Monitoring system for pediatric severe nursing based on ultrasonic waves

Abstract

The invention relates to the technical field of medical monitoring, and discloses a pediatric severe nursing monitoring system based on ultrasonic waves, which is characterized in that a flexible ultrasonic sensor array transmits ultrasonic pulses through a piezoelectric micromechanical ultrasonic transducer and receives pleura line echo signals, and meanwhile, senses the real-time temperature of a contact surface through a miniature thermocouple sensor; the multi-mode signal acquisition module processes echo signals and real-time temperature, and the self-adaptive beam forming and heat dose control module dynamically adjusts emission voltage and pulse repetition frequency of the ultrasonic transducer according to a comparison result of the real-time temperature and a preset safety threshold value, and controls temperature rise of a contact surface to be within a safety range. And a breathing analysis module based on sub-pixel displacement performs sub-pixel-level displacement vector tracking on the pleura line strong echo points to generate refined breathing waveform data. By designing the intelligent integrated pediatric special monitoring system, high-precision, low-risk and long-time continuous monitoring of respiratory movement of a severe pediatric patient is realized.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring technology, and in particular to an ultrasound-based monitoring system for pediatric intensive care. Background Technology

[0002] In pediatric intensive care, continuous monitoring of respiratory function is an important means of assessing changes in the condition of children. Pleural line movement directly reflects respiratory rhythm and amplitude, and its accurate tracking is of great clinical significance. However, traditional monitoring methods such as impedance respiration or manual observation have problems such as slow response, low accuracy and many interferences, which make it difficult to meet the real-time monitoring needs of critically ill children.

[0003] In recent years, ultrasound technology has been gradually introduced into the field of respiratory monitoring due to its advantages of being non-invasive, real-time, and repeatable. However, traditional ultrasound probes are mostly rigid structures, which are difficult to adapt to the curvature of children's chests. Prolonged contact can easily lead to signal loss or artifacts. In addition, ultrasound transducers generate heat during operation, and if the temperature rise is not properly controlled, it may cause thermal damage to pediatric patients with delicate skin.

[0004] To address the issue of ultrasound temperature rise, existing equipment often employs fixed power output or simple temperature alarm mechanisms, lacking dynamic assessment and predictive control of thermal dose. Especially during long-term continuous monitoring, it is impossible to adjust transmission parameters based on real-time temperature and energy output, posing safety hazards. Furthermore, pleural line motion extraction relies heavily on manual annotation or simple image processing, limiting accuracy and making it difficult to distinguish between combined movements caused by respiration and heartbeat. Currently, there is no pediatric-specific monitoring system that integrates flexible sensing, multimodal synchronous acquisition, adaptive thermal dose control, and sub-pixel respiratory analysis. A comprehensive intelligent solution is urgently needed. Summary of the Invention

[0005] This invention provides an ultrasound-based monitoring system for pediatric critical care to address existing technical problems.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides an ultrasound-based monitoring system for pediatric critical care, comprising: a flexible ultrasound sensor array, a multimodal signal acquisition module, an adaptive beamforming and thermal dose control module, and a respiratory analysis module based on subpixel displacement; The flexible ultrasonic sensor array is used to transmit ultrasonic pulses to the child's chest cavity through a piezoelectric micromechanical ultrasonic transducer and receive the original echo signal of the pleural line movement, and to sense the physical temperature value of the probe-skin contact surface in real time through a miniature thermocouple sensor integrated in the transducer gap. The multimodal signal acquisition module is used to digitize the original echo signal and the physical temperature value respectively to generate a digital echo data stream and a real-time temperature data packet; The adaptive beamforming and thermal dose control module is used to dynamically adjust the transmission voltage and pulse repetition frequency of the piezoelectric micromechanical ultrasonic transducer based on the comparison result of the received real-time temperature data packet and the preset safety threshold, so as to control the real-time temperature rise of the contact surface within a safe range. The subpixel displacement-based respiratory analysis module is used to perform subpixel-level displacement vector tracking on strong echo points representing the pleural line in the digital echo data stream, and to generate refined respiratory waveform data based on the periodic differences of the displacement vectors.

[0007] The beneficial effects of the technical solution provided by this invention include at least the following: This invention, through the design of a flexible ultrasound sensor array, employs a biocompatible silicone substrate and biomimetic microstructures, significantly improving the fit and comfort between the probe and the child's skin, effectively eliminating air bubbles on the contact surface, ensuring stable transmission of ultrasound signals, and is suitable for pediatric intensive care scenarios.

[0008] By introducing an adaptive beamforming and thermal dose control module, this invention enables the system to monitor the temperature changes at the probe-skin contact surface in real time and dynamically adjust the energy output in conjunction with the emission parameters to ensure that the tissue temperature rise is always within a safe threshold. Furthermore, by combining LSTM temperature rise prediction and multi-level energy regulation strategies, the predictability and precision of safety control are further improved.

[0009] This invention constructs a respiratory analysis module based on deep learning and subpixel displacement, which can automatically identify the pleural line region and perform edge enhancement. It combines pyramid optical flow and Gaussian surface fitting to achieve subpixel-level tracking of pleural line movement, and effectively distinguishes the displacement caused by breathing and heartbeat through multi-peak motion separation technology to generate high-fidelity respiratory waveforms, providing a reliable basis for clinical diagnosis.

[0010] The system of this invention also realizes the synchronous acquisition and frame-level alignment of ultrasound echo and temperature data, laying the foundation for subsequent multimodal fusion analysis and intelligent control, improving the availability of data and the reliability of clinical decision-making, and has significant clinical application value and promotion prospects. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0012] Figure 1 This is a system structure diagram provided in an embodiment of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0014] This embodiment provides an ultrasound-based monitoring system for pediatric intensive care. Please refer to [link / reference]. Figure 1 This is a system structure diagram provided in an embodiment of the present invention.

[0015] The pediatric critical care monitoring system based on ultrasound in this embodiment includes: a flexible ultrasound sensor array, a multimodal signal acquisition module, an adaptive beamforming and thermal dose control module, and a respiratory analysis module based on subpixel displacement. A flexible ultrasonic sensor array is used to transmit ultrasonic pulses to the child's chest cavity through a piezoelectric micromechanical ultrasonic transducer and receive the raw echo signal of the pleural line movement. It also senses the physical temperature value of the probe-skin contact surface in real time through a miniature thermocouple sensor integrated in the transducer gap. The multimodal signal acquisition module is used to digitize the raw echo signal and physical temperature value respectively, generating digital echo data stream and real-time temperature data packet; The adaptive beamforming and thermal dose control module is used to dynamically adjust the transmission voltage and pulse repetition frequency of the piezoelectric micromechanical ultrasonic transducer based on the comparison result of the received real-time temperature data packet and the preset safety threshold, so as to control the real-time temperature rise of the contact surface within a safe range. The subpixel displacement-based respiratory analysis module is used to perform subpixel-level displacement vector tracking of strong echo points representing the pleural line in the digital echo data stream, and to generate refined respiratory waveform data based on the periodic differences of the displacement vectors.

[0016] As one embodiment of the present invention, the flexible ultrasonic sensor array further includes: a flexible substrate material layer, an array transducer patch layer, and a top encapsulation layer. The flexible base material layer is made of biocompatible medical silicone, and its lower surface has a biomimetic microstructure texture to fit the skin of young children and eliminate tiny air bubbles on the contact surface. An array-type transducer patch layer is composited on the upper surface of a flexible substrate material layer. Piezoelectric micromechanical ultrasonic transducers are embedded in the array-type transducer patch layer in a matrix form. The spacing between adjacent piezoelectric micromechanical ultrasonic transducers is an integer multiple of half the wavelength of the ultrasonic pulse, which is used to synthesize a phased beam when transmitting ultrasonic pulses to focus on the focal area where the pleural line is located. The top encapsulation layer covers the array transducer patch layer, and the miniature thermocouple sensor is attached to the inner surface of the top encapsulation layer, located in the acoustic blind zone of the piezoelectric micromechanical ultrasonic transducer.

[0017] It should be noted that in this embodiment, the flexible substrate material layer is made of medical-grade platinum-catalyzed silicone. This type of material has passed the ISO 10993 cytotoxicity test and has excellent biocompatibility and acoustic impedance matching characteristics. The biomimetic microstructure texture on the lower surface is prepared by laser etching process, specifically an array of hexagonal pit structures. This structure mimics the microscopic morphology of a tree frog's foot. When attached to the skin, the capillary channels formed by the microstructure guide the air on the contact surface to the edge to be discharged, avoiding the scattering interference of air bubbles on the propagation of ultrasound. At the same time, the microstructure increases the contact area with the skin, achieving a firm adhesion without adhesive. It is painless and leaves no residue when removed, making it suitable for critically ill children with delicate skin.

[0018] In this embodiment, the piezoelectric micromechanical ultrasonic transducer (hereinafter referred to as PMUT) uses aluminum nitride as the piezoelectric material and is fabricated using MEMS technology. The size of a single transducer unit is 150μm×150μm, the working center frequency is designed to be 5MHz, and the spacing between adjacent piezoelectric micromechanical ultrasonic transducers is designed to be 154μm. The calculation basis for this value is: the speed of ultrasound propagation in human tissue is about 1540m / s, the wavelength corresponding to the center frequency of 5MHz is λ=1540 / 5e6=308μm, and half the wavelength is 154μm. The spacing is set to an integer multiple of half the wavelength (1 times in this embodiment). The purpose is to control the phase delay of each transducer unit so that the ultrasonic pulses emitted by each transducer unit are superimposed in phase at the pleural line depth (usually 1-3cm thick in children's chest walls), thereby achieving beam focusing and enhancing the signal-to-noise ratio in this depth region. During reception, the echo signal is delayed and superimposed to improve the lateral resolution.

[0019] In this embodiment, the flexible ultrasonic sensor array is designed to have a scale of 32×32, totaling 1024 PMUT elements. The imaging field covers an area of ​​20mm×20mm, which is sufficient to include the pleural line and part of the intercostal space. The miniature thermocouple sensor uses a thin-film T-type thermocouple, which is attached to the inner surface of the top encapsulation layer. The specific location is selected at the center intersection of four adjacent PMUT elements. This location has been verified by finite element simulation as the zero point region of the sound pressure field (i.e., the acoustic blind zone). Therefore, the presence of the thermocouple will not block or reflect the ultrasonic transmission and reception beams.

[0020] The above material selection, size parameters, and process steps are preferred examples that can be implemented by those skilled in the art based on the concept of the present invention. In practical applications, the specific parameters can be adjusted according to factors such as the child's body size and the monitoring site.

[0021] As one embodiment of the present invention, the array transducer patch layer also integrates a time gain compensation circuit, which is connected to a piezoelectric micromechanical ultrasonic transducer for gain compensation of the received original echo signal to counteract the attenuation of the ultrasonic signal in the thoracic tissue. The top encapsulation layer also integrates a temperature compensation reference circuit, which is connected to a miniature thermocouple sensor to receive physical temperature values ​​and generate a temperature reference voltage. The compensation slope of the time gain compensation circuit is dynamically adjusted according to the temperature reference voltage.

[0022] It should be noted that, in order to compensate for the attenuation of ultrasound signals propagating in the pleural tissue, this embodiment adopts a piecewise linear ramp control strategy. It is known that the attenuation coefficient of ultrasound propagation in tissue is approximately 0.5 dB / cm / MHz. For ultrasound with a center frequency of 5 MHz, the attenuation is approximately 2.5 dB / cm. The chest wall thickness of children is typically in the range of 1-3 cm, and the pleural line is located 1-3 cm deep subcutaneously. Therefore, the maximum attenuation is approximately 7.5 dB. Based on the above principle, the specific implementation of the time gain compensation circuit is as follows: (1) The original echo signal received by the PMUT array element first passes through a preamplifier stage consisting of four parallel low noise amplifiers (LNAs). This preamplifier stage adopts a differential input structure and has a common mode rejection ratio ≥80dB to suppress common mode interference. (2) The preamplified signal enters the main amplification stage of the variable gain amplifier (hereinafter referred to as VGA). The gain of VGA is linearly adjusted by the external control voltage V_TGC. The core amplifier chip selected in this embodiment has a gain adjustment range of -4.5dB to +43.5dB and a gain control slope of 40dB / V. (3) The time gain compensation circuit has a built-in ramp generator. With the time of ultrasound pulse emission as the time zero point, within a time window of 0-40μs (corresponding to an ultrasound propagation depth of about 0-6cm), V_TGC is linearly increased from 0.2V to 1.0V. Since VGA gain is linearly related to V_TGC, the corresponding amplifier gain is linearly increased from 10dB to 42dB. This compensation curve matches the exponential characteristics of tissue attenuation, so that the amplitude of the pleural line echo signal at different depths is normalized to the same level, which is convenient for subsequent signal processing.

[0023] The core components of the temperature compensation reference circuit are a high-precision digital-to-analog converter (DAC) and a microcontroller (MCU), and its workflow is as follows: (1) The miniature thermocouple sensor outputs a weak voltage signal of about 40μV / ℃. This signal is first amplified to the range of 0-2.5V by an instrumentation amplifier. The amplified voltage signal is then converted into a digital temperature value by a 12-bit analog-to-digital converter (ADC) and sent to the MCU. (2) The MCU has a temperature-compensation slope lookup table pre-stored inside. This lookup table is obtained through experimental determination (i.e., simulating tissue attenuation characteristics at different temperatures in a constant temperature water bath and recording the optimal compensation slope required to keep the pleural line echo amplitude constant). The MCU looks up the corresponding compensation slope according to the current real-time temperature value and writes the corresponding digital value to the DAC. The DAC outputs the temperature reference voltage V_TEMP.

[0024] The effect of temperature on ultrasound propagation is mainly reflected in the following: for every 1°C increase in tissue temperature, the ultrasound propagation speed increases by approximately 0.1%, and the attenuation coefficient changes by approximately 0.2%-0.5%. During long-term monitoring, the temperature of the probe-skin contact surface may rise from the initial 36.5°C to 38.5°C, which will cause a slight change in the attenuation characteristics of the ultrasound echo, introducing measurement errors in subsequent sub-pixel level displacement tracking. Therefore, this embodiment achieves dynamic adjustment of time gain compensation through temperature compensation using the following mechanism: Using V_TEMP as a bias voltage superimposed on the reference voltage terminal of the ramp generator (usually implemented by an integrator circuit consisting of an operational amplifier, resistor R, and capacitor C), the output voltage V_TGC(t) changes with time t as follows: V_TGC(t) = [V_REF / (R·C)]·t, where V_REF is the input reference voltage of the integrator circuit, V_REF = V_TEMP, and R and C are the time constant values ​​of the resistor and capacitor. When the temperature changes, V_REF = V_TEMP changes accordingly, and the ramp generator's output voltage rise slope k = V_REF / (R·C) is dynamically adjusted accordingly to compensate for the additional attenuation.

[0025] In one embodiment of the present invention, the multimodal signal acquisition module further includes: an ultrasonic simulation front end, a temperature acquisition unit, and a synchronous sampling clock generator; The ultrasonic simulation front end is connected to a piezoelectric micromechanical ultrasonic transducer, which is used to amplify and convert the original echo signal into digital data to generate a digital echo data stream. The temperature acquisition unit, connected to a miniature thermocouple sensor, is used to perform analog-to-digital conversion on physical temperature values ​​and generate real-time temperature data packets. A synchronous sampling clock generator is connected to the ultrasonic simulation front end and the temperature acquisition unit respectively. It is used to generate a first sampling clock to supply the ultrasonic simulation front end, and to generate a second sampling clock with an integer multiple frequency division relationship with the first sampling clock to supply the temperature acquisition unit. At the same time, it generates a frame synchronization signal according to the frame period of the first sampling clock to align the digital echo data stream and the real-time temperature data packet in time.

[0026] It should be noted that in this embodiment, the sampling rate of the ultrasound simulation front end is set to 40 MSPS, outputting 12-bit parallel data. 2048 sampling points are collected in each ultrasound pulse emission cycle, covering a detection depth of about 6 cm. The sampling rate of the temperature acquisition unit is set to 100 SPS, outputting 16-bit serial data. The digital echo data stream output by the ultrasound simulation front end and the real-time temperature data packet output by the temperature acquisition unit are sent together to the data stream encapsulation unit (such as FPGA or CPLD). Based on the frame synchronization signal, the echo data of 2048 sampling points in each frame and the temperature data at the corresponding time of the frame are combined into an encapsulated data packet, which is then sent to the subsequent adaptive beamforming and thermal dose control module and respiratory analysis module through the LVDS interface.

[0027] As one embodiment of the present invention, the adaptive beamforming and thermal dose control module further includes: a thermal dose accumulation calculation unit and a temperature rise trend prediction unit; The thermal dose accumulation calculation unit is connected to the flexible ultrasonic sensor array and the multimodal signal acquisition module respectively. It has a pediatric tissue thermal dose equation preset inside, which is used to calculate the tissue thermal dose accumulation value at the current moment based on the real-time temperature data packet and the current transmission voltage and pulse repetition frequency of the piezoelectric micromechanical ultrasonic transducer. The temperature rise trend prediction unit, connected to the heat dose accumulation calculation unit, is used to predict the tissue temperature rise trend curve within a future preset time window based on the historical tissue heat dose accumulation value sequence and an internally integrated temperature rise predictor based on a time-series neural network. If it is determined that the tissue temperature rise trend within the future preset time window exceeds the safety threshold, the adjustment of the emission voltage and pulse repetition frequency is triggered.

[0028] It should be noted that in this embodiment, the heat dose accumulation calculation unit is implemented using a field-programmable gate array (FPGA). It receives real-time temperature data packets from the multimodal signal acquisition module, as well as the current transmission voltage (unit: V) and pulse repetition frequency (unit: Hz) of the piezoelectric micromechanical ultrasonic transducer from the flexible ultrasonic sensor array. This embodiment employs a heat dose accumulation model based on the Arrhenius equation, widely used in the field of biological heat transfer, to convert the time-temperature history into an equivalent heat dose accumulation value. The calculation formula is as follows: C(t)=∫_(0~t) r·A^[T(τ)−T_0] dτ; In the formula, t is the cumulative time, τ is the time element, T(τ) is the tissue temperature at time τ, T_0 is the normal reference temperature for the human body (usually set to T_0=37℃), r is the equivalent temperature rise coefficient, A is the temperature coefficient, when T(τ)≥T_0, A=0.5, when T(τ)<T_0, A=0.25, C(t) is the cumulative heat dose value, and the unit of C(t) is equivalent minutes (i.e., the time equivalent to exposure at the reference temperature T_0). In this embodiment, considering the tissue sensitivity of critically ill pediatric patients, the preset safety threshold is 120 equivalent minutes, that is, the cumulative heat dose value shall not exceed the level equivalent to 120 minutes of exposure at 37℃.

[0029] The emission voltage and pulse repetition frequency directly affect the ultrasound energy output, and thus the tissue temperature rise. In this embodiment, both are incorporated into the thermal dose calculation. The thermal dose accumulation calculation unit performs the following operations at a refresh rate of 100Hz: (1) Read the current real-time temperature value T_current, the current transmit voltage V_current, and the pulse repetition frequency PRF_current; (2) Calculate the equivalent temperature rise coefficient r=(V_current / V_ref)^2×(PRF_current / PRF_ref) at the current moment, where V_ref is the reference voltage and PRF_ref is the reference pulse repetition frequency. In this embodiment, V_ref=10V and PRF_ref=1kHz are taken. (3) Substitute the equivalent temperature rise coefficient r and T_current into the formula: ΔC=r·A(T_current-T0)·Δt, and calculate the heat dose increment ΔC(t) within the current time step Δt; (4) Add ΔC(t) to the historical cumulative heat dose value and output the current cumulative heat dose value C(t) to the temperature rise trend prediction unit.

[0030] In this embodiment, the temperature rise trend prediction unit is implemented using an embedded neural network processor (NPU). Preferably, the temporal neural network used in this embodiment is a Long Short-Term Memory (LSTM) network, and its structure is as follows: Input layer: Receives a historical cumulative heat dose value sequence of length 60 (corresponding to the data of the past 60 seconds, with 6000 points at a sampling rate of 100Hz, and 60 representative points after downsampling). Hidden layers: Two LSTM layers, each containing 128 memory units, used to extract long-term dependencies in time series; Output layer: Fully connected layer, outputs the predicted temperature rise trend curve for the next 30 seconds (30 prediction points at 1-second intervals).

[0031] During actual operation, the temperature rise trend prediction unit performs the following operations: (1) Read the historical cumulative heat dose value sequence for the most recent 60 seconds from the heat dose cumulative calculation unit; (2) Input the historical cumulative heat dose sequence into the trained LSTM network for forward inference calculation and output the predicted temperature rise trend curve for the next 30 seconds; (3) Compare the maximum value in the predicted temperature rise trend curve with the preset safety threshold. If the predicted temperature rise at any time within the next 30 seconds will exceed the safety threshold, then immediately trigger the command to adjust the launch parameters.

[0032] The above implementation methods, parameter settings, and decision-making mechanisms are all preferred examples that can be implemented by those skilled in the art based on the present invention. In practical applications, parameters such as the heat dose threshold and prediction time window can be optimized and adjusted according to factors such as the child's weight and monitoring site. As one embodiment of the present invention, the adaptive beamforming and thermal dose control module further includes: a multi-parameter energy regulator and a beamforming controller; The multi-parameter energy regulator is connected to the heat dose accumulation calculation unit, the temperature rise trend prediction unit, and the beamforming controller. It has a set of energy control strategies for different heat dose accumulation value ranges. It is used to dynamically select the target energy control strategy from the set of energy control strategies based on the received current heat dose accumulation value and the future temperature rise prediction curve, and generate the corresponding energy regulation command. A beamforming controller, connected to a multi-parameter energy regulator and a piezoelectric micromechanical ultrasonic transducer, is used to synchronously adjust the transmission voltage and pulse repetition frequency of the piezoelectric micromechanical ultrasonic transducer according to energy regulation commands.

[0033] It should be noted that the multi-parameter energy regulator has a pre-set set of energy control strategies corresponding to different cumulative heat dose ranges. This set of energy control strategies is pre-determined through offline experiments and simulations and stored as a lookup table. In this embodiment, five heat dose ranges are set, corresponding to five different energy control strategies, as shown in Table 1: Table 1: Lookup table for energy control strategies corresponding to cumulative heat dose ranges Cumulative heat dose range (equivalent minutes) Risk level Interval Name Energy control strategy Strategy Explanation ≥0 and <60 Low safe zone Normal monitoring strategy Keep the current launch parameters unchanged ≥60 and <90 lower Attention Area Mild energy-saving strategy Slightly reduce the pulse repetition frequency while appropriately increasing the transmit voltage. ≥90 and <110 middle restricted area Moderate energy saving strategy Significantly reduce pulse repetition frequency and moderately reduce transmit voltage. ≥110 and <120 high Critical region Forced energy-saving strategy Significantly reduce pulse repetition frequency and transmit voltage ≥120 Extremely high Over-limit area Emergency shutdown strategy Suspend ultrasonic emission until the thermal dose drops back to a safe level. In this embodiment, the multi-parameter energy regulator performs the following decision-making process at a frequency of 1Hz: (1) Receive the current cumulative heat dose C(t) from the heat dose accumulation calculation unit and the predicted temperature rise trend curve for the next 30 seconds from the temperature rise trend prediction unit; (2) Determine the current risk level based on the interval where C(t) is located, and at the same time analyze the prediction curve to predict the interval where C(t) will be located in the next 30 seconds, and determine the predicted risk level. (3) Select the energy control strategy corresponding to the higher of the current risk level and the predicted risk level as the target strategy; (4) Generate a specific energy adjustment command based on the parameter combination corresponding to the target strategy. The command includes two fields: target transmission voltage value and target pulse repetition frequency value.

[0034] The above implementation methods, parameter settings, and decision-making mechanisms are all preferred examples that can be implemented by those skilled in the art based on the concept of this invention. In practical applications, the division of thermal dose intervals and strategy parameters can be optimized and adjusted according to the specific condition of the child, monitoring site, and other factors.

[0035] In one embodiment of the present invention, the adaptive beamforming and thermal dose control module further includes: a watchdog comparator and a forced interrupt switch; The watchdog comparator, connected to the multimodal signal acquisition module, is used to trigger a forced interrupt switch to cut off the power supply circuit of the piezoelectric micromechanical ultrasonic transducer when the real-time temperature data packet exceeds the preset overheating threshold. The forced interrupt switch is connected to the watchdog comparator, and its normally closed contact is connected in series in the power supply circuit of the beamforming controller and the piezoelectric micromechanical ultrasonic transducer.

[0036] It should be noted that, in this embodiment, the core function of the watchdog comparator is to perform real-time hardware-level monitoring of the temperature of the probe-skin contact surface. This is achieved using an analog voltage comparator chip, which requires no clock or program. Once the temperature exceeds the preset overheat threshold, an interrupt is immediately triggered.

[0037] In this embodiment, the hardware overheat threshold is set to 42°C. The selection criteria for this value are as follows: (1) According to IEC 60601-1 General Standard for Safety of Medical Electrical Equipment, the surface temperature of the part of the medical probe that comes into contact with the patient shall not exceed 43°C to prevent low-temperature burns; (2) Considering that pediatric patients have delicate skin and high heat sensitivity, the overheating threshold is set at 42℃ with a safety margin of 1℃ based on the general standard; The software-preset safety threshold is usually lower than the overheat threshold (for example, the software-preset safety threshold may be 40°C). The software first performs predictive adjustment, and the overheat threshold is only triggered when the software control fails and the temperature continues to rise to a dangerous level.

[0038] As one embodiment of the present invention, the breathing analysis module based on subpixel displacement further includes: an intelligent region of interest extraction unit and a pleural line edge enhancement unit. The region of interest (ROI) intelligent extraction unit is connected to the multimodal signal acquisition module. It has a pre-set deep learning-based thoracic structure recognition model, which is used to automatically identify and extract rectangular ROI containing the pleural line from the B-mode image frame of the digital echo data stream and generate the corresponding ROI mask. The pleural line edge enhancement unit is connected to the region of interest intelligent extraction unit. It integrates an anisotropic diffusion filter to perform smoothing, noise reduction, and edge enhancement processing only on B-mode image frames within the region of interest mask, generating an enhanced pleural line feature image sequence.

[0039] It should be noted that the thoracic structure recognition model used in this embodiment is based on a lightweight semantic segmentation network and consists of the following parts: Encoder: MobileNetV2 is used as the backbone network, which contains depthwise separable convolutional layers to extract image features layer by layer. The input is a 256×256 pixel B-mode image frame, and after passing through the encoder, a 16×16 pixel high-dimensional feature map is obtained. Decoder: Employs an upsampling and skip connection structure to gradually restore the resolution of the high-dimensional feature map, and finally outputs a segmentation result map of 256×256 pixels with the same size as the input; Output layer: The Sigmoid activation function is used to output a 0-1 range value representing the probability that each pixel belongs to the pleural line.

[0040] The training dataset for this thoracic structure recognition model comes from thoracic ultrasound images of 500 pediatric patients collected under ethical review and informed consent conditions. The dataset includes original images of the pleural line from patients of different ages, body types, and respiratory states. Senior physicians manually annotated the pleural line regions, and the annotation results were processed into a binary mask (pleural line region as 1, background as 0). The annotated original pleural line images were then subjected to random rotation, translation, scaling, and brightness adjustment to expand the training set.

[0041] In actual operation, the region of interest intelligent extraction unit performs the following operations: (1) Receive digital echo data stream from the multimode signal acquisition module, extract B-mode image frames and normalize their size to 256×256 pixels, and normalize the pixel values ​​to the range of 0-1; (2) Input the preprocessed image into the MobileUNet model for forward inference calculation and output a 256×256 pixel probability map, where the value of each pixel represents the probability that the pixel belongs to the pleural line; (3) Set the probability threshold to 0.5 and convert the probability map into a binary mask, that is, set the pixels with a probability ≥ 0.5 to 1 (pleural line area) and the rest to 0 (background area). (4) Mark the connected components of the binary mask, extract the connected component with the largest area as the region of interest, and eliminate false positive points caused by noise; (5) Calculate the minimum bounding rectangle of the region of interest, extend it outward by 10 pixels as a safety boundary, and generate the final rectangular region of interest mask. The mask is output to the pleural line edge enhancement unit in coordinate form: [x-coordinate of the upper left corner, y-coordinate of the upper left corner, width, height].

[0042] As one embodiment of the present invention, the breathing analysis module based on subpixel displacement further includes: an optical flow field calculation unit and a subpixel peak fitting unit. The optical flow field calculation unit is connected to the pleural line edge enhancement unit. It integrates a pyramid iterative optical flow algorithm to calculate the optical flow field of pixels in the region of interest mask between two consecutive B-mode image frames in the enhanced pleural line feature image sequence, and obtain the initial displacement vector of each pixel. The subpixel peak fitting unit, connected to the optical flow field calculation unit, is used to perform subpixel interpolation fitting on the local peak region representing the strong echo point of the pleural line in the initial displacement vector through its internally integrated Gaussian surface fitter, thereby generating a subpixel displacement vector sequence.

[0043] It should be noted that the pyramid iterative optical flow algorithm is an improved version of the Lucas-Kanade optical flow method. Its core idea is to construct an image pyramid and calculate the optical flow layer by layer from coarse to fine, effectively solving the tracking loss problem under large displacement motion. The basic principle of the algorithm is as follows: The original image is downsampled layer by layer to form an image pyramid with resolution from high to low. The bottom layer is the original image and the top layer is the image with the lowest resolution. Optical flow is calculated starting from the top layer (low resolution) and the result is used as the initial value for the next layer (high resolution). The process is refined layer by layer, and the optical flow estimate is converged to sub-pixel accuracy through iterative optimization in each layer. After completing the calculations for all three pyramid layers, the optical flow field calculation unit outputs the initial displacement vector (including x-direction and y-direction displacements, in pixels) for each pixel within the region of interest, as well as the matrix condition number or residual size corresponding to each displacement vector. The output format is represented by 16-bit fixed-point numbers.

[0044] The subpixel peak fitting unit obtains the enhanced image from the pleural line edge enhancement unit, extracts the gray values ​​of pixels within the region of interest, sets the gray value threshold to 80% of the maximum gray value, and filters out pixels above the threshold as candidate strong echo points. Non-maximum suppression is applied to the candidate strong echo points, that is, only the point with the largest gray value is retained as a strong echo point in the 3×3 local neighborhood. Adjacent strong echo points are grouped into the same connected domain, and the centroid coordinates of each connected domain are calculated as the representative points of the pleural line segment. Finally, the initial displacement vectors of these representative points are found from the output of the optical flow field calculation unit.

[0045] Gaussian surface fitting assumes that the gray-level distribution of the target point follows a two-dimensional Gaussian distribution. The precise center position is estimated by fitting the surface. The specific steps are as follows: extract a 5×5 pixel neighborhood window with each strong echo point as the center, take the natural logarithm of the gray-level value of each pixel in the window, convert the two-dimensional Gaussian function into a quadratic surface form, and calculate the sub-pixel coordinates of the surface vertex based on the fitted quadratic surface parameters. The subpixel coordinates of strong echo points are obtained by fitting a Gaussian surface to the current frame image. The above process is repeated for the next frame image to obtain the subpixel coordinates of the corresponding points. The subtraction of the two is the subpixel displacement vector of the point. Since the pleural line usually contains multiple strong echo points, the average displacement of each strong echo point is calculated, and points that deviate from the average displacement by more than 2 standard deviations are removed. Finally, the subpixel displacement vector sequence representing the overall movement of the pleural line is output.

[0046] As one embodiment of the present invention, the breathing analysis module based on subpixel displacement further includes: a breathing waveform reconstruction unit, a multi-peak motion separation unit, and a waveform confidence evaluation unit. The respiratory waveform reconstruction unit, connected to the subpixel peak fitting unit, is used to extract the projection component of the subpixel displacement vector sequence in the direction perpendicular to the pleural line through a displacement-time transformer, and arrange the projection component values ​​in the order of image frame acquisition time to generate preliminary respiratory waveform data reflecting the movement of the pleural line with respiration. The multi-peak motion separation unit connects the respiratory waveform reconstruction unit and the optical flow field calculation unit. It is used to construct a separation matrix based on the differences in the local motion direction of pixels in the optical flow field, and then separate the initial respiratory waveform data into respiratory motion components and heartbeat vibration components. The waveform confidence assessment unit, connected to the multi-peak motion separation unit, is used to calculate the similarity score between the respiratory motion component and the normal respiratory morphology template through its internal preset waveform morphology template, and encapsulate the similarity score, preliminary respiratory waveform data, respiratory motion component and heartbeat vibration component into refined respiratory waveform data.

[0047] It should be noted that in this embodiment, respiratory motion refers to the periodic up-and-down displacement of the pleural line as a whole with respiration. This is manifested as large continuous areas of pixels within the region of interest having a motion direction that is perpendicular to the pleural line and basically consistent, with a large amplitude and low frequency. In contrast, heartbeat vibration refers to the transmission of heartbeats to the chest wall, resulting in local areas (especially pleural points near the heart) being superimposed with tiny vibrations that are different from the direction of respiration. These vibrations are usually more chaotic in direction, with smaller amplitude and higher frequency.

[0048] This embodiment constructs a separation matrix and performs signal separation based on the above principles. The specific process is as follows: The initial single-channel respiratory waveform data was expanded into multidimensional data and used as input for independent component analysis: Select p representative points from the region of interest, which typically include points in the central region of the pleural line (dominated by respiratory movement), points near the edge of the pleural line close to the cardiac projection area (mixed respiratory and cardiac movement), and points in the cardiac projection area (dominated by cardiac vibration). Extract the motion vector time series of these representative points, including displacement in the x and y directions, from the optical flow field output by the optical flow field calculation unit; The preliminary respiratory waveform data is combined with the motion vector time series to form an m×n observation matrix X, where m is the number of observed signals, m=1+2p, and n is the length of the motion vector time series; The observation matrix X is subjected to mean removal and whitening to eliminate the second-order correlation between signals, resulting in the whitened matrix Z. With the goal of maximizing negative entropy, a fixed-point iterative algorithm is used to solve for the separation matrix from Z. The iteration is terminated when the change in the separation matrix between two consecutive iterations is less than a preset change threshold (e.g., 10^−6), thus obtaining the separation matrix W. The observation matrix X is transformed using the separation matrix W, and the independent components are calculated as: Y = WX, where each row of Y represents an independent signal source; A fast Fourier transform is performed on each independent component Y. Components with spectral energy concentrated in 0.2~0.8Hz (corresponding to 12~48 beats / min) are identified as respiratory motion components, and components with spectral energy concentrated in 1~3Hz (corresponding to 60~180 beats / min) are identified as heartbeat vibration components. Other components are considered noise or artifacts and are removed.

[0049] Furthermore, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0050] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0051] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0052] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0053] Finally, it should be noted that the above description represents a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make various improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A monitoring system for pediatric critical care based on ultrasound, characterized in that, include: Flexible ultrasonic sensor array, multimodal signal acquisition module, adaptive beamforming and thermal dose control module, and respiratory analysis module based on subpixel displacement; The flexible ultrasonic sensor array is used to transmit ultrasonic pulses to the child's chest cavity through a piezoelectric micromechanical ultrasonic transducer and receive the original echo signal of the pleural line movement, and to sense the physical temperature value of the probe-skin contact surface in real time through a miniature thermocouple sensor integrated in the transducer gap. The multimodal signal acquisition module is used to digitize the original echo signal and the physical temperature value respectively to generate a digital echo data stream and a real-time temperature data packet; The adaptive beamforming and thermal dose control module is used to dynamically adjust the transmission voltage and pulse repetition frequency of the piezoelectric micromechanical ultrasonic transducer based on the comparison result of the received real-time temperature data packet and the preset safety threshold, so as to control the real-time temperature rise of the contact surface within a safe range. The subpixel displacement-based respiratory analysis module is used to perform subpixel-level displacement vector tracking on strong echo points representing the pleural line in the digital echo data stream, and to generate refined respiratory waveform data based on the periodic differences of the displacement vectors.

2. The ultrasound-based pediatric intensive care monitoring system according to claim 1, characterized in that: The flexible ultrasonic sensor array further includes: a flexible substrate material layer, an array transducer patch layer, and a top encapsulation layer; The flexible substrate material layer is made of biocompatible medical silicone, and its lower surface has a biomimetic microstructure texture to fit the skin of young children and eliminate tiny air bubbles on the contact surface. The array transducer patch layer is composited on the upper surface of the flexible substrate material layer. The piezoelectric micromechanical ultrasonic transducers are embedded in the array transducer patch layer in a matrix form. The spacing between adjacent piezoelectric micromechanical ultrasonic transducers is an integer multiple of half the wavelength of the ultrasonic pulse, which is used to synthesize a phased beam when transmitting ultrasonic pulses to focus on the focal area where the pleural line is located. The top encapsulation layer covers the array transducer patch layer, and the miniature thermocouple sensor is attached to the inner surface of the top encapsulation layer, located in the acoustic blind zone of the piezoelectric micromechanical ultrasonic transducer.

3. The ultrasound-based pediatric intensive care monitoring system according to claim 2, characterized in that: The array transducer patch layer also integrates a time gain compensation circuit, which is connected to the piezoelectric micromechanical ultrasonic transducer and is used to perform gain compensation on the received original echo signal to counteract the attenuation of the ultrasonic signal in the thoracic tissue. The top encapsulation layer also integrates a temperature compensation reference circuit, which is connected to the miniature thermocouple sensor. This circuit receives the physical temperature value, generates a temperature reference voltage, and dynamically adjusts the compensation slope of the time gain compensation circuit based on the temperature reference voltage.

4. The ultrasound-based pediatric intensive care monitoring system according to claim 1, characterized in that: The multimodal signal acquisition module further includes: an ultrasonic simulation front end, a temperature acquisition unit, and a synchronous sampling clock generator; The ultrasonic simulation front end is connected to the piezoelectric micromechanical ultrasonic transducer, which is used to amplify and convert the original echo signal into digital data to generate a digital echo data stream. The temperature acquisition unit is connected to the miniature thermocouple sensor and is used to perform analog-to-digital conversion on the physical temperature value to generate a real-time temperature data packet. The synchronous sampling clock generator is connected to the ultrasonic simulation front end and the temperature acquisition unit, respectively. It is used to generate a first sampling clock for the ultrasonic simulation front end, generate a second sampling clock with an integer multiple frequency division relationship with the first sampling clock for the temperature acquisition unit, and generate a frame synchronization signal according to the frame period of the first sampling clock to align the digital echo data stream and the real-time temperature data packet in time.

5. The ultrasound-based pediatric intensive care monitoring system according to claim 1, characterized in that: The adaptive beamforming and thermal dose control module further includes: a thermal dose accumulation calculation unit and a temperature rise trend prediction unit; The heat dose accumulation calculation unit is connected to the flexible ultrasonic sensor array and the multimodal signal acquisition module respectively. It has a pediatric tissue heat dose equation preset inside, which is used to calculate the tissue heat dose accumulation value at the current moment based on the real-time temperature data packet and the current transmission voltage and pulse repetition frequency of the piezoelectric micromechanical ultrasonic transducer. The temperature rise trend prediction unit is connected to the heat dose accumulation calculation unit. It is used to predict the tissue temperature rise trend curve within a future preset time window based on the historical tissue heat dose accumulation value sequence and an internally integrated temperature rise predictor based on a time-series neural network. If it is determined that the tissue temperature rise trend within the future preset time window exceeds the safety threshold, the adjustment of the emission voltage and pulse repetition frequency is triggered.

6. The ultrasound-based pediatric intensive care monitoring system according to claim 5, characterized in that: The adaptive beamforming and thermal dose control module further includes: a multi-parameter energy regulator and a beamforming controller; The multi-parameter energy regulator is connected to the heat dose accumulation calculation unit, the temperature rise trend prediction unit, and the beamforming controller, respectively. It has a preset set of energy control strategies corresponding to different heat dose accumulation value ranges. It is used to dynamically select the target energy control strategy from the set of energy control strategies based on the received current heat dose accumulation value and the future temperature rise prediction curve, and generate the corresponding energy regulation command. The beamforming controller is connected to the multi-parameter energy regulator and the piezoelectric micromechanical ultrasonic transducer, and is used to synchronously adjust the transmission voltage and pulse repetition frequency of the piezoelectric micromechanical ultrasonic transducer according to the energy regulation command.

7. The ultrasound-based pediatric intensive care monitoring system according to claim 5 or 6, characterized in that: The adaptive beamforming and thermal dose control module further includes: a watchdog comparator and a forced interrupt switch; The watchdog comparator is connected to the multimodal signal acquisition module and is used to trigger the forced interrupt switch to cut off the power supply circuit of the piezoelectric micromechanical ultrasonic transducer when the real-time temperature data packet exceeds the preset overheating threshold. The forced interrupt switch is connected to the watchdog comparator, and its normally closed contact is connected in series in the power supply circuit of the beamforming controller and the piezoelectric micromechanical ultrasonic transducer.

8. The ultrasound-based pediatric intensive care monitoring system according to claim 1, characterized in that: The breathing analysis module based on subpixel displacement further includes: an intelligent region of interest extraction unit and a pleural line edge enhancement unit; The intelligent region of interest extraction unit is connected to the multimodal signal acquisition module. It has a deep learning-based thoracic structure recognition model pre-set inside, which is used to automatically identify and extract rectangular regions of interest containing pleural lines from the B-mode image frames of the digital echo data stream, and generate corresponding region of interest masks. The pleural line edge enhancement unit is connected to the region of interest intelligent extraction unit. It integrates an anisotropic diffusion filter to perform smoothing, noise reduction, and edge enhancement processing only on B-mode image frames within the region of interest mask, generating an enhanced pleural line feature image sequence.

9. The ultrasound-based pediatric intensive care monitoring system according to claim 8, characterized in that: The breathing analysis module based on subpixel displacement further includes: an optical flow field calculation unit and a subpixel peak fitting unit; The optical flow field calculation unit is connected to the pleural line edge enhancement unit. It integrates a pyramid iterative optical flow algorithm to perform optical flow field calculation on the pixels in the region of interest mask between two consecutive B-mode image frames in the enhanced pleural line feature image sequence, and to obtain the initial displacement vector of each pixel. The subpixel peak fitting unit is connected to the optical flow field calculation unit and is used to perform subpixel interpolation fitting on the local peak region representing the strong echo point of the pleural line in the initial displacement vector through the Gaussian surface fitter integrated inside it, so as to generate a subpixel displacement vector sequence.

10. The ultrasound-based pediatric intensive care monitoring system according to claim 9, characterized in that: The breathing analysis module based on subpixel displacement further includes: a breathing waveform reconstruction unit, a multi-peak motion separation unit, and a waveform confidence evaluation unit; The respiratory waveform reconstruction unit is connected to the subpixel peak fitting unit. It is used to extract the projection component of the subpixel displacement vector sequence in the direction perpendicular to the pleural line through a displacement-time transformer, and arrange the projection component values ​​in the order of image frame acquisition time to generate preliminary respiratory waveform data reflecting the movement of the pleural line with respiration. The multi-peak motion separation unit, connected to the respiratory waveform reconstruction unit and the optical flow field calculation unit, is used to construct a separation matrix based on the differences in the local motion direction of pixels in the optical flow field, thereby separating the preliminary respiratory waveform data into respiratory motion components and heartbeat vibration components. The waveform confidence assessment unit is connected to the multi-peak motion separation unit and is used to calculate the similarity score between the respiratory motion component and the normal respiratory morphology template through its internal preset waveform morphology template, and encapsulate the similarity score, the preliminary respiratory waveform data, the respiratory motion component and the heartbeat vibration component into refined respiratory waveform data.

Images