is ​​processed by a three-layer convolution structure, wherein the first layer uses 32 3×3 convolution kernels to extract low-level features, the second layer uses 64 3×3 convolution kernels to extract intermediate features, and the third layer uses 128 3×3 convolution kernels to extract high-level features, and each layer of convolution is followed by a maximum pooling layer and a batch normalization layer to obtain a multi-dimensional stimulus feature map; The multi-dimensional stimulation feature map is subjected to dimensionality reduction processing through two fully connected layers, wherein the first fully connected layer has 256 neurons and the second fully connected layer has 128 neurons, and a ReLU activation function is used to map the high-dimensional features to a sleep parameter evaluation space to form a stimulation effect map; The stimulation effect mapping and the feedback feature vector library are weighted by cosine similarity to generate sleep improvement prediction indicators, physiological response prediction values ​​and subjective experience expectation scores to obtain prediction feedback data.

7. The method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation according to claim 1, characterized in that: The step of obtaining actual feedback data of a user stimulated by the ultrasound stimulation scheme, generating feedback difference data according to the predicted feedback data and the actual feedback data, and modifying the ultrasound stimulation scheme according to the feedback difference data to obtain a target stimulation scheme includes: After the user receives ultrasound stimulation, the user's sleep cycle data is monitored, and the frequency of sleep stage transitions, the proportion of deep sleep, and the duration of rapid eye movement sleep are extracted to form a sleep quality assessment index; Collect the user's physiological data changes, including heart rate variability, respiratory rhythm and body temperature curve, and build a physiological response record; Obtain the user's subjective report of sleep experience, including the difficulty of falling asleep, nighttime awakening, and morning mental state, and generate a subjective feedback form; integrating the sleep quality assessment index, the physiological response record and the subjective feedback form into actual feedback data; By comparing the deviation values ​​of the predicted feedback data and the actual feedback data in each dimension, a feedback difference matrix is ​​calculated to generate feedback difference data; According to the deviation direction and amplitude of each parameter in the feedback difference data, the stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform and stimulation focus position in the ultrasound stimulation scheme are accurately adjusted to obtain a target stimulation scheme.

8. A sleep-wake circadian rhythm disorder regulating device based on ultrasound stimulation, used to implement the sleep-wake circadian rhythm disorder regulating method based on ultrasound stimulation as described in any one of claims 1 to 7, characterized in that: The sleep-wake circadian rhythm disorder regulating device based on ultrasound stimulation comprises: The acquisition module is used to collect the user's breathing rate data, heart rate data, blood oxygen concentration data and electromyography data during sleep, obtain the user's facial image, record brain wave data, and construct a sleep cycle feature matrix; An analysis module, used to analyze the user's sleep stage distribution according to the sleep cycle characteristic matrix, measure the sleep latency and the number of awakenings at night, and generate a sleep-wake circadian rhythm disorder type determination result; A generation module, configured to generate an ultrasound stimulation parameter combination, including stimulation frequency, stimulation intensity, stimulation duration, stimulation waveform, and stimulation focus position, according to the sleep-wake circadian rhythm disorder type determination result, to obtain an ultrasound stimulation plan; A stimulation module, used to determine the stimulation coordinates of key brain regions according to the ultrasound stimulation scheme, including the coordinates of the pineal region, the hypothalamus region, the prefrontal region and the preoptic region, and generate spatial positioning data; A prediction module, configured to perform user feedback prediction based on a convolutional neural network according to the spatial positioning data and the ultrasound stimulation scheme to obtain predicted feedback data; The correction module is used to obtain actual feedback data of the user stimulated by the ultrasonic stimulation scheme, generate feedback difference data according to the predicted feedback data and the actual feedback data, and correct the ultrasonic stimulation scheme according to the feedback difference data to obtain a target stimulation scheme.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, the method for regulating sleep-wake circadian rhythm disorders based on ultrasonic stimulation as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the method for regulating sleep-wake circadian rhythm disorders based on ultrasound stimulation according to any one of claims 1 to 7.

Images