Intelligent ultrasonic body fat management system and method based on full-link data fusion

Abstract

The invention provides an intelligent ultrasonic body fat management system and method based on full-link data fusion, and relates to the technical field of digital health management. According to the system, an ultrasonic intervention module collects physical field coupling data and subcutaneous fat thickness data by utilizing a physical field sensing and in-situ measurement technology; the physiological monitoring module acquires whole body metabolic behavior data by using a physical sign tracking technology; the fusion decision module constructs a full-link fusion data set by using a space-time alignment technology, and identifies a dominant factor through a causal analysis algorithm to generate a feedback control instruction; and the execution response module analyzes the instruction by utilizing a self-adaptive control technology and generates physical field modulation output and interaction strategy window output. Through multi-dimensional data fusion and attribution closed loop, precise management from treatment perception and behavior tracking to software and hardware cooperative regulation is realized, and the problems of process black box and data island in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of digital health management technology, and in particular to an intelligent ultrasound body fat management system and method with end-to-end data fusion. Background Technology

[0002] Non-invasive ultrasonic fat reduction technology is widely used in medical aesthetics and home health management as a physical therapy method. Its working principle is mainly to destroy subcutaneous fat cells using the thermal or cavitation effects of ultrasound. Current body fat management models usually consist of a standalone ultrasound treatment device and a universal weight or body fat measurement device. Users use the ultrasound device for local treatment at specific times, while monitoring changes in overall body fat percentage or weight in daily life using a body fat scale.

[0003] With the development of the Internet of Things (IoT) and digital healthcare, related devices are gradually transforming towards intelligence. Current technological trends include the miniaturization and home-based application of treatment devices, as well as connecting devices to mobile terminals via wireless communication technology to achieve simple on / off control or duration recording. Meanwhile, the widespread adoption of wearable devices such as smart bracelets makes it possible to collect daily activity data such as heart rate and steps, and health management is evolving from single-point treatment to continuous monitoring.

[0004] Existing ultrasound-guided body fat management systems suffer from significant data fragmentation and process opacity, making it difficult to quantify and guarantee treatment efficacy. Specifically, current ultrasound equipment lacks closed-loop monitoring of key treatment parameters such as real-time contact impedance between the probe and skin and actual subcutaneous acoustic power. Treatment effectiveness is highly dependent on operator technique, and the reasons for ineffective treatment cannot be traced. Furthermore, user inputs such as diet and exercise, process interventions in ultrasound treatment parameters, and physiological outputs of changes in fat thickness are scattered across different dimensions and lack spatiotemporal correlation. Users cannot know whether fat reduction is due to dietary control or effective treatment, and the system cannot optimize treatment parameters based on changes in fat thickness, thus failing to form a truly personalized management closed loop. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the purpose of this invention is to provide an intelligent ultrasound body fat management system and method with end-to-end data fusion. This invention solves the problems of lack of closed-loop monitoring of the treatment process and lack of effective fusion of multi-source data in the prior art.

[0006] To achieve the above objectives, the present invention provides the following solution: A smart ultrasound body fat management system with end-to-end data fusion includes: The power module and the ultrasound intervention module, physiological monitoring module, fusion decision-making module and execution response module, all connected to the power module; The power supply module is used to supply power to the ultrasound intervention module, the physiological monitoring module, the fusion decision module and the execution response module; The ultrasound intervention module is used to collect data on the treatment process and results using physical field sensing technology and in-situ measurement technology, and to obtain physical field coupling data and subcutaneous fat thickness data. The physiological monitoring module is used to monitor the user's metabolic behavior using vital sign tracking technology to obtain whole-body metabolic behavior data; The fusion decision module is used to perform vector mapping on the physical field coupling data, the subcutaneous fat thickness data and the whole body metabolic behavior data using spatiotemporal alignment technology to obtain a full-link fusion dataset. It then uses a causal analysis algorithm to perform attribution analysis on the full-link fusion dataset to obtain the dominant factor and determine the corresponding feedback control command based on the dominant factor. The execution response module is used to parse the feedback control command using adaptive control technology, obtain the execution driving parameter set, and determine the corresponding target control output based on the execution driving parameter set. The target control output includes: physical field modulation output and interactive strategy window output.

[0007] A smart ultrasound-based body fat management method with end-to-end data fusion includes: The treatment process and results were collected using physical field sensing technology and in-situ measurement technology to obtain physical field coupling data and subcutaneous fat thickness data. By using vital sign tracking technology to monitor users' metabolic behavior, data on whole-body metabolic behavior can be obtained. Using spatiotemporal alignment technology, vector mapping is performed on the physical field coupling data, the subcutaneous fat thickness data, and the whole-body metabolic behavior data to obtain a full-link fusion dataset. Causal analysis algorithm is then used to perform attribution analysis on the full-link fusion dataset to obtain the dominant factor, and the corresponding feedback control command is determined based on the dominant factor. The feedback control command is parsed using adaptive control technology to obtain the execution driving parameter set, and the corresponding target control output is determined based on the execution driving parameter set. The target control output includes: physical field modulation output and interactive strategy window output.

[0008] The present invention discloses the following technical effects: This invention provides an intelligent ultrasound body fat management system and method with end-to-end data fusion. By constructing an end-to-end data fusion closed loop, this invention significantly solves the pain points of black box treatment process, data silos, and feedback lag in existing technologies. First, the system uses physical field sensing technology to monitor contact impedance and energy output in real time, breaking the "process blind spot" of traditional equipment and achieving quantitative transparency of treatment safety and efficacy. Second, it uses spatiotemporal alignment technology to deeply integrate discrete metabolic behavior, physical intervention, and fat change data, establishing cross-dimensional quantitative correlations and completely breaking down data barriers at each stage. Finally, based on causal analysis algorithms, precise attribution and adaptive feedback can proactively identify the dominant factors affecting efficacy and generate targeted regulatory instructions, upgrading passive general suggestions to real-time personalized closed-loop decision-making, greatly improving the scientific nature and user compliance of body fat management. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of an intelligent ultrasound body fat management system with end-to-end data fusion provided in an embodiment of the present invention.

[0011] Figure label: 1-Power supply module, 2-Ultrasound intervention module, 3-Physiological monitoring module, 4-Fusion decision-making module, 5-Execution response module. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0014] like Figure 1 As shown, this invention provides an intelligent ultrasound body fat management system with end-to-end data fusion, comprising: Power module 1 and ultrasound intervention module 2, physiological monitoring module 3, fusion decision module 4 and execution response module 5, all connected to power module 1; The power module 1 is used to supply power to the ultrasound intervention module 2, the physiological monitoring module 3, the fusion decision module 4 and the execution response module 5; The ultrasound intervention module 2 is used to collect data on the treatment process and results using physical field sensing technology and in-situ measurement technology, and to obtain physical field coupling data and subcutaneous fat thickness data. The physiological monitoring module 3 is used to monitor the user's metabolic behavior using vital sign tracking technology to obtain whole-body metabolic behavior data; The fusion decision module 4 is used to perform vector mapping on the physical field coupling data, the subcutaneous fat thickness data and the whole body metabolic behavior data using spatiotemporal alignment technology to obtain a full-link fusion dataset, and to perform attribution analysis on the full-link fusion dataset using a causal analysis algorithm to obtain the dominant factor and determine the corresponding feedback control command based on the dominant factor. The execution response module 5 is used to parse the feedback control command using adaptive control technology, obtain the execution driving parameter set, and determine the corresponding target control output based on the execution driving parameter set. The target control output includes: physical field modulation output and interactive strategy window output.

[0015] Specifically, regarding hardware integration and comprehensive data acquisition: This embodiment constructs a hardware collaborative network centered on an intelligent power supply. Power module 1 adopts a DC-DC multi-rail regulated topology and BMS management technology to provide the system with suitable high-voltage drive and low-voltage logic power supply, and achieves energy efficiency scheduling based on load characteristics. Building upon this, ultrasound intervention module 2 innovatively integrates a piezoelectric array and a thermistor at the energy emission end. While performing targeted fat reduction, it simultaneously acquires physical field coupling data characterizing energy transfer efficiency and in-situ subcutaneous fat thickness data using physical field sensing and pulse echo detection technologies. Combined with a physiological monitoring module 3 worn on the limb, it continuously tracks the user's activity expenditure and heart rate variability using an accelerometer and PPG photoelectric probe, thereby achieving real-time capture of comprehensive data from physical parameters and physiological structural results during treatment to daily metabolic behavior.

[0016] Intelligent Decision-Making and Two-Way Closed-Loop Control: At the data processing level, the fusion decision module 4 utilizes a unified time base protocol and interpolation model to map the aforementioned heterogeneous data into a spatiotemporally aligned end-to-end fusion dataset. It then applies a causal analysis algorithm to accurately calculate the contribution weight of each variable to the therapeutic effect, thereby identifying dominant factors such as improper operation or low metabolism. The execution response module 5 establishes an adaptive feedback closed loop based on this, generating dual-control outputs through table lookup parsing and PID tuning: on the one hand, it automatically optimizes the physical field emission parameters of the ultrasound transducer using PWM modulation; on the other hand, it renders personalized interactive strategy windows on mobile devices through a GUI engine. Ultimately, this achieves a "hardware and software dual-drive" closed-loop management system, from adaptive calibration of equipment hardware parameters to precise intervention in user behavior, effectively overcoming the shortcomings of traditional treatment processes being black boxes and data silos.

[0017] Furthermore, in this embodiment, when the coupling sensing submodule is running, physical field sensing technology is activated through a piezoelectric sensor array or impedance monitoring circuit integrated on the energy emission end face. High-frequency impedance scanning is performed on the treatment interface during the ultrasonic wave emission gap or by superimposing a low-voltage detection signal. This embodiment collects the acoustic impedance change characteristics between the probe and the skin contact surface in real time to form an interface impedance signal set, and uses a signal demodulation algorithm to extract the amplitude and perform variance analysis on this signal set. This embodiment compares the demodulated real-time impedance value with a preset standard acoustic impedance threshold range for human soft tissue. When the detected impedance value is within the effective coupling range and the fluctuation rate is lower than a preset stable value, it is determined that the current contact state is effective and a quantified contact quality score is generated, thereby establishing the contact quality data and solving the problem of energy transfer loss caused by improper operation.

[0018] In this embodiment, when the thermal monitoring submodule is running, the voltage and current sampling circuit and the thermal sensing unit are simultaneously activated to capture the energy emission process in multiple dimensions. A high-speed analog-to-digital converter records the driving voltage waveform and loop current waveform at both ends of the ultrasonic transducer in real time, and combines this with the end-face temperature gradient data fed back by the thermal element to construct a physical field state signal set. Based on this, this embodiment uses a power integral algorithm to perform time-domain convolution on the voltage and current waveforms to obtain the effective acoustic power value of the actual injected load. Simultaneously, a heat conduction model is used to predict the thermal accumulation trend of the temperature gradient data. This embodiment weights and encapsulates the calculated effective acoustic power value and the predicted thermal accumulation value to finally generate dynamic energy output data containing real-time power parameters and a thermal safety index, achieving precise quantification of the treatment dose.

[0019] In this embodiment, when operating the structural detection submodule, the pulse echo detection logic in in-situ measurement technology drives the ultrasonic transducer to switch to transmit / receive full-duplex mode, transmitting a high-frequency short-pulse detection beam towards the subcutaneous target area. This embodiment utilizes the transducer to receive sound wave signals reflected from different tissue interfaces such as skin, fat, and muscle, and forms a biological tissue echo set through signal amplification and bandpass filtering. Next, this embodiment applies time-frequency analysis technology to identify the time-of-flight characteristics of the echo set, accurately locating the high-echo, strong-reflection interface between the fat and muscle layers. Based on the sound wave propagation speed constant in adipose tissue and the time difference between the reflection interface, the physical depth of the subcutaneous fat layer is calculated, thereby determining the subcutaneous fat thickness data. This completes the real-time in-situ measurement of treatment effects without changing the equipment.

[0020] Furthermore, in this embodiment, when operating the inertial measurement sensing submodule and the photoelectric perfusion detection submodule, the microelectromechanical system inertial sensor unit of the wearable end is first activated to perform high-frequency sampling of the instantaneous displacement acceleration and minute vibration frequencies of the user's limbs in three-dimensional space. After removing gravitational component interference through low-pass filtering, a pure limb acceleration signal set is formed. The gait recognition algorithm and metabolic equivalent formula are then applied to integrate this signal set to output quantitative limb movement data including step count and dynamic calorie consumption. Simultaneously, this embodiment drives a reflective photoelectric sensor to emit green light or infrared beams of specific wavelengths towards the user's skin. A photodiode continuously receives the light intensity reflected from subcutaneous capillaries, which varies with the cardiac cycle. The light intensity changes are converted into analog electrical signals and converted into photoplethysmography (PPG) wave signal sets through analog-to-digital conversion. Then, a peak detection algorithm is used to extract pulse rate interval features, establishing hemodynamic characteristic data including real-time heart rate and heart rate variability indicators.

[0021] This embodiment, in its runtime temporal analysis submodule, aims to address the limitation of a single data source in accurately determining a user's physiological state. It simultaneously retrieves the aforementioned limb acceleration signal set and photoplethysmography (PPG) signal set through a resting activity pattern classifier. This embodiment performs temporal feature coupling analysis on these two heterogeneous signal sets, specifically comparing limb activity and cardiac load levels within the same time window to identify states such as anxiety (low activity / high heart rate) or deep sleep (low activity / low heart rate), thereby generating a continuous user activity temporal map. Furthermore, based on the temporal distribution characteristics of this map, this embodiment utilizes a sleep staging algorithm to define the periods when the user is awake, in light sleep, or in deep sleep, thus accurately generating circadian rhythm phase data characterizing the biological clock state.

[0022] In this embodiment, when running the multi-source data integration submodule, a multi-channel information fusion protocol is initiated to uniformly encapsulate the scattered monitoring results, mapping the aforementioned limb movement quantification data, hemodynamic feature data, and diurnal rhythm phase data into the same data structure. This embodiment utilizes a standardized interface to perform dimensionless processing and timestamp alignment on each data item, constructing a multi-dimensional physiological feature vector group containing dimensions of exercise intensity, cardiopulmonary function, and sleep quality. Finally, this embodiment serializes and encodes this vector group to generate whole-body metabolic behavior data that can be used for subsequent system decisions, achieving a complete transformation from low-level physical signals to high-level physiological semantics.

[0023] Furthermore, in this embodiment, when running the heterogeneous clock synchronization submodule and the feature vector interpolation submodule, time-domain standardization is first performed on the high-frequency millisecond-level sampling characteristics of the physical field coupling data, the minute-level continuous characteristics of the whole-body metabolic behavior data, and the day-level discrete characteristics of the subcutaneous fat thickness data. This embodiment uses a globally unified time base protocol to parse the timestamp header information of each data source, uniformly converting it to Greenwich Mean Time format and constructing a linear and continuous unified time reference axis. Based on this, this embodiment calls a linear regression interpolation model to smoothly fill in and up-map low-frequency or missing data points, precisely anchoring discrete fat thickness measurements and dynamically fluctuating ultrasound parameters and metabolic data to the same time coordinate point. This constructs a highly aligned spatiotemporal synchronization feature matrix set and establishes a full-link fusion dataset, effectively overcoming the technical obstacle of being unable to jointly analyze multi-source heterogeneous data due to inconsistent sampling frequencies.

[0024] In this embodiment, when running the correlation dimension calculation submodule and the dominant weight ranking submodule, a multivariate statistical analysis model is initiated to deeply mine the entire fused dataset. This embodiment defines the user's dietary calorie intake, exercise metabolic expenditure, and ultrasound treatment dose as behavioral input variables, and the rate of change in subcutaneous fat thickness as a physiological outcome variable. Covariance calculation and regression analysis are used to quantify the linear correlation between each input variable and the outcome variable, generating a correlation coefficient matrix containing the weight values ​​of each variable, and determining the efficacy contribution accordingly. Subsequently, this embodiment uses a bubble sort algorithm or quicksort logic to numerically rank and descend the efficacy contribution of all variables, generating a priority list of influencing factors, and locking the variable with the highest weight or the strongest correlation in the list as the dominant factor causing the current change in efficacy.

[0025] In this embodiment, when running the intervention rule index submodule, a pre-set expert decision lookup table stored in non-volatile memory is activated. This table contains a complete mapping logic from the root cause of the problem to the solution. This embodiment uses the identified dominant factor as the index key, traversing the lookup table to search and match the corresponding intervention logic path. For example, when the dominant factor points to poor ultrasound probe contact, this embodiment matches a combination of hardware parameter compensation and user prompts; when the dominant factor points to a decrease in basal metabolic rate, this embodiment matches a suggested logic for increasing exercise intensity. This embodiment converts the matched path into a machine-readable set of control strategy codes and ultimately parses and generates feedback control instructions containing specific physical parameter adjustment values ​​or interactive text content, achieving an automated closed loop from data attribution to precise decision-making.

[0026] Specifically, in this embodiment, during the operation of the fusion decision module 4, in order to solve the problem of inconsistent sampling frequencies between ultrasound treatment data (high frequency), physiological monitoring data (continuous), and body fat measurement data (low frequency), a linear regression interpolation model is used for data standardization. The specific implementation logic of this model is as follows: The system first determines the interpolation target time based on the globally unified time base protocol. This parameter is the specific time point on the unified time base axis that needs to be filled (e.g., every hour).

[0027] To calculate the value at that moment, the system retrieves two key reference points from the asynchronous data stream uploaded by the original sensors: one is the original sampling timestamp immediately preceding the interpolation target time and its corresponding original feature observations (sourced from historical data actually collected by the sensors); the other is the original sampling timestamp immediately following the interpolation target time and its corresponding original feature observations. The model's computational logic is to use the observation value of the previous moment as a base and add a change increment; this change increment is obtained by multiplying the "data change slope" by the "time interval". The "data change slope" is calculated by subtracting the observation value of the previous moment from the observation value of the later moment, and dividing by the difference between the timestamps of the later and earlier moments; it characterizes the rate of change of the data between the two sampling points. The "time interval" is the difference between the interpolation target time and the timestamp of the previous moment. Through this calculation, the system ultimately obtains the synchronous feature values, i.e., standardized data mapped onto a unified time axis, completing the spatiotemporal alignment of the data.

[0028] After data alignment, this embodiment utilizes a multivariate statistical analysis model to deeply mine the entire fused dataset to identify the dominant factors influencing fat loss. This model essentially performs covariance standardization to derive a correlation coefficient, a quantitative indicator measuring the linear correlation between a specific behavior (such as diet, exercise, or treatment parameters) and the final therapeutic effect (fat reduction). The coefficient typically ranges from -1 to +1, with a value closer to one indicating a greater influence. The specific computational process is as follows: The system first traverses the entire fused dataset, extracting the total number of valid sample pairs participating in the computation. This determines the sample size and confidence level for the statistical analysis.

[0029] For each sample point, the system acquires the values ​​of specific behavioral or process variables (e.g., daily steps or treatment power) and the corresponding subcutaneous fat thickness changes. Simultaneously, the system pre-calculates the arithmetic mean of the behavioral variable and the subcutaneous fat thickness change over the statistical period, serving as a baseline for measuring data fluctuation. During calculation, the system first calculates the deviation of the behavioral variable value from its mean for each sample point, multiplies it by the deviation of the fat thickness change value from its mean, and sums this product across all samples to obtain a covariance term reflecting the co-variance trend. Subsequently, the system calculates the square root of the sum of squares of the behavioral variable deviations and the square root of the sum of squares of the fat thickness change deviations across all samples; these two values ​​represent the degree of dispersion of the two variables, respectively. Finally, the system divides the covariance term by the product of these two dispersion values ​​to eliminate the influence of dimensions, obtaining the final correlation coefficient.

[0030] Furthermore, in this embodiment, when running the instruction protocol parsing submodule and the parameter adaptive tuning submodule, it first receives feedback control instructions from the decision layer and calls the control command code table stored in non-volatile memory. This embodiment maps high-level semantic instructions to a sequence of binary operation control frames recognizable by the underlying hardware through a table lookup method, thereby establishing the basic logic instruction set. Subsequently, this embodiment starts the PID closed-loop controller, introducing the current device state feedback value as a correction variable, and performs dynamic gain calculation on the target parameters in the basic logic instruction set. This embodiment uses the proportional-integral-derivative algorithm to automatically calculate the optimal control step size and response rate, generating an execution drive parameter set containing precise power values ​​and interactive text encoding, thereby ensuring that the system can suppress overshoot oscillations and achieve rapid convergence during control execution.

[0031] In this embodiment, when the energy field drive submodule is running, the power adjustment parameters from the execution drive parameter set are extracted and input to the pulse width modulation generator. Based on the target power value, this embodiment calculates the corresponding pulse duty cycle value using an internal clock counter, converting the digitized power command into a continuously varying analog drive signal in the time domain. This embodiment further utilizes a power amplifier circuit to perform voltage amplitude modulation and impedance matching on this analog drive signal, directly driving the ultrasonic transducer to change its mechanical vibration amplitude, thereby forming a physical field modulation output at the treatment head end face. This process achieves microsecond-level fine adjustment of the ultrasonic energy intensity, ensuring that the energy density actually injected into the human tissue strictly follows the system's control strategy.

[0032] In this embodiment, when running the visualization rendering submodule, text prompts and strategy type codes are extracted simultaneously from the execution driver parameter set, and the mobile terminal's graphical user interface engine is activated. This embodiment adaptively typesets and assembles components for the text information according to a preset UI layout template, transforming the abstract data stream into an interactive view data package containing text descriptions, dynamic charts, and operation guidance animations. This embodiment uses the graphics rendering pipeline to instantiate this data package into a visual floating window or full-screen pop-up interface, i.e., the interactive strategy window output. This output is directly presented at the top layer of the user terminal screen, providing intuitive visual feedback to the user on the current treatment status or behavioral intervention suggestions, achieving real-time mapping from the device's underlying control logic to the user's perception level.

[0033] Specifically, in this embodiment, when running the digital duty cycle mapping unit, it first receives the execution drive parameter set generated by the upper-level control algorithm and calls the numerical mapping protocol stored in the microcontroller. This embodiment uses this protocol to parse the parameter set, extract the target power value representing the expected output intensity, and calculates it based on a preset power-duty cycle linear conversion formula or lookup table curve. This embodiment converts the target power value into the corresponding conduction time ratio and calculates a precise count value in conjunction with the hardware clock frequency, thereby determining the digital duty cycle instruction. This instruction clearly defines the ratio of the high-level duration to the period width of the subsequent pulse waveform, completing the digital quantization of the energy control logic.

[0034] In this embodiment, when running the analog waveform synthesis unit, the aforementioned digital duty cycle instruction is written into the comparison register of the pulse width modulation generator, and an internal high-frequency counter is started for time-domain modulation. This embodiment uses the counter's count value to compare with a preset threshold in the register in real time. When the count value is less than the threshold, a high level is output; otherwise, a low level is output, thereby synthesizing a basic pulse sequence with a constant frequency but variable pulse width at the I / O port. This embodiment uses a signal shaping circuit to correct the steepness of the rising and falling edges of this pulse sequence, eliminating glitches and interference, thus determining a standard analog drive signal and realizing the physical conversion from digital logic instructions to low-voltage analog electrical signals.

[0035] In this embodiment, when the acoustic-to-electric conversion excitation unit is running, a low-voltage analog drive signal is input to a high-voltage gain circuit (such as an H-bridge power amplifier stage or a Class D amplifier). This embodiment utilizes the high-speed switching characteristics of power MOSFETs to apply the DC high voltage provided by power module 1 to the output terminal, significantly amplifying the analog drive signal to generate a high-energy drive waveform with peak-to-peak values ​​reaching tens or even hundreds of volts. This embodiment uses an impedance matching network to apply this high-energy waveform to the piezoelectric ceramic terminals of the ultrasonic transducer, utilizing the inverse piezoelectric effect to drive the ceramic to generate synchronous mechanical vibrations, thereby determining the physical field modulation output and ultimately efficiently converting electrical energy into ultrasonic physical field energy capable of penetrating human tissue.

[0036] More specifically, in this embodiment, when running the information decoupling and extraction unit, the parameter parsing protocol is first invoked to perform deep traversal and field separation operations on the received execution-driven parameter set. This embodiment decouples non-display control instructions from display text information in the parameter set, extracting a clean feedback text stream and identifying its metadata attributes, such as urgency, text type, and numerical precision. Based on these metadata attributes, this embodiment instantiates the feedback text stream into independent display-level content elements. For example, high-priority warning information is marked as a modal pop-up element, and regular suggestion information is marked as a floating bubble element, thus completing the initial separation of data content and business logic.

[0037] In this embodiment, when running the layout topology calculation unit, the aforementioned display-level content elements are input into the graphical user interface engine, and an adaptive layout algorithm is activated for component matching. Based on the current screen size and resolution, and the orientation (portrait or landscape), this embodiment dynamically calculates the relative coordinates and hierarchical relationships of each content element within the view container, constructing a UI component topology tree containing parent-child node references. This embodiment traverses each node of this topology tree, determining its corresponding control type and spatial constraints, thereby generating a visual layout skeleton that does not contain specific styles but possesses complete structural information, ensuring that interface elements maintain the correct logical arrangement structure across different terminal devices.

[0038] In this embodiment, when running the view encapsulation generation unit, a pre-built visual style library is loaded and rendering serialization technology is initiated. This embodiment binds style attributes to each component node in the visual layout skeleton, specifically including filling background color values, setting font size, configuring rounded corners, and attaching interactive event listeners. This embodiment serializes the rendered view structure into an interactive view data package in binary or JSON format and pushes it to the operating system's display buffer. Finally, it determines the interaction strategy window output, allowing the user to see a fully rendered and responsive graphical interface on the top layer of the screen.

[0039] Furthermore, in this embodiment, when the energy storage management submodule is running, the BMS chip integrated on the main control circuit board is activated to perform millisecond-level polling of the physical state of the built-in energy storage medium. This embodiment utilizes a high-precision current sampling resistor and a voltage sensing probe to read the charging and discharging current values, terminal voltage, and cell surface temperature of the battery pack in real time. It then calculates a set of state-of-the-art parameters, including the remaining charge percentage, battery health, and cycle count, using the open-circuit voltage method combined with the ampere-hour integration method from the internal algorithm library. This embodiment logically compares this parameter set with preset safety thresholds. Once overcharging, over-discharging, or over-temperature anomalies are detected, the metal-oxide-semiconductor field-effect transistor in the main power supply circuit is immediately disconnected, thereby determining the on / off state of the power supply link and ensuring the operational safety of the equipment at the physical level.

[0040] In this embodiment, when operating the multi-track voltage regulator submodule, a DC-DC buck-boost converter topology circuit is used to handle a wide range of floating input voltages from the energy storage medium. This embodiment controls the high-frequency on / off of the power switching transistors, and, in conjunction with the energy storage inductor and filter capacitor, temporarily stores and releases electrical energy, performing independent multi-track regulation of the input voltage. This embodiment generates high-voltage drive levels for the ultrasonic transducer, low-noise analog voltages for the analog signal chain, and standard core voltages for the digital logic chips, thereby obtaining a sequence of logic voltage levels. This embodiment performs ripple suppression and load regulation calibration on each output, ultimately determining independent power supply voltages for the ultrasonic intervention module 2, physiological monitoring module 3, fusion decision module 4, and execution response module 5, achieving power isolation and stable supply between different functional units.

[0041] In this embodiment, when running the power consumption scheduling and allocation submodule, a system-level energy efficiency monitoring daemon is established using dynamic voltage and frequency adjustment logic. This embodiment periodically samples the current workload of each functional module through a bus polling mechanism, for example, detecting whether the ultrasound module is in a high-power transmission state or whether the Bluetooth module is in an idle state, thereby obtaining a power load characteristic value reflecting real-time energy consumption requirements. Based on this characteristic value, this embodiment dynamically adjusts the main processor's operating clock frequency and core power supply voltage, automatically reducing the frequency under low load to reduce ineffective energy consumption, or issuing a command to suspend unnecessary peripherals and enter a low-power mode after idle timeout, thereby determining the system's sleep or wake-up energy efficiency control strategy and maximizing device battery life while ensuring performance responsiveness.

[0042] Specifically, this application scenario targets a 35-year-old sedentary male user for deep abdominal fat reduction. During a 20-minute ultrasound treatment, this embodiment uses the in-situ measurement technology of the ultrasound intervention module 2 to measure the initial thickness of the subcutaneous fat in the user's abdomen as 28mm. Simultaneously, the coupling sensing submodule collected interface impedance signals, showing multiple high impedance abrupt changes exceeding 1500Ω, indicating poor contact between the probe and the skin. At the same time, the physiological monitoring module 3 synchronously transmitted the user's whole-body metabolic behavior data over the past 24 hours. The specific values ​​showed that his average daily steps were only 3200 steps (far below the standard value), and his deep sleep duration was less than 0.8 hours. Photoplethysmography data showed that his stress index remained consistently high in the 85-point range.

[0043] This embodiment utilizes the fusion decision module 4 to perform spatiotemporal alignment and attribution analysis on the aforementioned heterogeneous data. The system maps 20 minutes of treatment process data to a unified time axis and finds a strong positive correlation of 0.92 between the high impedance mutation moment and the drop in energy output data (power decreasing from 10W to 3W). Simultaneously, the multivariate statistical analysis model calculates a correlation coefficient of 0.78 between "increased cortisol levels due to sleep deprivation" and "lag in fat thickness change rate." Based on this, this embodiment uses the dominant weight ranking submodule to determine that the dominant factor causing the current poor single treatment effect is "operational-level coupling loss," while the secondary dominant factor hindering long-term efficacy is "metabolic-level sleep deprivation."

[0044] Based on the above analysis results, this embodiment immediately executes a dual-track closed-loop response strategy. At the hardware control level, the execution response module 5 calculates compensation parameters through the parameter adaptive tuning submodule. When the driving energy field driving submodule detects that the impedance has returned to normal, it temporarily increases the driving voltage of the ultrasonic transducer by 15% and maintains it for 30 seconds to compensate for the dose loss caused by poor contact in the early stage. At the interaction level, the visualization rendering submodule outputs a tactile guidance window with the message "Probe suspension detected, please increase the pressure by 0.5kg" in real time on the user's APP, and simultaneously generates a personalized strategy card with the message "It is recommended to go to bed 1 hour earlier tonight to reduce cortisol resistance". This achieves precise management of the entire chain from millisecond-level hardware compensation to lifestyle intervention.

[0045] This embodiment also provides an intelligent ultrasound body fat management method based on end-to-end data fusion, including: The treatment process and results were collected using physical field sensing technology and in-situ measurement technology to obtain physical field coupling data and subcutaneous fat thickness data. By using vital sign tracking technology to monitor users' metabolic behavior, data on whole-body metabolic behavior can be obtained. Using spatiotemporal alignment technology, vector mapping is performed on the physical field coupling data, the subcutaneous fat thickness data, and the whole-body metabolic behavior data to obtain a full-link fusion dataset. Causal analysis algorithm is then used to perform attribution analysis on the full-link fusion dataset to obtain the dominant factor, and the corresponding feedback control command is determined based on the dominant factor. The feedback control command is parsed using adaptive control technology to obtain the execution driving parameter set, and the corresponding target control output is determined based on the execution driving parameter set. The target control output includes: physical field modulation output and interactive strategy window output.

[0046] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0047] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A smart ultrasound body fat management system with end-to-end data fusion, characterized in that, include: The power module and the ultrasound intervention module, physiological monitoring module, fusion decision-making module and execution response module, all connected to the power module; The power supply module is used to supply power to the ultrasound intervention module, the physiological monitoring module, the fusion decision module and the execution response module; The ultrasound intervention module is used to collect data on the treatment process and results using physical field sensing technology and in-situ measurement technology, and to obtain physical field coupling data and subcutaneous fat thickness data. The physiological monitoring module is used to monitor the user's metabolic behavior using vital sign tracking technology to obtain whole-body metabolic behavior data; The fusion decision module is used to perform vector mapping on the physical field coupling data, the subcutaneous fat thickness data and the whole body metabolic behavior data using spatiotemporal alignment technology to obtain a full-link fusion dataset. It then uses a causal analysis algorithm to perform attribution analysis on the full-link fusion dataset to obtain the dominant factor and determine the corresponding feedback control command based on the dominant factor. The execution response module is used to parse the feedback control command using adaptive control technology, obtain the execution driving parameter set, and determine the corresponding target control output based on the execution driving parameter set. The target control output includes: physical field modulation output and interactive strategy window output.

2. The intelligent ultrasound body fat management system with end-to-end data fusion as described in claim 1, characterized in that, The physical field coupling data includes: Contact quality data and energy output data.

3. The intelligent ultrasound body fat management system with end-to-end data fusion according to claim 2, characterized in that, The ultrasound intervention module includes: Coupled sensing submodule, thermal monitoring submodule, and structural detection submodule; The coupling sensing submodule is used to sample the contact state of the treatment interface in real time using the physical field sensing technology, obtain the interface impedance signal set, and determine the contact quality data based on the interface impedance signal set. The thermal monitoring submodule is used to synchronously capture power fluctuations and thermal effects during the energy emission process, obtain a set of physical field state signals, and determine the energy output data based on the set of physical field state signals. The structure detection submodule is used to perform structural scanning of the subcutaneous target area using the in-situ measurement technology, obtain biological tissue echo sets, and determine the subcutaneous fat thickness data based on the biological tissue echo sets.

4. The intelligent ultrasound body fat management system with end-to-end data fusion as described in claim 1, characterized in that, The physiological monitoring module includes: The system includes an inertial measurement sensing submodule, an optoelectronic injection detection submodule, a state timing analysis submodule, and a multi-source data integration submodule. The inertial measurement and sensing submodule is used to perform three-axis inertial capture of the user's limb spatial displacement and vibration frequency, obtain a set of limb acceleration signals, and determine limb motion quantification data based on the set of limb acceleration signals. The photoelectric perfusion detection submodule is used to optically sample the changes in blood volume pulsation of the user's subcutaneous capillaries using a reflective photoelectric transceiver probe, obtain a photoplethysmography (PPG) signal set, and determine hemodynamic characteristic data based on the PPG signal set. The state-time analysis submodule is used to perform time-domain feature coupling analysis on the limb acceleration signal set and the photoplethysmography pulse wave signal set using a resting-activity pattern classifier, to obtain a user activity time series map and determine the diurnal rhythm phase data based on the user activity time series map; The multi-source data integration submodule is used to encapsulate the limb motion quantification data, the hemodynamic feature data and the diurnal rhythm phase data into data packets using a multi-channel information fusion protocol, to obtain a physiological feature vector group, and to determine the whole-body metabolic behavior data based on the physiological feature vector group.

5. The intelligent ultrasound body fat management system with end-to-end data fusion according to claim 1, characterized in that, The fusion decision module includes: Heterogeneous clock synchronization submodule, feature vector interpolation submodule, correlation dimension calculation submodule, dominant weight ranking submodule and intervention rule index submodule; The heterogeneous clock synchronization submodule is used to perform timestamp standardization processing on the physical field coupling data, the subcutaneous fat thickness data and the whole body metabolic behavior data using a global unified time base protocol, to obtain a standard time series set and determine a unified time reference axis based on the standard time series set. The feature vector interpolation submodule is used to map the physical field coupling data, the subcutaneous fat thickness data, and the whole-body metabolic behavior data onto the unified time reference axis based on the linear regression interpolation model, to obtain a synchronous feature matrix set, and to determine the full-link fusion dataset based on the synchronous feature matrix set. The correlation dimension calculation submodule is used to perform covariance calculation on the behavioral variables and physiological outcome variables in the full-link fusion dataset based on a multivariate statistical analysis model, to obtain a correlation coefficient matrix, and to determine the efficacy contribution of each variable based on the correlation coefficient matrix. The dominant weight ranking submodule is used to perform numerical classification and screening of the effectiveness contribution, obtain an impact factor priority list, and determine the dominant factor according to the impact factor priority list. The intervention rule index submodule is used to perform logical path matching on the dominant factor using a preset expert decision lookup table, obtain the regulation strategy code set, and determine the feedback control instruction based on the regulation strategy code set. The expression for the linear regression interpolation model is: ； in, This is the synchronization feature value mapped to the target time on the unified time reference axis; and These are the original feature observations of two adjacent sampling points located before and after the target time in the original asynchronous data stream, respectively; and These are the original sampling timestamps corresponding to the original feature observations; The interpolation target time is determined on the unified time reference axis according to the globally unified time base protocol; The expression for the multivariate statistical analysis model is: ； in, The calculated correlation coefficient; For the first in the full-link fusion dataset The numerical values ​​of specific behavioral or process variables after alignment at each time point; For the first The numerical values ​​of subcutaneous fat thickness changes after alignment at each time point; This is the arithmetic mean of the behavioral variable over the statistical period; This is the arithmetic mean of the subcutaneous fat thickness change over the statistical period; This represents the total number of valid sample pairs participating in the computation within the full-link fusion dataset.

6. The intelligent ultrasound body fat management system with end-to-end data fusion according to claim 1, characterized in that, The execution response module includes: The module includes a command protocol parsing submodule, a parameter adaptive tuning submodule, an energy field driving submodule, and a visualization rendering submodule. The instruction protocol parsing submodule is used to perform binary translation of the feedback control instruction using a preset control command code table, to obtain an operation control frame sequence, and to determine the basic logic instruction set based on the operation control frame sequence. The parameter adaptive tuning submodule is used to perform dynamic gain calculation on the basic logic instruction set using a PID closed-loop controller to obtain the execution drive parameter set. The energy field driving submodule is used to perform digital-to-analog conversion on the power parameters in the execution driving parameter set using a pulse width modulation generator to obtain an analog driving signal and determine the physical field modulation output based on the analog driving signal. The visualization rendering submodule is used to generate a graphic layout of the text information in the execution driver parameter set using a graphical user interface engine, to obtain an interactive view data package, and to determine the interactive strategy window output based on the interactive view data package.

7. The intelligent ultrasound body fat management system with end-to-end data fusion according to claim 6, characterized in that, The energy field driving submodule includes: Digital duty cycle mapping unit, analog waveform synthesis unit, and acoustic-to-electrical conversion excitation unit; The digital duty cycle mapping unit is used to extract features from the execution drive parameter set using a numerical mapping protocol, obtain a target power value, and determine a digital duty cycle instruction based on the target power value. The analog waveform synthesis unit is used to perform time-domain modulation of the digital duty cycle command using the pulse width modulation generator to obtain a basic pulse sequence and determine the analog drive signal based on the basic pulse sequence. The acoustic-electric conversion excitation unit is used to amplify the amplitude of the analog drive signal using a high-voltage gain circuit to obtain a high-energy drive waveform and determine the physical field modulation output based on the high-energy drive waveform.

8. The intelligent ultrasound body fat management system with end-to-end data fusion according to claim 6, characterized in that, The visualization rendering submodule includes: Information decoupling and extraction unit, layout topology calculation unit, and view encapsulation generation unit; The information decoupling and extraction unit is used to perform field separation on the execution driver parameter set using a parameter parsing protocol, obtain a feedback text stream, and determine the display-level content elements based on the feedback text stream. The layout topology calculation unit is used to perform component matching on the display-level content elements using the adaptive typesetting algorithm in the graphical user interface engine, to obtain a UI component topology tree and to determine the visual layout skeleton based on the UI component topology tree. The view encapsulation generation unit is used to bind style attributes to the visual layout skeleton using rendering serialization technology, obtain the interactive view data package, and determine the interactive strategy window output based on the interactive view data package.

9. The intelligent ultrasound body fat management system with end-to-end data fusion according to claim 1, characterized in that, The power module includes: Energy storage management submodule, multi-track voltage regulation and conversion submodule, and power consumption scheduling and allocation submodule; The energy storage management submodule is used to monitor the charging and discharging status of the built-in energy storage medium in real time using the BMS chip, obtain a set of power status parameters, and determine the on / off status of the power supply link based on the set of power status parameters. The multi-track voltage regulator submodule is used to perform multi-track regulation of the input voltage using a DC-DC buck-boost converter topology circuit to obtain a sequence of logic voltage levels and determine the independent power supply voltage for the ultrasound intervention module, the physiological monitoring module, the fusion decision module and the execution response module based on the logic voltage level sequence. The power consumption scheduling and allocation submodule is used to poll and sample the load demand of each module using dynamic voltage and frequency adjustment logic, obtain power consumption load characteristic values, and determine the energy efficiency control strategy for system hibernation or wake-up based on the power consumption load characteristic values.

10. A smart ultrasound-based body fat management method with end-to-end data fusion, characterized in that, include: The treatment process and results were collected using physical field sensing technology and in-situ measurement technology to obtain physical field coupling data and subcutaneous fat thickness data. By using vital sign tracking technology to monitor users' metabolic behavior, data on whole-body metabolic behavior can be obtained. Using spatiotemporal alignment technology, vector mapping is performed on the physical field coupling data, the subcutaneous fat thickness data, and the whole-body metabolic behavior data to obtain a full-link fusion dataset. Causal analysis algorithm is then used to perform attribution analysis on the full-link fusion dataset to obtain the dominant factor, and the corresponding feedback control command is determined based on the dominant factor. The feedback control command is parsed using adaptive control technology to obtain the execution driving parameter set, and the corresponding target control output is determined based on the execution driving parameter set. The target control output includes: physical field modulation output and interactive strategy window output.

Images