Non-contact vital sign detection system and device based on optical sensing for emergency ICU (intensive care unit)

Abstract

The invention relates to the technical field of body temperature or heat measurement, in particular to an optical sensing-based non-contact vital sign detection system and device for an emergency ICU (intensive care unit), and the system introduces active thermal stimulation, spectrum infrared thermal response analysis, environmental factor real-time compensation and physiological thermal model fusion technologies. The accurate, stable and specific non-contact monitoring of the core body temperature and endogenous metabolic calorie of the patient is realized, the heat coupling obstacle is effectively overcome, the endogenous heat and the endogenous heat are distinguished, and the monitoring stability and precision are improved.

Description

Technical Field

[0001] This invention relates to the field of body temperature or heat measurement technology, and in particular to a non-contact vital signs detection system and device based on optical sensing in emergency ICUs. Background Technology

[0002] The goal of vital sign monitoring for patients in the Emergency Intensive Care Unit (ICU) is to obtain real-time and accurate physiological data to provide objective evidence for clinical diagnosis, treatment planning, and prognostic assessment. Non-contact vital sign monitoring technologies, with their advantages of being non-invasive, reducing the risk of cross-infection, and improving patient comfort, have become an important trend in research and application in this field. In early applications, infrared thermal imaging technology successfully solved the problems of cumbersome operation, potential patient discomfort, and even the risk of cross-infection associated with traditional contact temperature measurement methods (such as mercury thermometers and electronic thermometers). It has demonstrated significant efficiency advantages, particularly in large-scale screening or preliminary assessment of fever symptoms.

[0003] As emergency ICU settings place more stringent, precise, and dynamic demands on vital sign monitoring, the inherent characteristics of the aforementioned non-contact temperature measurement technologies based on simple infrared radiation reception or surface spectral analysis are gradually revealing their deep limitations in addressing new challenges. Summary of the Invention

[0004] The purpose of this invention is to provide a non-contact vital sign detection system and device based on optical sensing for emergency ICUs. By introducing active thermal stimulation, multispectral infrared thermal response analysis, real-time compensation for environmental factors, and physiological thermal model fusion technology, it solves the technical problem that existing technologies cannot overcome, which is that the non-contact monitoring results are inaccurate due to insufficient coupling between body surface temperature and core physiological thermal information and difficulty in distinguishing between endogenous and exogenous heat.

[0005] One aspect of the present invention provides a non-contact vital sign detection system based on optical sensing for emergency ICUs, comprising: The non-contact optical sensing module is used to actively thermally stimulate the preset monitoring area and simultaneously collect its multispectral infrared thermal radiation response data, visible light or near-infrared image data, and distance information. The environmental context awareness module is used to monitor key physical parameters in the emergency ICU environment in real time. The signal acquisition and preprocessing unit is used to receive the raw electrical signals from the non-contact optical sensing module and the environmental context awareness module, and to digitize, filter, amplify and perform preliminary feature extraction on them. The data fusion and dynamic thermal modeling unit is used to integrate data and, by establishing and solving physical / physiological thermal models, to achieve compensation for environmental disturbances, inference of deep physiological heat, and accurate estimation of core body temperature. The display and alarm unit is used to display the patient's core body temperature and metabolic heat production trends in real time. The control and communication unit coordinates the operation of all the aforementioned modules and enables secure data exchange with external systems.

[0006] In some embodiments, the non-contact optical sensing module includes: An active thermal stimulation submodule is used to emit precisely controlled thermal radiation energy into the preset monitoring area to induce a transient thermal response. The active thermal stimulation submodule is equipped with an infrared radiation emitter array, which consists of multiple high-power, narrow-band infrared light-emitting diodes. The emitter array focuses the light spot onto the preset monitoring area through an optical lens group, and the light spot diameter is adjustable from 2 mm to 10 mm. The spectral infrared detection submodule is used to collect transient multispectral infrared radiation data of the preset monitoring area at different time points before, during and after active thermal stimulation in real time. The spectral infrared detection submodule is equipped with an infrared detector array, which includes at least two infrared detectors with different spectral response characteristics. The acquisition frame rate of the detector array is adjustable from 50Hz to 500Hz.

[0007] In some embodiments, the data fusion and dynamic thermal modeling unit includes: The environmental interference compensation submodule, based on the real-time environmental parameters provided by the environmental context perception module, constructs a dynamic thermal balance model to mathematically model the instantaneous heat exchange process on the surface of the preset monitoring area, thus separating the influence of exogenous heat on the body surface temperature. The environmental interference compensation submodule uses an adaptive Kalman filter algorithm to introduce environmental parameters as state variables and update the thermal balance equation of the patient's body surface in real time, thereby separating pure physiological heat radiation from the environmentally affected measurement values. The output of the environmental interference compensation submodule is the environmentally corrected net heat radiation value of the body surface. The thermal response analysis submodule receives transient multispectral infrared radiation data provided by the spectral infrared detection submodule and stimulation parameters from the active thermal stimulation submodule. It analyzes the dynamic response of tissue to active thermal stimulation using an inverse heat conduction algorithm and a multilayer tissue heat conduction model. The physiological heat inference submodule receives deep tissue thermal parameters provided by the thermal response analysis submodule and, in conjunction with a patient-specific physiological heat model, infers changes in the patient's core body temperature and endogenous metabolic heat production. The physiological heat model is an improved version based on the Penness physiological model, and the accuracy of the real-time core body temperature estimate output by the physiological heat inference submodule is ±0.2℃. The motion compensation submodule uses image registration and feature tracking algorithms to detect and quantify the patient's head or chest movements in real time, and dynamically adjusts the monitoring areas of the spectral infrared detection submodule and the active thermal stimulation submodule.

[0008] In some embodiments, the dynamic heat balance model is based on convective heat transfer based on ambient temperature and local airflow velocity, radiative heat transfer based on background infrared radiation and ambient temperature, and evaporative heat dissipation based on ambient humidity and patient sweating. The adaptive Kalman filter algorithm updates the heat balance equation of the patient's body surface in real time by using environmental parameters as state variables to isolate the influence of exogenous heat on the body surface temperature. The inverse heat conduction algorithm uses finite element analysis to establish a multilayer biological tissue heat conduction model, which includes the epidermis, dermis, and subcutaneous fat layer. It considers the density, specific heat capacity, thermal conductivity, and blood perfusion rate of each tissue layer. The Levenberg-Marquardt algorithm is used to optimize the parameters of the measured transient temperature curve to inversely derive the effective thermal conductivity and blood perfusion rate of the deep tissue.

[0009] In some embodiments, the environment context awareness module includes: The ambient temperature sensor uses a high-precision platinum resistance temperature sensor with a measurement range of -10℃ to 60℃ and a measurement accuracy of ±0.1℃; the ambient humidity sensor uses a capacitive humidity sensor with a measurement range of 0% to 100% relative humidity and a measurement accuracy of ±2%RH; the local airflow sensor uses a hot-wire anemometer with a measurement range of 0.05m / s to 5m / s and a measurement accuracy of ±0.05m / s; and the background infrared radiation sensor uses a broadband infrared thermopile detector with a spectral response range of 2μm to 20μm.

[0010] Another aspect of the present invention provides a non-contact vital sign detection device based on optical sensing for emergency ICU, used to implement a non-contact vital sign detection system, including a robotic arm having at least six rotary joints, the range of motion of which can cover the patient monitoring area on the emergency ICU bed; the non-contact optical sensing module is mounted on the robotic arm.

[0011] In some embodiments, the control system of the robotic arm works in conjunction with the motion compensation submodule to adjust the posture and position of the non-contact optical sensing module in real time according to the patient's body position and minute movements.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The active thermal stimulation of this invention, combined with spectral infrared detection, analyzes the dynamic thermal response of tissue to controlled thermal stimulation, rather than relying solely on static surface radiation temperature, thus enabling a deeper understanding of the thermal properties of the tissue's deeper layers. The environmental context perception module and environmental interference compensation submodule of the present invention accurately isolate the influence of external environmental factors on body surface temperature, so that the measured temperature changes can more realistically reflect the patient's own physiological condition, thereby achieving accurate estimation of core body temperature and effectively avoiding the significant deviation between body surface temperature and core body temperature in the prior art. The physiological heat inference submodule of this invention can accurately identify and quantify these clinically significant endogenous metabolic heat production changes from aliased signals by solving complex models, providing a reliable basis for early warning of local inflammation, tissue necrosis or metabolic abnormalities. The motion compensation submodule of this invention can accurately track changes in patient position and minute movements, and make real-time posture adjustments through a robotic arm to ensure continuous alignment of the monitoring area and data consistency. At the same time, the environmental context awareness module continuously provides environmental parameters, which are dynamically corrected by the environmental interference compensation submodule to effectively offset the impact of environmental fluctuations on monitoring stability. The energy output of the active thermal stimulation submodule of this invention is precisely controlled and protected by multiple safety features to ensure no harm to the patient during monitoring. The linkage between the robotic arm's torque sensor and motion compensation submodule prevents misoperation and accidental collisions. In addition, the intelligent algorithm and adaptive correction function integrated into the system of this invention reduce the requirements for the professional skills of operators and improve the ease of use and reliability of the system. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a system block diagram of the present invention.

[0015] Figure 2 This is a diagram illustrating the usage scenario of the device of the present invention. Detailed Implementation

[0016] Implementation Overview In the ICU environment, patients are often in a complex situation involving ventilator-assisted breathing, continuous intravenous infusion, repositioning, and being surrounded by multiple medical devices. These external factors, such as ambient temperature fluctuations, local airflow disturbances, infusion temperatures, and even the heat dissipation of surrounding medical equipment and the radiant heat from monitoring lights, directly affect the local thermal balance of the patient's body surface. This leads to a significant deviation between the "surface temperature" obtained by technologies such as infrared thermal imaging and the patient's true physiological body temperature, especially their core body temperature. This deviation is not a simple systematic error, but rather the superposition of external interference signals and physiological signals at the radiation level. This causes temperature readings based solely on radiation intensity to lose their reliability and diagnostic specificity as core physiological indicators. For example, when a patient is undergoing active cooling or warming treatment, or when peripheral perfusion is insufficient due to shock, the gradient between their surface temperature and core body temperature will change drastically. In such cases, relying solely on surface temperature for judgment can easily lead to misdiagnosis and delayed treatment.

[0017] Furthermore, existing optical sensing methods face fundamental difficulties in distinguishing between endogenous metabolic heat production and exogenous heat absorption. Many pathophysiological processes in ICU patients, such as severe infectious inflammatory responses, metabolic abnormalities caused by organ failure, and even tumor necrosis, are accompanied by localized changes in metabolic heat production in specific tissues or organs. This type of heat production is a deep, endogenous physiological signal, and its changes often precede significant changes in body surface temperature. Traditional infrared thermal imaging technology primarily captures heat radiated from the skin surface, and its sensitivity and ability to distinguish weak thermal signals on the body surface caused by changes in metabolic heat in deep tissues are extremely limited. In other words, it is difficult to accurately isolate and identify these clinically significant subtle thermal changes caused by the patient's own physiological and pathological processes from various environmental noises and exogenous heat in the complex ICU environment.

[0018] Finally, the conditions of patients in the emergency ICU can change rapidly, and the monitoring of vital signs needs to reflect trends rather than instantaneous changes. Existing technologies are easily affected by factors such as changes in patient position, sweating, and skin dressings, making it difficult to guarantee the stability and consistency of continuous monitoring. Each interference may require recalibration or adjustment of the measurement position, which not only increases the workload of medical staff, but more importantly, may lead to the obscuring or misinterpretation of important physiological trends, thereby affecting the timeliness and accuracy of clinical decisions.

[0019] The following will be based on embodiments of the present invention. Figures 1-2 The technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0020] Example 1 The emergency ICU non-contact vital sign detection system based on optical sensing in this embodiment includes a non-contact optical sensing module, an environmental context awareness module, a signal acquisition and preprocessing unit, a data fusion and dynamic thermal modeling unit, a control and communication unit, and a display and alarm unit.

[0021] Specifically: The non-contact optical sensing module is used to actively thermally stimulate a preset monitoring area and simultaneously acquire its multispectral infrared thermal radiation response data, visible or near-infrared image data, and distance information. It can be understood that the non-contact optical sensing module includes an active thermal stimulation submodule, a spectral infrared detection submodule, a visible / near-infrared imaging submodule, and a distance measurement submodule. The core component of the active thermal stimulation submodule is an array of infrared radiation emitters, consisting of multiple high-power, narrow-band infrared light-emitting diodes (IR-LEDs) or semiconductor laser diodes. To achieve thermal stimulation of tissues at different depths, the center wavelengths of these IR-LEDs or semiconductor laser diodes are carefully selected to achieve different penetration depths and absorption characteristics in superficial skin tissues. Specifically, one set of emitters operates in the approximately 940nm wavelength range, with its peak absorption primarily concentrated in oxyhemoglobin in the blood, resulting in a relatively shallow penetration depth; while the other set of emitters operates in the approximately 1450nm wavelength range, with its energy mainly absorbed by water molecules in the tissue, resulting in a slightly deeper penetration depth. This dual-wavelength design enables the system to probe the thermal properties of different tissue layers.

[0022] Each unit of the transmitter array is precisely controlled by an independent drive circuit for current and pulse width modulation (PWM). This drive circuit, based on high-frequency switching power supply technology, enables pulse duration from nanoseconds to microseconds and peak power regulation from milliwatts to tens of watts, ensuring the accuracy, safety, and repeatability of the stimulation energy.

[0023] In practical implementation, a current source driven by a high-speed MOSFET can be used, combined with a precise digital pulse generator, to achieve, for example, a pulse width accuracy of 100ns and a peak power stability of ±0.1W. The transmitter array focuses the light spot onto the preset monitoring area through a multi-piece combined optical lens group. This lens group uses high-performance aspherical glass lenses, which have low aberration and high transmittance characteristics. By driving a micro stepper motor to adjust the lens spacing, the light spot diameter can be continuously adjusted within the range of 2mm to 10mm to adapt to different monitoring needs. For example, a 2mm light spot can be selected for monitoring a small area of ​​the face, while a 10mm light spot can be selected for monitoring a large area of ​​the chest.

[0024] The core of the spectral infrared detection submodule is an infrared detector array, which contains at least two infrared detectors with different spectral response characteristics. The first detector is a long-wave infrared (LWIR) focal plane array detector with a working wavelength range of 8μm to 14μm. It has a resolution of no less than 384×288 pixels and achieves a noise equivalent temperature difference (NETD) of less than 50mK, which enables it to accurately capture the temperature distribution and transient thermal response of the skin surface. In practical use, the first detector can be a detector based on uncooled vanadium oxide microbolometer technology. This technology operates at room temperature, has a compact structure, and is suitable for ICU environments.

[0025] The second detector is either a mid-wave infrared (MWIR) or short-wave infrared (SWIR) detector, with an operating wavelength range of 3μm to 5μm (MWIR) or 1μm to 2.5μm (SWIR). MWIR detectors typically use mercury cadmium telluride (MCT) or indium antimonide (InSb) materials, which are sensitive to moderate-depth thermal radiation; SWIR detectors mostly use InGaAs materials, which are more suitable for detecting changes in infrared radiation that penetrate superficial skin tissue to reach deeper tissues. The second detector is equipped with a resolution of no less than 320×256 pixels to ensure sufficient spatial information.

[0026] The detector array is equipped with custom-made infrared optical lenses made of high-transmittance materials such as germanium or chalcogenide glass and treated with multi-layer anti-reflection coatings to minimize energy loss. These lenses all have variable focal length capabilities, and the focal length can be continuously adjusted by precisely controlling the stepper motor, thereby ensuring that clear, high-contrast infrared images can be acquired at different monitoring distances from 0.5 meters to 2.0 meters. The acquisition frame rate of the detector array can be adjusted from 50Hz to 500Hz to capture the rapid thermal transient response caused by active stimulation.

[0027] As an internal calibration mechanism, the spectral infrared detection submodule also integrates a miniature blackbody calibration source. This calibration source is composed of a microporous structure made of a high-emissivity material, and its temperature is stabilized at preset values, such as 35°C and 40°C, by a high-precision PID controller. In this embodiment, the system can periodically or automatically point the detector at the blackbody source for two-point or multi-point radiation calibration when specific environmental conditions change, in order to compensate for the non-uniformity and drift of the detector itself, thereby ensuring the absolute accuracy of temperature measurement.

[0028] The core of the visible light / near infrared imaging submodule uses a high-resolution CMOS image sensor with a native resolution of no less than 1920×1080 pixels, which can capture rich detail information. The sensor has an adjustable frame rate function, ranging from 15 frames / second to 60 frames / second, to adapt to different application scenarios.

[0029] The visible / near-infrared imaging submodule is equipped with a dual-band optical filter for both visible and near-infrared light. This filter selectively allows the passage of visible light (400nm to 700nm) or near-infrared light (780nm to 900nm) through electrochromic or mechanical switching. The sensor mounts a high-performance zoom lens, such as a motorized zoom lens with a focal length range of 12mm-36mm, via a standard C / CS interface, enabling flexible scaling and precise focusing of the monitored area.

[0030] In addition, to provide uniform illumination and enhance image contrast and signal-to-noise ratio in near-infrared mode, the visible / near-infrared imaging submodule also integrates a set of narrowband near-infrared LED light sources, such as a matrix LED array with a center wavelength of 850nm. These LEDs provide uniform, spotless illumination through a diffuser.

[0031] The distance measurement submodule is used to accurately measure the distance between the non-contact optical sensing module and the preset monitoring area. It is equipped with a laser time-of-flight (TOF) sensor with a measurement range of 0.1 meters to 2.0 meters and a measurement accuracy of ±1 mm within this range. The TOF sensor calculates the distance by emitting a narrow pulse laser and accurately measuring the time it takes for the light signal to travel from emission to reflection from the target surface.

[0032] The environmental context awareness module is used to monitor key physical parameters in the emergency ICU environment in real time and provide necessary environmental interference parameters for the data fusion and dynamic thermal modeling unit. This module includes an environmental temperature sensor, an environmental humidity sensor, a local airflow sensor, and a background infrared radiation sensor.

[0033] The ambient temperature sensor uses a high-precision four-wire platinum resistance (Pt100) temperature sensor, which is connected to a high-resolution ADC through a precision Wheatstone bridge circuit. Its measurement range is -10℃ to 60℃, with a measurement accuracy of ±0.1℃ over this wide range, and it has long-term stability and high linearity.

[0034] The ambient humidity sensor is a capacitive humidity sensor with a measurement range of 0% to 100% relative humidity (RH) and a measurement accuracy of ±2%RH. This sensor derives humidity by measuring changes in the dielectric constant and integrates temperature compensation to ensure measurement accuracy under different ambient temperatures.

[0035] The local airflow sensor uses a hot-wire anemometer, whose core principle is to measure the change in the rate at which airflow carries away heat from the heating element. Its measurement range is 0.05 m / s to 5 m / s, with a measurement accuracy of ±0.05 m / s. This sensor is used to monitor local airflow velocities that may affect convective heat transfer on the patient's skin, such as minute airflows from air conditioning vents or caused by the movement of medical staff.

[0036] The background infrared radiation sensor employs a broadband infrared thermopile detector with a spectral response range of 2 μm to 20 μm. This detector has a wide field of view for measuring the intensity of background infrared radiation emitted from surrounding medical equipment, walls, ceilings, and other environmental objects. By accurately measuring this exogenous radiation, the system can separate it from the total radiation emitted from the patient's body surface, thus distinguishing exogenous radiant heat from the heat generated physiologically by the patient.

[0037] The signal acquisition and preprocessing unit is responsible for receiving analog signals from the non-contact optical sensing module and the environmental context awareness module, and digitizing, filtering, amplifying and performing preliminary feature extraction on them. The unit has multiple analog-to-digital converter (ADC) channels, which are used to convert analog signals from various sensors into digital signals.

[0038] This unit also integrates a low-noise amplifier circuit, a programmable gain amplifier (PGA), and an anti-aliasing filter. The low-noise amplifier circuit ensures the integrity of weak signals, the anti-aliasing filter prevents the introduction of high-frequency noise during sampling, and the PGA automatically adjusts the gain according to the signal strength to maximize the dynamic range utilization of the ADC.

[0039] It should be noted that the core processing capabilities of the signal acquisition and preprocessing unit are provided by an embedded field-programmable gate array (FPGA). This FPGA is responsible for performing real-time processing of high-speed data streams, including Bayer demosaicing, color correction, geometric distortion correction, image stabilization, and region of interest (ROI) extraction based on preset algorithms or medical personnel input for image sensor data; non-uniformity correction (NUC), bad pixel removal, and radiation intensity to temperature conversion based on calibration curves for infrared detector data; and high-precision synchronous acquisition and timestamping of all sensor data, as well as preliminary filtering and noise reduction processing.

[0040] The Data Fusion and Dynamic Thermal Modeling unit is responsible for integrating data from all sensor submodules and, by establishing and solving physical / physiological thermal models, achieving compensation for environmental disturbances, inference of deep physiological heat, and accurate estimation of core body temperature. The Data Fusion and Dynamic Thermal Modeling unit includes an environmental disturbance compensation submodule, a thermal response analysis submodule, a physiological heat inference submodule, and a motion compensation submodule.

[0041] The environmental interference compensation submodule constructs a dynamic thermal balance model based on the real-time environmental parameters provided by the environmental context perception module. The dynamic thermal balance model mathematically models the instantaneous heat exchange process on the surface of the preset monitoring area, thereby accurately separating the influence of exogenous heat on the body surface temperature measurement.

[0042] The dynamic heat balance model considers multiple heat exchange pathways, including convective heat transfer (based on ambient temperature and local airflow velocity, using Newton's law of cooling or its improved model), radiative heat transfer (based on background infrared radiation and ambient temperature, using the Steffen-Boltzmann law and its modifications), and evaporative heat dissipation (based on ambient humidity and assessment of patient sweating, skin moisture can be obtained from visible light image analysis, for example, indirectly inferred through skin gloss or reflectivity).

[0043] The environmental interference compensation submodule employs an adaptive Kalman filter algorithm, introducing environmental parameters (such as ambient temperature, humidity, airflow velocity, and background radiation intensity) as state variables to update the patient's surface heat balance equation in real time. The Kalman filter's prediction step predicts the surface temperature based on the current heat balance model, while the update step uses the difference between the measured infrared temperature and the predicted value to correct the state variables, thereby achieving an accurate estimate of the net surface heat radiation and effectively separating purely physiological heat radiation from environmentally influenced measurements. The output of the environmental interference compensation submodule is a sequence of environmentally corrected net surface heat radiation values ​​that reflects the patient's physiological condition.

[0044] The thermal response analysis submodule receives transient multispectral infrared radiation data from the spectral infrared detection submodule and stimulation parameters from the active thermal stimulation submodule. It then analyzes the dynamic response of tissue to active thermal stimulation using the inverse heat conduction algorithm and a multilayer tissue heat conduction model.

[0045] The specific workflow of the thermal response analysis submodule is as follows: First, before the active stimulation begins, baseline infrared radiation data is collected for a period of time (e.g., 5 seconds) to establish an initial thermal background reference for the stimulation area without external interference. Secondly, during active stimulation and for a period of time after stimulation (e.g., within 2 seconds after stimulation), multispectral infrared radiation sequences are continuously acquired at high speed to capture the complete transient response of tissue to stimulation. Furthermore, the acquired thermal image sequence is finely processed, for example, by averaging the pixels within a preset stimulation area, or by extracting the transient temperature curve of a specific core pixel. Then, a refined multilayer biological tissue heat conduction model is established using finite element analysis (FEA) or finite difference method (FDM). This model typically includes at least three typical tissue layers: the epidermis, the dermis (including the vascular network), and the subcutaneous fat layer, and can be extended to the muscle layer as needed. The model considers the density, specific heat capacity, thermal conductivity, and key physiological parameters such as blood perfusion rate of each tissue layer. Next, the energy inputs of the active thermal stimulation submodule (e.g., spot size, pulse duration, peak power density, wavelength, etc.) are used as boundary conditions for the model on the skin surface. The measured transient temperature curve, compensated for environmental disturbances, is used as the model output. By solving the inverse heat conduction problem, the equivalent thermal conductivity, blood perfusion rate of change, and potential local metabolic heat generation intensity of deep tissues are inverted. The inverse heat conduction problem can be solved using efficient numerical optimization algorithms, such as the gradient-based Levenberg-Marquardt algorithm or a population-based genetic algorithm for parameter optimization. The iterative convergence accuracy of this algorithm is set to 0.01℃ to ensure the accuracy of the inversion results. Spectral data are analyzed collaboratively in this stage through weighted averaging or independent modeling. The differences in penetration depth of different wavelengths of infrared light make it possible to analyze the transient thermal response of tissues at different depths. For example, 940nm stimulation mainly affects superficial blood flow, while 1450nm stimulation is more absorbed by water and affects tissue heat diffusion. Combining the responses of the two wavelengths can more comprehensively characterize the tissue thermal properties.

[0046] The physiological heat inference submodule receives deep tissue thermal parameters (such as equivalent thermal conductivity and blood perfusion rate of change) provided by the thermal response analysis submodule and auxiliary physiological parameters (such as heart rate, respiratory rate, and local blood perfusion index estimated by image processing) provided by the visible light / near infrared imaging submodule. Combined with a patient-specific physiological heat model, it infers the changes in the patient's core body temperature and endogenous metabolic heat production.

[0047] The physiological thermal model is an improved version of the Penness physiological model or Gagge two-compartment model. It divides the human body into core cavities (mainly including the trunk and head) and peripheral cavities (limbs and skin), and comprehensively considers blood circulation, cell metabolism, and various heat transfer pathways such as conduction, convection, radiation and evaporation. The model is trained using machine learning algorithms.

[0048] The physiological thermodynamic model takes as input the aforementioned inverted physical parameters and auxiliary physiological parameters, and outputs an estimated core body temperature and the metabolic heat production rate for a specific tissue region (e.g., the central chest, abdomen, or a specific lesion area designated by a physician). This metabolic heat production rate is typically expressed as heat flux density (W / m³). 2 The physiological thermal model initializes initial parameters based on the patient's age, gender, height, weight, and underlying medical history when the patient uses the system for the first time. In subsequent monitoring, it performs adaptive correction and optimization based on real-time acquired physiological data to improve the accuracy of personalized monitoring. The accuracy of the patient's real-time core body temperature estimate output by the physiological thermal inference submodule is ±0.2℃.

[0049] The motion compensation submodule continuously utilizes the continuous image sequence acquired by the visible / near-infrared imaging submodule to detect and quantify the patient's head or chest movements in real time through image registration and feature tracking algorithms. For example, feature point matching algorithms based on scale-invariant feature transform (SIFT) or accelerated robust feature transform (SURF) can be used, combined with the random sample consensus (RANSAC) algorithm to remove mismatched points, and then a Kalman filter is used to predict and smooth the motion trajectory of the target area. Once patient movement is detected, the motion compensation submodule quickly calculates the displacement and rotation of the target area and generates corresponding compensation commands. These commands are sent to the spectral infrared detection submodule and the active thermal stimulation submodule to dynamically adjust their monitoring areas, while simultaneously coordinating with the robotic arm to adjust the attitude and position of the optical sensing module to ensure that the monitored target remains centered in the field of view.

[0050] In addition, the motion compensation submodule also performs algorithmic compensation for measurement deviations caused by motion, such as reducing the distortion of the thermal response curve caused by pixel displacement through interpolation or weighted averaging. The response time of motion compensation is less than 50ms to adapt to the rapid and irregular movements that may occur in ICU patients.

[0051] The control and communication unit is equipped with a high-performance embedded controller, such as a multi-core processor based on the ARM Cortex-A series architecture, which runs a real-time operating system (RTOS) to ensure the priority scheduling and timing accuracy of various tasks. The controller communicates with sensors and actuators through various industry-standard interfaces such as SPI, I2C, UART, and USB to achieve precise power and pulse control of the active thermal stimulation submodule (e.g., controlling the drive current through a DAC), configuration of acquisition parameters (such as frame rate, exposure time, and gain) of the spectral infrared detection submodule, exposure and focus control of the visible / near-infrared imaging submodule, and triggering and data reading of the distance measurement submodule.

[0052] The control and communication unit also integrates a wireless LAN module compliant with the IEEE 802.11ac standard and a gigabit Ethernet interface compliant with the IEEE 802.3ab standard, for securely and reliably transmitting processed vital sign data and system status information to the central monitoring station or the hospital's electronic medical record (EHR) system.

[0053] The display and alarm unit is equipped with a 10.1-inch high-brightness IPS LCD screen with a resolution of 1120×800 pixels, ensuring clear images and accurate color reproduction. The screen presents data in an intuitive graphical interface, including real-time numerical display, historical trend graphs (such as 24-hour body temperature curves), color thermograms (displaying the transient thermal response of the stimulated area), and other auxiliary parameters (such as heart rate, respiratory rate, etc.).

[0054] The display and alarm unit also includes a three-color LED indicator (e.g., green for normal, yellow for warning, and red for emergency) and a built-in speaker, which is used to issue an audible and visual alarm when vital signs exceed the preset safety range. The alarm threshold can be personalized by clinicians according to the patient's condition. The alarm levels are divided into three levels: low, medium, and high, which are distinguished by different tones, frequencies, and LED colors.

[0055] To better understand this invention, a specific clinical example is provided below: Situation introduction: The system was deployed in the emergency ICU ward to monitor the core body temperature and local metabolic heat production changes in a patient with severe pneumonia. The patient was on mechanical ventilation and was in a relatively stable position, but still had slight trunk swaying and spontaneous breathing.

[0056] Implementation process: First, the visible / near-infrared imaging submodule locates the monitoring area at the fourth intercostal space on the left side of the patient's chest, corresponding to the lung lesion. The robotic arm adjusts the non-contact optical sensing module to a distance of 0.8 meters from the patient's chest. The spectral infrared detection submodule acquires a 10-second baseline thermal image sequence, while the environmental context awareness module simultaneously records environmental parameters: ambient temperature 24.5℃, relative humidity 55%, local airflow velocity 0.15m / s, and background infrared radiation intensity 350W / m². 2 .

[0057] Subsequently, the active thermal stimulation submodule stimulates the region with the following parameters: 940nm infrared LED array, pulse duration 150ms, peak power 4W, repetition frequency 1Hz, lasting for 3 seconds. The spectral infrared detection submodule synchronously acquires transient multispectral infrared thermogram sequences during stimulation and within 2 seconds after stimulation at a frame rate of 250Hz. The signal acquisition and preprocessing unit synchronizes, filters, and performs NUC correction on the data, and extracts the average transient temperature curve of the stimulated region.

[0058] The environmental disturbance compensation submodule in the data fusion and dynamic thermal modeling unit uses Kalman filtering to perform environmental correction on the original thermal image sequence. After correction, the thermal response analysis submodule inputs the net thermal radiation data into a preset multilayer tissue heat conduction model (including epidermis, dermis, subcutaneous fat, and superficial pectoral muscle), and uses the Levenberg-Marquardt algorithm to invert the equivalent thermal conductivity and blood perfusion rate of change in the region. In this embodiment, the equivalent thermal conductivity of the region is inverted to 0.48 W / (m·K), and the blood perfusion rate of change is 15% higher than the baseline. The physiological heat inference submodule 17 combines the patient's age (65 years), gender (male), weight (70 kg), and heart rate (85 beats / min) and respiratory rate (18 breaths / min) extracted from the visible light image, and uses a physiological thermal model based on a deep neural network to infer that the patient's core body temperature is 38.6℃, and estimates the endogenous metabolic heat production rate of the local monitoring area to be 55 W / m². 2 The motion compensation submodule continuously monitored the patient's minute movements throughout the process and coordinated with the robotic arm to make five minor posture adjustments to ensure stable alignment of the monitoring area.

[0059] The entire monitoring and processing cycle takes 4.8 seconds. The display and alarm unit shows the core body temperature and metabolic heat production trend in real time, and has a set high-temperature alarm of 38.5℃. When the core body temperature reaches 38.6℃, the system issues a yellow warning.

[0060] Example 2 Based on the same inventive concept as the optical-sensor-based non-contact vital sign detection system for emergency ICUs in Embodiment 1 above, this invention also provides an optical-sensor-based non-contact vital sign detection device for emergency ICUs. This device, used to implement the non-contact vital sign detection system, includes a robotic arm with at least six rotary joints, whose range of motion covers the patient monitoring area on the emergency ICU bed. A non-contact optical sensing module is mounted on the robotic arm. The robotic arm's control system and motion compensation submodule work in conjunction to adjust the posture and position of the non-contact optical sensing module in real time according to the patient's position and minute movements.

[0061] Specifically, the non-contact optical sensing module is mounted on a multi-degree-of-freedom robotic arm with at least six rotational joints, such as a serial or parallel robotic arm structure, with a range of motion sufficient to cover the patient monitoring area on the emergency ICU bed, including multiple key areas such as the head, chest, and abdomen.

[0062] The robotic arm is driven by a high-precision servo motor, and each joint is equipped with a high-resolution encoder for position feedback. Its positioning accuracy is better than ±0.5mm, and its repeatability is better than ±0.1mm.

[0063] The robotic arm's control system works in conjunction with the motion compensation submodule to adjust the posture and position of the sensing modules in real time based on the patient's positional changes and minute movements, ensuring continuous and accurate monitoring. When the patient turns over, the robotic arm can automatically adjust itself into position within seconds. To ensure patient safety, the robotic arm also integrates high-sensitivity torque sensors at all joints. Once an accidental collision or contact force exceeding a preset threshold is detected, the robotic arm can immediately stop all movement and enter a safety mode, while simultaneously issuing an alarm signal.

[0064] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0065] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A non-contact vital signs detection system based on optical sensing for emergency ICU, characterized in that, include: The non-contact optical sensing module is used to actively thermally stimulate the preset monitoring area and simultaneously collect its multispectral infrared thermal radiation response data, visible light or near-infrared image data, and distance information. The environmental context awareness module is used to monitor key physical parameters in the emergency ICU environment in real time. The signal acquisition and preprocessing unit is used to receive the raw electrical signals from the non-contact optical sensing module and the environmental context awareness module, and to digitize, filter, amplify and perform preliminary feature extraction on them. The data fusion and dynamic thermal modeling unit is used to integrate data and, by establishing and solving physical / physiological thermal models, to achieve compensation for environmental disturbances, inference of deep physiological heat, and accurate estimation of core body temperature. The display and alarm unit is used to display the patient's core body temperature and metabolic heat production trends in real time. The control and communication unit coordinates the operation of all the aforementioned modules and enables secure data exchange with external systems.

2. The system according to claim 1, characterized in that, The non-contact optical sensing module includes: An active thermal stimulation submodule is used to emit precisely controlled thermal radiation energy into the preset monitoring area to induce a transient thermal response. The active thermal stimulation submodule is equipped with an infrared radiation emitter array, which consists of multiple high-power, narrow-band infrared light-emitting diodes. The emitter array focuses the light spot onto the preset monitoring area through an optical lens group, and the light spot diameter is adjustable from 2 mm to 10 mm. The spectral infrared detection submodule is used to collect transient multispectral infrared radiation data of the preset monitoring area at different time points before, during and after active thermal stimulation in real time. The spectral infrared detection submodule is equipped with an infrared detector array, which includes at least two infrared detectors with different spectral response characteristics. The acquisition frame rate of the detector array is adjustable from 50Hz to 500Hz.

3. The system according to claim 1, characterized in that, The data fusion and dynamic thermal modeling unit includes: The environmental interference compensation submodule, based on the real-time environmental parameters provided by the environmental context perception module, constructs a dynamic thermal balance model to mathematically model the instantaneous heat exchange process on the surface of the preset monitoring area, thus separating the influence of exogenous heat on the body surface temperature. The environmental interference compensation submodule uses an adaptive Kalman filter algorithm to introduce environmental parameters as state variables and update the thermal balance equation of the patient's body surface in real time, thereby separating pure physiological heat radiation from the environmentally affected measurement values. The output of the environmental interference compensation submodule is the environmentally corrected net heat radiation value of the body surface. The thermal response analysis submodule receives transient multispectral infrared radiation data provided by the spectral infrared detection submodule and stimulation parameters from the active thermal stimulation submodule. It analyzes the dynamic response of tissue to active thermal stimulation using an inverse heat conduction algorithm and a multilayer tissue heat conduction model. The physiological heat inference submodule receives deep tissue thermal parameters provided by the thermal response analysis submodule and, in conjunction with a patient-specific physiological heat model, infers changes in the patient's core body temperature and endogenous metabolic heat production. The physiological heat model is an improved version based on the Penness physiological model, and the accuracy of the real-time core body temperature estimate output by the physiological heat inference submodule is ±0.2℃. The motion compensation submodule uses image registration and feature tracking algorithms to detect and quantify the patient's head or chest movements in real time, and dynamically adjusts the monitoring areas of the spectral infrared detection submodule and the active thermal stimulation submodule.

4. The system according to claim 3, characterized in that, The dynamic heat balance model is based on convective heat transfer based on ambient temperature and local airflow velocity, radiative heat transfer based on background infrared radiation and ambient temperature, and evaporative heat dissipation based on ambient humidity and patient sweating. The adaptive Kalman filter algorithm updates the heat balance equation of the patient's body surface in real time by using environmental parameters as state variables, so as to isolate the influence of exogenous heat on the body surface temperature. The inverse heat conduction algorithm uses finite element analysis to establish a multilayer biological tissue heat conduction model, which includes the epidermis, dermis, and subcutaneous fat layer. It considers the density, specific heat capacity, thermal conductivity, and blood perfusion rate of each tissue layer. The Levenberg-Marquardt algorithm is used to optimize the parameters of the measured transient temperature curve to inversely derive the effective thermal conductivity and blood perfusion rate of the deep tissue.

5. The system according to claim 1, characterized in that, The environment context awareness module includes: The ambient temperature sensor uses a high-precision platinum resistance temperature sensor with a measurement range of -10℃ to 60℃ and a measurement accuracy of ±0.1℃; the ambient humidity sensor uses a capacitive humidity sensor with a measurement range of 0% to 100% relative humidity and a measurement accuracy of ±2%RH; the local airflow sensor uses a hot-wire anemometer with a measurement range of 0.05m / s to 5m / s and a measurement accuracy of ±0.05m / s; and the background infrared radiation sensor uses a broadband infrared thermopile detector with a spectral response range of 2μm to 20μm.

6. An optically-sensor-based non-contact vital sign detection device for emergency ICUs, used to implement the non-contact vital sign detection system as described in any one of claims 1-5, characterized in that, It includes a robotic arm with at least six rotary joints, whose range of motion can cover the patient monitoring area on the emergency ICU bed; the non-contact optical sensing module is mounted on the robotic arm.

7. The apparatus according to claim 6, characterized in that, The control system of the robotic arm works in conjunction with the motion compensation submodule to adjust the posture and position of the non-contact optical sensing module in real time according to the patient's body position and minute movements.

Images