Non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis

Abstract

This invention discloses a non-invasive real-time monitoring and intervention system for lower extremity deep vein thrombosis (DVT). The system comprises an impedance monitoring module, a microcurrent stimulation module, and a data analysis and display module, employing a closed-loop "sensing-analysis-intervention" architecture. The impedance monitoring module utilizes a tri-frequency bioelectrical impedance analysis system with a 4-site, 8-electrode distributed array to accurately distinguish between blood flow stasis and tissue edema. The microcurrent stimulation module provides biphasic constant current dual-mode stimulation, adapting to both endothelial dysfunction and muscle pump insufficiency. The data analysis module runs an embedded lightweight random forest algorithm for real-time risk assessment and adaptively adjusts stimulation parameters based on impedance feedback. This invention integrates non-invasive continuous monitoring, intelligent early warning, and personalized active intervention. It is highly safe, comfortable to wear, and cost-effective, making it suitable for high-risk DVT populations such as postoperative patients, bedridden individuals, and those traveling long distances. It possesses promising prospects for clinical translation and industrialization.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and more specifically to a non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis. Background Technology

[0002] Deep vein thrombosis (DVT) of the lower extremities is a common and critical complication caused by abnormal blood clotting within the deep veins of the lower extremities, leading to obstruction of venous return. Thrombus detachment can cause fatal pulmonary embolism (PE). Early monitoring, risk warning, and timely intervention of DVT have irreplaceable clinical value. Current DVT detection and prevention technologies have the following limitations:

[0003] 1. Limitations of detection methods: Ultrasound is the gold standard for diagnosis, but it cannot be continuously monitored at the bedside, relies on professional physicians, is costly, and has no early warning capability; Venography is an invasive examination with risks of contrast agent allergy and radiation; D-dimer requires repeated blood sampling, has low specificity, and cannot achieve dynamic tracking.

[0004] 2. Passive and limited preventative devices: Intermittent inflation devices (IPC) and gradient compression stockings only provide passive physical protection, lack real-time monitoring and risk assessment functions, are bulky, noisy, and uncomfortable, and their parameters are fixed and cannot be adjusted individually.

[0005] 3. Drug prophylaxis poses safety risks: Anticoagulants carry a risk of bleeding, requiring regular monitoring of coagulation function. They are not suitable for postoperative patients or patients with bleeding tendencies, and lack a feedback regulation mechanism.

[0006] 4. Disconnect between monitoring and intervention: Existing bioelectrical impedance technology only provides static amplitude information, with low signal-to-noise ratio and single feature. It cannot distinguish between blood stasis and tissue edema, and it does not have the ability to actively intervene, thus failing to form a closed-loop management of "sensing-analysis-early warning-intervention-feedback".

[0007] In summary, existing technologies cannot simultaneously meet the clinical needs of non-invasive continuous monitoring, high-precision identification, real-time early warning, adaptive intervention, and safety and comfort. There is an urgent need for an integrated, wearable, and intelligent closed-loop monitoring and intervention system. Summary of the Invention

[0008] The purpose of this invention is to provide a stable, reliable, highly accurate, and safe non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis in order to overcome the shortcomings of existing technologies.

[0009] The present invention achieves the above objectives by adopting the following technical solution: a non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis, characterized in that it includes an impedance monitoring module, a microcurrent stimulation module, and a data analysis and display module. The impedance monitoring module is connected to the data analysis and display module via an SPI / I2C bus, and the microcurrent stimulation module is connected to the data analysis and display module via a UART interface. The modules constitute a closed-loop architecture of perception-analysis-intervention.

[0010] The impedance monitoring module includes an impedance measurement circuit and a ring-shaped flexible electrode patch, enabling 50kHz / 100kHz / 1MHz tri-frequency scanning and four-site eight-electrode distributed measurement.

[0011] The microcurrent stimulation module includes a control unit, a signal generation unit, a power amplification unit, and a protection unit. It outputs a biphasic constant current stimulation waveform and has dual modes of low-frequency endothelial modulation and high-frequency muscle pump activation.

[0012] The data analysis and display module is equipped with an embedded lightweight random forest algorithm and adaptive closed-loop control logic, which automatically starts, adjusts or stops the intervention based on the thrombosis risk level.

[0013] As a further explanation of the above scheme, the annular flexible electrode patch has a medical-grade silicone rubber substrate and includes four attachment points: proximal thigh, distal thigh, proximal calf, and distal calf. Each point has a pair of Ag / AgCl electrodes, forming a spatially distributed measurement network of eight electrodes with a contact pressure of 0.5-1 N / cm. 2 This ensures measurement stability.

[0014] Furthermore, the impedance measurement circuit uses AD5933 as its core and 74HC4051 multiplexer to achieve tri-frequency automatic scanning. A constant AC excitation current is injected from the far electrode, with the excitation current <100μA. The output impedance amplitude and phase angle are measured, and the response voltage difference is detected by the near electrode. After sampling by the AD5933's built-in 12-bit ADC, the impedance amplitude |Z| and phase angle theta are output through the I2C bus.

[0015] Furthermore, the AD5933 is connected to an external crystal oscillator, and the VOUT pin of the AD5933 is connected to the common terminal COM of the 74HC4051 through a precision resistor; the eight select terminals S0-S7 of the 74HC4051 are respectively connected to the eight Ag / AgCl electrodes on the four ring flexible electrode patches through shielded wires; the RFB pin is connected to an external feedback resistor, and a decoupling capacitor and a tantalum capacitor are connected in parallel between VCC and GND; AVDD and DVDD are independently powered by a voltage regulator to reduce digital noise coupling.

[0016] Furthermore, the microcurrent stimulation module includes an STM32F103 control unit, a DDS+H bridge biphase wave unit, an OPA549 constant current source unit, and a MAX30004 overcurrent protection unit, outputting a 0.1-5mA biphase symmetrical square wave, with a hardware threshold of 5mA, and immediate shutdown when the contact impedance is >10kΩ.

[0017] Furthermore, the data analysis and display module includes a main controller, a display, and a wireless communication unit. It is the core processing hub and human-machine interface of the system. The main controller uses an STM32H743VIT6 microprocessor, reads the impedance data of AD5933 through the SPI bus, controls the microcurrent stimulation module through UART, and runs the embedded thrombosis risk assessment and closed-loop control algorithm.

[0018] The display uses a TFT-LCD touch screen to display impedance waveforms, risk probability curves, stimulation parameters and alarm information of each segment of the lower limb in real time; the wireless communication unit supports Bluetooth 5.0 and USB-OTG dual-mode transmission, which can upload monitoring data to the hospital information system (HIS) or cloud monitoring platform.

[0019] Furthermore, the embedded thrombosis risk assessment algorithm includes: signal preprocessing, ΔZ / PFR / SampEn / trend slope four-dimensional feature extraction, lightweight random forest inference, and three-level risk classification.

[0020] Furthermore, the adaptive dual-mode control logic is as follows: when PFR < 0.15, the 1Hz low-frequency endothelial modulation mode is activated; when PFR ≥ 0.15, the 30Hz high-frequency muscle pump activation mode is activated; the current is dynamically adjusted in a 2-minute cycle according to the ΔZ rise amplitude ± 0.2mA; a time-division multiplexing architecture for measurement / stimulation electrodes is adopted, sharing an eight-electrode array and a 74HC4051 multiplexer to avoid signal crosstalk; the closed-loop exit condition is that the thrombosis risk probability is < 10% for two consecutive monitoring cycles, and the impedance change is stable within the baseline ± 5%.

[0021] The beneficial effects that can be achieved by adopting the above-mentioned technical solution in this invention are:

[0022] 1. This invention employs a non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis, mainly composed of an impedance monitoring module, a microcurrent stimulation module, and a data analysis and display module. It is the first to achieve integrated "monitoring-early warning-intervention-feedback" for DVT, enabling proactive intervention in the very early stage of blood flow stasis.

[0023] 2. This invention has high monitoring accuracy, with three-frequency impedance + distributed electrodes + multi-dimensional features, effectively distinguishing between stasis and edema. Offline verification showed an accuracy of 92.3%, specificity of 94.1%, and AUC=0.96.

[0024] 3. Personalized intervention, dual-mode adaptation to different pathological mechanisms, closed-loop parameter dynamic adjustment, avoiding "one-size-fits-all" approach, and improving compliance.

[0025] 4. High safety: Dual-phase constant current, hardware overcurrent protection, contact impedance monitoring, and excitation / stimulation current all comply with IEC60601-1 medical safety standards.

[0026] 5. Widely compatible with wearable devices: Small size, light weight, and controllable cost (BOM 800-1200 yuan), suitable for multiple scenarios such as hospitalization, home, long-distance travel, and microgravity in aerospace. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0028] Figure 2 This is a schematic diagram of the annular flexible electrode patch structure of the present invention.

[0029] Figure 3 This is a distribution diagram of the attachment sites for the annular flexible electrode patch of the present invention.

[0030] Figure 4 This is a block diagram of the microcurrent stimulation module circuit.

[0031] Figure 5 This is a flowchart of an embedded thrombosis risk assessment algorithm software.

[0032] Figure 6 This is a state transition diagram for adaptive dual-mode closed-loop stimulus control.

[0033] Figure labeling: 1. Impedance monitoring module 1-1, Impedance measurement circuit 1-1; 1. Multiplex analog switch 74HC4051 1-2, Ring flexible electrode patch 1-2; 1. Patch A 1-2; 1. Patch B 1-2; 1. Patch C 1-2; 1. Patch D 1-2; 1. Distal current injection electrode (E+) 1-2; 1. Proximal voltage detection electrode (E-) 2. Microcurrent stimulation module 2-1, Control unit 2-2, Signal generation unit 2-3, Power amplification unit 2-4, Protection unit 3. Data analysis and display module 3-1, Main controller 3-2, Display 3-3, Wireless communication unit. Detailed Implementation

[0034] In the description of this invention, it should be noted that the directional terms such as "center", "lateral", "longitudinal", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", and "counterclockwise" indicate the orientation and positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. They should not be construed as limiting the specific protection scope of this invention.

[0035] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features. Thus, the use of "first" and "second" to define a feature may explicitly or implicitly include one or more of that feature, and in the description of this invention, "at least" means one or more, unless otherwise explicitly specified.

[0036] In this invention, unless otherwise explicitly specified and limited, the terms "assembly," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can also refer to a mechanical connection; they can refer to a direct connection or a connection through an intermediate medium; or they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0037] In this invention, unless otherwise specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "below," and "over" the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Above," "below," and "below" the second feature includes the first feature being directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0038] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings, making the technical solution and beneficial effects of the present invention clearer and more explicit. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.

[0039] like Figures 1-6As shown, the present invention is a non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis, including an impedance monitoring module 1, a microcurrent stimulation module 2, and a data analysis and display module 3. The impedance monitoring module is connected to the data analysis and display module via an SPI / I2C bus, and the microcurrent stimulation module is connected to the data analysis and display module via a UART interface. The modules constitute a closed-loop architecture of perception-analysis-intervention.

[0040] The impedance monitoring module includes an impedance measurement circuit 1-1 and a ring-shaped flexible electrode patch 1-2. The ring-shaped flexible electrode patch 1-2 employs a ring-shaped multi-electrode array composed of multiple sets of electrodes, which are respectively attached to four anatomical sites on the patient's lower limb: the proximal thigh (patch A 1-21), the distal thigh (patch B 1-22), the proximal calf (patch C 1-23), and the distal calf (patch D 1-24). Each site is equipped with a flexible conductive patch, and each flexible conductive patch has a pair of Ag / AgCl electrodes, namely a distal current injection electrode (E+) 1-25 and a proximal voltage detection electrode (E-) 1-26. A total of eight electrodes are used at the four sites, forming a spatially distributed measurement network. The ring-shaped flexible electrode patch 1-2 uses a medical-grade silicone rubber substrate with a thickness of 0.5 mm. Ag / AgCl disk electrodes with a diameter of 10 mm are arranged on it using a screen printing process, with an electrode spacing of 20 mm. Ag / AgCl electrode materials possess characteristics such as low polarization potential, high electrochemical stability, and minimal long-term drift, making them suitable for precise bioelectrical signal detection applications. The fixing buckle utilizes a medical-grade non-woven fabric strap and Velcro structure to ensure a constant contact pressure between the electrode and the skin (0.5-1.0 N / cm²). 2 To reduce motion artifacts and maintain contact impedance below 5 kOhm, the spatial arrangement between the current injection electrode (E+) and the voltage detection electrode (E-) employs a two-terminal measurement topology. By rationally designing the electrode spacing and the pairing method of the injection / detection electrodes, the penetration depth of the current in the tissue can be effectively controlled, so that the measurement results include information from both superficial subcutaneous tissue and deep muscle tissue hemodynamic changes.

[0041] Impedance measurement circuit 1-1 uses the AD5933 impedance analysis chip as its core. It is sequentially connected to the eight electrodes via a 74HC4051 multi-channel analog switch 1-11 to achieve automatic scanning measurement at three frequency points: 50kHz, 100kHz, and 1MHz. A two-terminal measurement topology is employed. A constant AC excitation current (amplitude strictly controlled below 100µA, meeting medical microcurrent safety standards) is injected through the distal electrode. The proximal electrode detects the response voltage difference. After sampling by the AD5933's built-in 12-bit ADC, the impedance amplitude |Z| and phase angle theta are output via the I2C bus. The AD5933 is connected to an external 16MHz crystal oscillator with a frequency resolution of 0.1Hz, providing an impedance measurement resolution better than 0.1Ohm, meeting the requirements for detecting subtle impedance changes in biological tissues. The AD5933's VOUT pin is connected to the COM common terminal of the 74HC4051 via a 10kOhm precision resistor. The 74HC4051's eight select pins S0-S7 are connected to the eight Ag / AgCl electrodes on four ring-shaped flexible electrode patches via shielded wires. The RFB pin is connected to an external 100kOhm feedback resistor. A 100nF decoupling capacitor and a 10uF tantalum capacitor are connected in parallel between VCC and GND. AVDD and DVDD are independently powered by 3.3V LDO regulators to reduce digital noise coupling.

[0042] The selection of these three frequencies is based on the following physiological principles: 50kHz primarily reflects changes in extracellular fluid impedance and is sensitive to tissue edema; 100kHz takes into account both extracellular fluid and cell membrane capacitance characteristics; and a 1MHz high-frequency signal can penetrate the cell membrane, reflecting the combined impedance characteristics of intracellular fluid and deep tissues. The circuit's impedance measurement resolution is better than 0.1 Ohm, meeting the requirements for accurate detection of minute impedance changes (such as changes of a few ohms caused by early edema).

[0043] The microcurrent stimulation module includes a control unit 2-1 (STM32F103 coprocessor), a signal generation unit 2-2 (DDS submodule and H-bridge biphasic wave synthesis circuit), a power amplification unit 2-3 (OPA549 constant current source circuit), a protection unit 2-4 (MAX30004 overcurrent detection chip), and an output interface. It outputs a biphasic constant current square wave with a frequency of 1Hz or 30Hz and an adjustable current intensity of 0.1-5mA. The module employs a hierarchical architecture design, with each functional subunit working collaboratively to generate a biphasic constant current pulsed electrical stimulation signal that meets medical safety standards.

[0044] The control unit 2-1 uses an STM32F103C8T6 coprocessor and receives stimulation control command frames from the main controller 3-1 of the data analysis and display module 3 via a UART serial interface. The command content includes stimulation mode selection (low frequency / high frequency), output current intensity setting (within the range of 0.1-5mA), pulse repetition frequency setting, stimulation duration, and start / stop control bit.

[0045] The coprocessor performs frame verification and parameter validity checks on the received instructions. Once confirmed to be correct, it parses and executes the instructions: Based on the stimulation mode configuration, it sends the corresponding waveform synthesis parameters (including positive phase pulse width, negative phase pulse width, pulse interval, etc.) to the signal generation unit 2-2; simultaneously, it outputs the current setpoint to the digitally controlled potentiometer or DAC of the power amplifier unit 2-3 to control the output level of the constant current source. It configures the TIM2 timer to generate a PWM at the corresponding frequency, which is then converted into a biphase square wave via the H-bridge.

[0046] Signal generation unit 2-2 includes a Direct Digital Frequency Synthesis (DDS) submodule and a biphase wave synthesis circuit. The DDS submodule generates a square wave signal with a precise frequency based on the STM32's built-in timer and DMA channel. It allows for flexible configuration of the pulse repetition frequency (supporting multiple frequency settings within the range of 1Hz to 50Hz), offering high frequency resolution and good long-term stability. The biphase wave synthesis circuit converts a single-phase square wave into a biphase square wave using an H-bridge topology. Positive-phase pulses (+I, pulse width 200µs) and negative-phase pulses (-I, pulse width 200µs) are output alternately, balancing the total positive and negative charges and preventing electrode polarization and skin electrolysis damage. The H-bridge consists of an IR2104 half-bridge driver and four IRF540N MOSFETs. The H-bridge control signals undergo dead-time insertion logic processing to ensure that the power switches of the upper and lower bridge arms do not conduct simultaneously, preventing power supply short circuits.

[0047] Power amplifier unit 2-3 includes a constant current source circuit and an impedance matching network. The constant current source circuit is based on an operational amplifier OPA549 and a power MOSFET to construct a voltage-to-current conversion topology, ensuring that the output current is stably adjustable within the range of 0.1mA to 5mA, and the output impedance is >1MΩ, so that the stimulation current is not affected by the impedance fluctuations of the electrode-skin contact. The impedance matching network includes a series DC blocking capacitor (100uF) and a parallel RC buffer circuit to filter out the DC component and suppress high-frequency harmonics. The non-inverting input of the OPA549 receives the bi-phase voltage signal output from the H-bridge, and the inverting input forms a current negative feedback with the output through a 1Ohm precision sampling resistor to achieve constant current output. The large-capacity DC blocking coupling capacitor (100uF) connected in series in the output path blocks the DC component, allowing only the AC pulse signal to pass through, avoiding electrolytic damage to skin tissue and electrode polarization effects caused by DC current.

[0048] Protection units 2-4 include a MAX30004 bioelectric front-end overcurrent detection chip and a real-time contact impedance monitoring circuit. The IN+ and IN- pins of the MAX30004 are connected across a 1 Ohm sampling resistor to monitor the stimulation current in real time. If the contact impedance > 10 kOhm or the stimulation current exceeds the 5 mA hardware threshold, the power MOSFET gate drive signal is immediately cut off, triggering an audible and visual alarm and reporting to the data analysis module 3. The INT pin of the MAX30004 triggers an external interrupt on the STM32F103, immediately shutting down the SD shutdown pin of the IR2104 and cutting off the H-bridge drive. The overcurrent protection function uses a dedicated overcurrent detection chip to monitor the output current in real time. When the output current exceeds the set safety threshold (5 mA), the overcurrent detection chip triggers a protection action within microseconds, immediately cutting off the H-bridge drive signal and interrupting the stimulation output through hard-wired logic. This hardware-level protection does not rely on software response, ensuring the reliability of the protection. The microcurrent stimulation module 2 uses an independent isolated DC-DC power supply, achieving electrical isolation from other modules in the system.

[0049] The data analysis and display module 3 includes a main controller 3-1 (STM32H743 microprocessor, ARM Cortex-M7 core, 480MHz clock speed, 2MB Flash, 1MB RAM), a display 3-2 (2.8-inch TFT-LCD touchscreen, 320*240 resolution), and a wireless communication unit 3-3, running embedded thrombosis risk assessment and closed-loop control algorithms. The Cortex-M7 core adopts a superscalar pipelined design, integrating a single-precision floating-point unit (FPU) and a digital signal processing instruction set (DSP), possessing powerful numerical computation capabilities. The wireless communication unit 3-3 supports dual-mode transmission via Bluetooth 5.0 and USB-OTG, enabling the uploading of monitoring data to a hospital information system (HIS) or a cloud monitoring platform. Compared to previous versions, the Bluetooth 5.0 protocol offers longer transmission distances, higher data transmission rates, and lower power consumption, making it suitable for long-term continuous data uploads in medical monitoring scenarios. Transmitted data is protected using an encryption protocol to ensure the security of patient privacy data. The embedded thrombosis risk assessment algorithm comprises a signal preprocessing layer S1, a time-frequency feature extraction layer S2, a machine learning inference layer S3, and a risk grading decision layer S4. The algorithm design adheres to the principles of low computational complexity, low memory consumption, and high real-time performance, ensuring efficient operation on resource-constrained embedded platforms.

[0050] Signal preprocessing layer S1: The system triggers a multi-frequency scan every 5 minutes, with a sampling rate of 200Hz and a sampling duration of 10 seconds for each frequency point, to obtain the original impedance time series Z_raw(t). The following steps are performed on Z_raw(t) in sequence: (1) Bandpass filtering (0.05Hz-10Hz, fourth-order Butterworth digital filter) to remove baseline drift and high-frequency noise; (2) Motion artifact removal, using a sliding window anomaly detection algorithm based on the 3sigma criterion. If the amplitude of a certain sampling point deviates from the mean of the window by more than 3 times the standard deviation, it is marked as an artifact and replaced with linear interpolation; (3) Baseline correction, using the mean impedance of the patient in the first calm state after admission as the individualized baseline Z_baseline, and calculating the relative impedance change Delta Z(t)=[Z_raw(t)-Z_baseline] / Z_baseline x 100%.

[0051] Time-frequency feature extraction layer S2: Extracts a four-dimensional feature vector from the preprocessed impedance signal.

[0052] (1) Impedance change Delta Z: The average value of Delta Z at each frequency point in the current cycle is taken to reflect the comprehensive changes in lower limb blood volume and tissue edema.

[0053] (2) Low-frequency power ratio (PFR): Perform a 1024-point FFT transformation on the impedance signal to calculate the power spectral density of the 0.1-0.3Hz frequency band (corresponding to the respiratory fluctuations and Mayer wave range of venous blood flow) to the total power of the entire frequency band (0.05-1Hz); a decrease in PFR indicates venous blood stasis and decreased vascular compliance.

[0054] (3) Waveform entropy SampEn: The complexity of impedance waveforms is quantified by the sample entropy algorithm (embedding dimension m=2, similarity tolerance r=0.2 times the standard deviation); hemodynamic disturbances in the pre-thrombotic stage are often manifested as enhanced signal regularity and decreased entropy value;

[0055] (4) Historical trend characteristics: A 10-minute sliding window (including the two most recent monitoring periods) was used to perform a first-order linear regression on Delta Z, and the regression slope k and the coefficient of determination R2 were extracted to reflect the dynamic evolution trend of thrombosis risk.

[0056] Machine Learning Inference Layer S3: The main controller 3-1 loads a pre-trained lightweight random forest classifier model. This model is trained and pruned offline using a publicly available DVT clinical dataset (a cohort of high-risk patients with Caprini scores >5, including both diagnosed and undiagnosed control groups). Hyperparameters are: number of decision trees n_estimators = 100, maximum tree depth max_depth = 10, feature subset sampling ratio max_features = sqrt, minimum number of leaf node samples min_samples_leaf = 5. The model input is a four-dimensional feature vector X = [Delta Z, PFR, SampEn, k], and the output is the thrombosis risk probability P_risk (0%-100%).

[0057] To avoid overfitting and adapt to the computing power limitations of embedded platforms, a post-pruning process based on Gini impurity importance ranking is employed: decision branches with importance below 1% are removed, compressing the model size to <200KB and achieving an inference latency of <50ms. Offline validation using 10-fold cross-validation yielded the following results for the lightweight random forest model with four-dimensional feature input: accuracy 92.3%, sensitivity 89.7%, specificity 94.1%, and AUC=0.96. The single-sample inference latency is 42ms (STM32H743@480MHz).

[0058] Risk grading decision layer S4: The main controller 3-1 classifies risk levels based on the P_risk value and executes corresponding decisions: Low risk (P_risk<30%): Green indicator light, continuous monitoring only, no stimulation triggered; Medium risk (30%<=P_risk<60%): Yellow indicator light, triggering microcurrent stimulation intervention; High risk (P_risk>=60%): Red indicator light and buzzer alarm, triggering enhanced stimulation and prompting medical staff to intervene.

[0059] Adaptive Dual-Mode Closed-Loop Stimulation Control: When the system determines that it has entered a medium-risk or high-risk state, it automatically activates adaptive dual-mode closed-loop stimulation control, which includes three sub-stages: stimulation mode selection, dynamic parameter adjustment, and intervention effect feedback. The system selects a targeted stimulation mode based on the current PFR value: If PFR < 0.15, it indicates that the patient's low-frequency blood flow oscillation is significantly insufficient, suggesting vascular endothelial dysfunction and decreased vasodilation capacity. The system selects the "low-frequency endothelial modulation mode": outputting a biphasic square wave with a frequency of 1 Hz, an initial current intensity of 0.5 mA, and a duty cycle of 50%. This parameter improves vasodilation and antiplatelet adhesion function by stimulating vascular endothelial cells to release nitric oxide (NO) and prostacyclin (PGI2). If PFR>=0.15, it indicates that the patient's vasomotor function is acceptable but the muscle pump return force is insufficient. The system selects "high-frequency muscle pump activation mode": output biphasic square wave, frequency 30Hz, initial current intensity 2.0mA, duty cycle 50%; this parameter induces rhythmic contraction of skeletal muscle, simulates the physiological "muscle pump" effect, and promotes centripetal return of venous blood.

[0060] The system uses a 2-minute evaluation cycle and immediately performs impedance retesting after each stimulus, calculating the rise in Delta Z before and after stimulation: Delta Z_rec = Delta Z_post - Delta Z_pre. Based on Delta Z_rec, the stimulation current intensity I_stim for the next cycle is dynamically adjusted: if Delta Z_rec < 5%, the current intensity is increased by 0.2 mA (upper limit 5 mA); if 5% <= Delta Z_rec <= 15%, the current intensity is maintained; if Delta Z_rec > 15%, the current intensity is decreased by 0.2 mA (lower limit 0.1 mA).

[0061] Simultaneously, the system records the PFR change trend for three consecutive assessment periods: if the PFR increases from <0.15 to >=0.15, it automatically switches from low-frequency mode to high-frequency mode; conversely, if the PFR continues to decrease and Delta Z_rec remains <5%, a high-risk alarm is triggered, indicating the need for drug or mechanical intervention. The mode switching process is designed with a hysteresis interval and a gradual transition mechanism to avoid frequent mode switching caused by small fluctuations in PFR near the threshold.

[0062] When P_risk is below 10% for two consecutive monitoring cycles (10 minutes) and Delta Z remains stable within ±5% of baseline, the system automatically stops stimulation and switches to pure monitoring mode, completing one closed-loop intervention cycle. The entire closed-loop control process runs continuously in cycles of tens of seconds to several minutes, enabling dynamic tracking of the patient's DVT risk and automatic optimization and adjustment of intervention measures. The system defines a complete closed-loop path from "monitoring-assessment-intervention-reassessment," ensuring that the intervention effect can be quantified in real time and dynamically optimized.

[0063] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis, characterized in that, It includes an electrical impedance monitoring module, a microcurrent stimulation module, and a data analysis and display module. The electrical impedance monitoring module is connected to the data analysis and display module via an SPI / I2C bus, and the microcurrent stimulation module is connected to the data analysis and display module via a UART interface. The modules together form a closed-loop architecture of perception-analysis-intervention. The impedance monitoring module includes an impedance measurement circuit and a ring-shaped flexible electrode patch, enabling 50kHz / 100kHz / 1MHz tri-frequency scanning and four-site eight-electrode distributed measurement. The microcurrent stimulation module includes a control unit, a signal generation unit, a power amplification unit, and a protection unit. It outputs a biphasic constant current stimulation waveform and has dual modes of low-frequency endothelial modulation and high-frequency muscle pump activation. The data analysis and display module is equipped with an embedded lightweight random forest algorithm and adaptive closed-loop control logic, which automatically starts, adjusts or stops the intervention based on the thrombosis risk level.

2. The non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis according to claim 1, characterized in that, The annular flexible electrode patch has a medical-grade silicone rubber substrate and includes four attachment points: proximal thigh, distal thigh, proximal calf, and distal calf. Each point has a pair of Ag / AgCl electrodes, forming a spatially distributed measurement network of eight electrodes. The contact pressure is 0.5-1 N / cm. 2 This ensures measurement stability.

3. The non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis according to claim 1, characterized in that, The impedance measurement circuit uses AD5933 as its core and 74HC4051 multiplexer to achieve tri-frequency automatic scanning. A constant AC excitation current is injected from the far electrode (excitation current <100μA), and the impedance amplitude and phase angle are output. The near electrode detects the response voltage difference. After sampling by the AD5933's built-in 12-bit ADC, the impedance amplitude |Z| and phase angle theta are output through the I2C bus.

4. The non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis according to claim 1, characterized in that, The AD5933 is connected to an external crystal oscillator. The VOUT pin of the AD5933 is connected to the common terminal COM of the 74HC4051 through a precision resistor. The eight selector pins S0-S7 of the 74HC4051 are connected to the eight Ag / AgCl electrodes on the four ring flexible electrode patches through shielded wires. The RFB pin is connected to an external feedback resistor. A decoupling capacitor and a tantalum capacitor are connected in parallel between VCC and GND. AVDD and DVDD are independently powered by a voltage regulator to reduce digital noise coupling.

5. The non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis according to claim 1, characterized in that, The microcurrent stimulation module includes an STM32F103 control unit, a DDS+H bridge biphase wave unit, an OPA549 constant current source unit, and a MAX30004 overcurrent protection unit. It outputs a 0.1-5mA biphase symmetrical square wave with a hardware threshold of 5mA. It immediately shuts off when the contact impedance is greater than 10kΩ.

6. The non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis according to claim 1, characterized in that, The data analysis and display module includes a main controller, a display, and a wireless communication unit. It is the core processing center and human-machine interface of the system. The main controller uses an STM32H743VIT6 microprocessor, reads the impedance data of AD5933 through the SPI bus, controls the microcurrent stimulation module through UART, and runs an embedded thrombosis risk assessment and closed-loop control algorithm. The display uses a TFT-LCD touch screen to display impedance waveforms, risk probability curves, stimulation parameters and alarm information of each segment of the lower limb in real time; the wireless communication unit supports Bluetooth 5.0 and USB-OTG dual-mode transmission, which can upload monitoring data to the hospital information system or cloud monitoring platform.

7. The non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis according to claim 1, characterized in that, The embedded thrombosis risk assessment algorithm includes: signal preprocessing, ΔZ / PFR / SampEn / trend slope four-dimensional feature extraction, lightweight random forest inference, and three-level risk classification.

8. The non-invasive real-time monitoring and intervention system for lower extremity venous thrombosis according to claim 1, characterized in that, The adaptive dual-mode control logic is as follows: when PFR < 0.15, the 1Hz low-frequency endothelial modulation mode is activated; when PFR ≥ 0.15, the 30Hz high-frequency muscle pump activation mode is activated; the current is dynamically adjusted in a 2-minute cycle according to the ΔZ rise amplitude ± 0.2mA; a time-division multiplexing architecture for measurement / stimulation electrodes is adopted, sharing an eight-electrode array and a 74HC4051 multiplexer to avoid signal crosstalk; the closed-loop exit condition is that the probability of thrombosis risk is < 10% for two consecutive monitoring cycles, and the impedance change is stable within the baseline ± 5%.

Images