Pressure sore prevention early warning method and system based on millimeter wave radar

Abstract

The invention discloses a millimeter wave radar-based pressure sore prevention early warning method and system, and the method comprises the steps: collecting a radar echo signal of a bed body region through a millimeter wave radar, and generating point cloud data containing distance and speed information; whether the patient is in bed is judged; when it is judged that the patient is in bed, processing the point cloud data to track the trunk and limbs of the patient; based on the tracking result, large-amplitude body movement and small-amplitude body movement of the patient are recognized through a multi-feature fusion body movement algorithm, and the effectiveness of body position change is judged; and according to the body movement identification and validity judgment result, performing pressure sore risk assessment, and sending out early warning information when the risk exists. According to the method, the defects of a traditional contact sensor and artificial nursing are overcome. Complete non-contact monitoring is achieved, direct contact with the body of a patient is avoided, discomfort, oppression and skin stimulation are eliminated, the risk of cross infection is reduced, information transmission timeliness and response speed are improved, and support is provided for data-driven decision making and medical research.

Description

Technical Field

[0001] This invention relates to the fields of biomedical engineering and wireless sensing equipment technology, and in particular to a method and system for early warning of pressure ulcers based on millimeter-wave radar. Background Technology

[0002] Pressure ulcers, also known as bedsores, are tissue ulcers and necrosis caused by prolonged pressure on local tissues, obstructed blood circulation, leading to tissue ischemia, hypoxia, and malnutrition. This is a common and serious complication, especially in patients who are bedridden for extended periods, have limited mobility, or impaired sensory function (such as paralyzed, comatose, elderly, or intensive care patients).

[0003] Traditional preventative measures rely primarily on manual care, such as medical staff or family members turning and changing the patient's position every few hours. However, this manual intervention is labor-intensive and difficult to ensure strict adherence at night or when staff is insufficient. Furthermore, even when turning is performed, the effectiveness of the positional change cannot be guaranteed; sometimes only slight movements fail to truly relieve pressure on critical areas.

[0004] To address the limitations of manual care, several products for monitoring pressure ulcer risk have emerged on the market. These products typically employ contact sensors, such as pressure sensor pads and skin moisture sensors, to acquire physiological data about the patient's body. However, these existing technologies have the following significant drawbacks: Patient discomfort and potential risks: Contact sensors need to be placed directly under the patient's body, which may cause discomfort, especially for patients with sensitive skin or pressure sores. They also pose a potential risk of cross-infection due to prolonged contact.

[0005] Inability to determine the effectiveness of repositioning: This is one of the biggest drawbacks of current technology. Even if a change in patient position is detected, the system cannot determine whether it is an "effective" positional change sufficient to relieve stress. This forces healthcare professionals to rely on experience rather than obtaining systematic, quantitative feedback.

[0006] Therefore, there is an urgent need for a new pressure ulcer prevention technology that can achieve non-contact real-time monitoring and intelligently judge the effectiveness of each positional change, thereby completely solving the shortcomings of traditional manual care and existing passive monitoring technology, and achieving truly intelligent and efficient pressure ulcer prevention.

[0007] Therefore, there is a need for a method and system for early warning of pressure sores based on millimeter-wave radar. Summary of the Invention

[0008] In view of this, the purpose of the present invention is to provide a pressure ulcer prevention and early warning method and system based on millimeter-wave radar. The method uses millimeter-wave radar to monitor the body movement information of long-term bedridden people in a non-contact manner to provide pressure ulcer early warning.

[0009] To achieve the above objectives, the present invention provides the following technical solution: The pressure ulcer prevention early warning method based on millimeter-wave radar provided by this invention includes the following steps: Step S1: Collect radar echo signals of the bed area in a non-contact manner using millimeter-wave radar, and process the echo signals to generate point cloud data containing distance and velocity information; Step S2: Based on the point cloud data, determine whether the patient is in bed; Step S3: When it is determined that the patient is in bed, the point cloud data is processed to track the patient's torso and limbs; Step S4: Based on the tracking results, identify the patient's large and small body movements using a multi-feature fusion body movement algorithm, and determine the effectiveness of the positional changes; Step S5: Based on the results of body movement recognition and effectiveness judgment, conduct pressure ulcer risk assessment and issue early warning information when a risk exists.

[0010] Furthermore, in step S2, a bed area presence detection algorithm is used to determine whether the patient is in bed. This algorithm includes: Clustering and tracking of point clouds; Calculate the average z-axis height, average signal-to-noise ratio, and range variation over time of the tracked target point cloud; Based on the comprehensive judgment of the above calculation results, it is determined whether there is a human body in the bed area.

[0011] Furthermore, the tracking step in step S3 is a hierarchical human body tracking algorithm, including: The position of the torso is determined by identifying the point cloud of "fixed points" in the torso. Based on the position of the torso, the point cloud of other objects on the bed that can form clusters is tracked, and the limbs are identified according to their relative positional relationship with the torso. Based on the identified trunk and limb point clouds, the distance gate range of the patient can be inferred.

[0012] Furthermore, the multi-feature fusion volumetric motion algorithm in step S4 includes: The changes in the number of point clouds on the torso and limbs were statistically analyzed to make a preliminary judgment on whether large or small body movements had occurred. Calculate the 3D coordinate changes of the trunk point cloud cluster centers, the smooth curve changes of the distance information, and the changes of the average signal-to-noise ratio of the point cloud; When the above-mentioned changes exceed the set threshold, the body position change is determined to be a valid rollover; otherwise, it is determined to be an invalid rollover.

[0013] Furthermore, step S5 includes: If no significant body movement occurs within a preset time after the patient is first detected in bed, a pressure ulcer warning will be issued. When an invalid large-amplitude body movement is detected, an invalid turning warning is issued.

[0014] The pressure ulcer prevention early warning system based on millimeter-wave radar provided by the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned method.

[0015] The pressure ulcer prevention early warning system based on millimeter-wave radar provided by the present invention includes a sensor module, a data processing and analysis module, and an alarm and interaction module; The sensor module is a millimeter-wave radar, configured to acquire radar point cloud data of the patient in a non-contact manner in real time. The data processing and analysis module is configured to process the radar point cloud data to identify the patient's large and small body movements and determine the effectiveness of the body position changes. The alarm and interaction module is configured to send a warning message when it detects that the patient has not turned over for a long time or that turning over is ineffective.

[0016] Furthermore, the data processing and analysis module is integrated on the millimeter-wave radar board.

[0017] Furthermore, the alarm and interaction module sends warning information to the terminal device via a mini-program. The warning information includes at least one of the following: the time when the patient needs to be turned over, feedback on the effectiveness of the turning over, and the patient's historical position data.

[0018] Furthermore, the data processing and analysis module is also configured to execute a bed area presence detection algorithm to determine whether the patient is in bed.

[0019] The beneficial effects of this invention are as follows: This invention provides a pressure ulcer prevention and early warning method and system based on millimeter-wave radar. This method utilizes non-contact sensors such as millimeter-wave radar, completely avoiding direct contact with the patient's body and fundamentally eliminating the discomfort, pressure, and skin irritation caused by traditional contact sensors. This feature is particularly suitable for patients with sensitive skin, those already suffering from pressure ulcers, or those at high risk of infection, greatly improving patient comfort and safety. Simultaneously, the non-contact design avoids the cleaning and disinfection issues of the sensors themselves, reducing the risk of cross-infection. Furthermore, thanks to the characteristics of millimeter-wave radar, patient privacy is absolutely guaranteed. The system can automatically identify patient position and perform real-time risk assessment, transforming traditional manual, periodic checks relying on the experience and time of medical staff into intelligent, on-demand intervention. Based on a preset algorithm model, the system automatically issues turning suggestions and warnings, freeing caregivers from tedious, repetitive tasks, allowing them to devote more energy to more important nursing tasks. This not only greatly reduces the heavy workload of medical staff but also helps optimize the allocation of nursing resources and improve overall nursing efficiency.

[0020] Among its features, the quantitative assessment of turning effectiveness addresses the common problem in existing technologies where patients are turned but not properly. Through in-depth analysis of millimeter-wave radar data, the system can quantitatively and objectively determine the effectiveness of each manual turning, displaying the percentage of pressure transfer or the duration of new pressure zones. This real-time, quantitative feedback mechanism effectively guides caregivers to perform correct procedures, ensuring the true implementation of pressure ulcer prevention measures and significantly improving prevention outcomes.

[0021] This system can implement the above methods through mobile applications, clients, or WeChat mini-programs, and can be deployed on any terminal device (such as smartphones, tablets, and computers). This allows nursing staff to receive real-time warnings and alerts from the system immediately, regardless of whether they are at the bedside, at the nursing station, or anywhere on the ward floor. This cross-platform, anytime-and-where interactive method greatly improves the timeliness and accuracy of information transmission, ensuring that nursing staff can respond to potential risks to patients in the most timely manner and effectively avoid the risk of pressure ulcers caused by information delays.

[0022] This system automatically records and analyzes key data such as patient position changes, turning frequency, and effectiveness, generating detailed nursing reports. This data provides healthcare professionals with a scientific, data-driven basis for decision-making, helping them evaluate the effectiveness of nursing plans and adjust and optimize them according to individual patient conditions. The long-term accumulation of historical data also provides valuable information for medical research, promoting a deeper understanding of pressure ulcer prevention mechanisms.

[0023] Therefore, this method can achieve non-contact real-time monitoring and intelligently judge the effectiveness of each position change, thereby completely solving the shortcomings of traditional manual care and existing passive monitoring technology, and achieving truly intelligent and efficient pressure ulcer prevention.

[0024] This system utilizes millimeter-wave radar technology for non-contact human body monitoring. Compared to current products that rely on pressure sensors for pressure ulcer prevention and early warning, this technology offers the advantage of eliminating direct contact with the human body, thus avoiding the inconvenience and discomfort associated with traditional monitoring methods. Furthermore, millimeter-wave radar has a natural advantage in privacy protection. By transmitting and receiving high-frequency electromagnetic waves to generate anonymous abstract point clouds, it does not collect any optical image information, maximizing the protection of the dignity of elderly people who are bedridden. It is highly suitable for privacy-sensitive settings such as nursing homes and hospitals. This system aims to reduce the workload of medical staff and significantly improve the automation and effectiveness of pressure ulcer prevention.

[0025] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0026] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following drawings are provided for illustration.

[0027] Figure 1 This is a flowchart illustrating the overall implementation of this embodiment.

[0028] Figure 2 This is an installation diagram of an actual scenario in this embodiment.

[0029] Figure 3 This is a flowchart of the human body echo signal processing in this embodiment.

[0030] Figure 4 This is a flowchart of the detection algorithm for the bed area in this embodiment.

[0031] Figure 5 This is a diagram showing the smoothing result of the z-axis height of the target point cloud and the determination of whether the bed is on or off.

[0032] Figure 6 This is a flowchart of the hierarchical human body tracking algorithm in this embodiment.

[0033] Figure 7 This is a schematic diagram of the "fixed point" in this embodiment.

[0034] Figure 8 The flowchart is for the multi-feature fusion dynamic algorithm.

[0035] Figure 9 This is a diagram of a pressure ulcer prevention and early warning system based on millimeter-wave radar. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0037] Example 1 like Figure 1 As shown in the figure, the pressure ulcer prevention early warning method based on millimeter-wave radar provided in this embodiment can accurately monitor changes in body position without direct contact with the patient's body. This invention innovatively utilizes radar point cloud data and intelligent signal processing algorithms to accurately determine the effectiveness of assisted turning operations and achieve real-time assessment of pressure ulcer risk. Furthermore, users can receive real-time warnings of ineffective turning and real-time pressure ulcer risk assessment reports via a WeChat mini-program.

[0038] This embodiment provides a method and system for preventing pressure sores based on millimeter-wave radar. The main steps are as follows: Step 1: Use millimeter-wave radar to transmit and receive electromagnetic waves and collect echo signals in the current scene; process the signals through algorithms to obtain distance dimension information and 5D point cloud data (i.e., point cloud containing three-dimensional coordinates, velocity and signal-to-noise ratio information) in the current scene. A millimeter-wave radar is mounted on the wall above the bed, continuously emitting electromagnetic waves into the bed area, receiving the corresponding echoes, and transmitting them to the data processing module. The data processing module performs a Fourier transform in the range dimension to generate range data for the bed area. Specifically, it performs a Fast Fourier Transform on the echo sequence received by the radar to obtain the echo energy distribution on each range cell (i.e., the range cell arrangement based on the radar resolution), and then further processes it to obtain 5D point cloud data.

[0039] Step 1 has the following sub-steps: Step 1.1: like Figure 2As shown, the millimeter-wave radar in this embodiment is installed at a 45-degree angle to the horizontal plane (or wall). This ensures that the radar beam can effectively cover the entire bed area, thereby ensuring accurate capture of the patient's body movement signals (such as turning over, limb movement, etc.). The 45-degree angle helps reduce blind spots and improves the accuracy of signal acquisition. The radar is installed at a height of 0.8 meters to 1.2 meters above the ground to ensure that the radar can monitor the patient's trunk and limb movements from the optimal viewing angle. Installation height that is too high or too low may affect signal quality; therefore, this range was determined experimentally. In this embodiment, the radar module and data processing module are integrated (or closely connected). This not only collects echo signals but also performs preliminary data processing (such as filtering and point cloud generation). This integrated design simplifies the installation process and improves the system's real-time performance and reliability. Therefore, the detection device integrating the millimeter-wave radar module and data processing module is installed on the wall above the headboard, responsible for transmitting and receiving electromagnetic waves and performing subsequent algorithmic processing.

[0040] The main echo term of a millimeter-wave radar for a target with multiple scattering centers: in, Indicates the first Complex reflection coefficient of each scattering center; This indicates the total echo signal received by the radar from the target; Indicates the original transmitted signal After the first After reflection from each scattering center, a delay is introduced due to the different propagation paths. ; Indicates the first Doppler frequency shift of each scattering center; The time reference of the radar system is a continuous time variable referenced in the entire radar echo signal observation and processing. It is used to describe the signal acquisition, transmission and reception process of the system at any given moment.

[0041] Due to hardware leakage, the radar's transmitted signal directly enters the receiving channel, manifesting as a strong interference signal with no time delay and no Doppler effect, i.e., the antenna coupling term: in, This is the coupling coefficient, which is usually -30dB. Antenna coupling term refers to the interference component that directly enters the receiving channel due to hardware leakage in the radar system (such as direct path from transmission to reception or inter-antenna coupling). It is a strong interference component in the received signal that does not have time delay or Doppler frequency shift. This refers to the original signal emitted by the radar system.

[0042] Coupled echoes often have high amplitudes (potentially much higher than the target echo), which can affect radar's near-target detection.

[0043] When the target echo encounters objects such as buildings or the ground and is reflected back to the receiving antenna, multipath effects occur, leading to problems such as artifacts and blurred images. The multipath echo term is represented as: in, Indicates the multipath attenuation factor; This represents a multipath echo signal, which is the sum of all multipath echo components; Indicates the first The propagation delay of a multipath path is related to the path length and the propagation speed; This indicates Doppler in different directions due to multipath propagation. It's worth noting that the parameters have a superscript indicating the direction of the Doppler effect. The symbols used specifically refer to physical quantities related to multipath propagation paths, distinguishing them from the parameters of the main echo (direct path). This facilitates effective differentiation and targeted analysis during signal modeling, processing, and algorithm implementation. Because multipath paths are longer, there are... .

[0044] Environmental clutter is mostly generated by the ground, walls, or buildings. It is characterized by being approximately stationary and densely distributed, and can be modeled as a non-Gaussian random process. (Rayleigh distribution, K-distribution, etc.): in, It represents the clutter signal, which is the sum of clutter echoes generated by environmental objects such as the ground and walls in the receiving channel; Indicates the first The complex amplitude coefficients of each clutter component are random processes that vary with time (such as Rayleigh distribution, K-distribution, etc.), used to describe the intensity, phase variation, and statistical characteristics of the clutter. Taking Gaussian white noise into account The radar received signal model is as follows: in, This indicates the received signal. It represents the radar receiving channel in time. The sum of signals actually observed at any given moment. It includes all components such as target echoes, clutter, multipath, coupling, and noise, and is the input signal for radar signal processing.

[0045] Step 1.2: Based on the above echo, design as follows Figure 3 The human body echo signal processing flow shown is as follows: Echo data: Raw signals received by millimeter-wave radar; Range-dimensional FFT: A Fast Fourier Transform (FFT) is performed on each transmitted linear frequency modulated continuous wave (Chirp signal). Since echoes from targets at different distances have different time delays (i.e., phase differences), the FFT transforms the signal from the time domain to the frequency domain. The frequency directly corresponds to the distance to the target, thus obtaining the target's range information. This step generates the "range spectrum."

[0046] Clutter filtering: Filtering out static or slowly changing clutter (such as echoes from walls and furniture); Doppler FFT: Perform FFT on multiple consecutive chirp signals in each distance cell; Channel incoherent accumulation: Millimeter-wave radars typically have multiple receiving antennas (channels). The signals detected by multiple antenna channels are superimposed and combined. "Incoherent" accumulation is mainly used to improve the signal-to-noise ratio, so that weak targets that might otherwise be obscured by noise (such as limbs far from the radar) can be clearly detected.

[0047] OS-CFAR (Ordered Statistical Constant False Alarm Rate Detection) employs an adaptive threshold detection algorithm. It dynamically sets a detection threshold based on the level of background noise around the target. Only when the signal strength exceeds this threshold is the target considered a valid target. This effectively suppresses random noise interference, ensuring stable operation of the system in various environments and avoiding false alarms.

[0048] Angular-dimensional FFT: Utilizing the phase difference between multiple receiving antennas, FFT is performed on the azimuth (horizontal direction) and elevation (vertical direction) angles respectively. By calculating the phase difference of the same target on different antennas, the target's azimuth can be accurately calculated, achieving spatial positioning.

[0049] Target point cloud: After the above CFAR and angle FFT, the system has detected individual "points", each containing distance, velocity, and azimuth information. The collection of these points constitutes the preliminary "target point cloud".

[0050] Coordinate transformation: The point cloud is transformed from the radar's polar coordinate system (distance, angle) to a more intuitive Cartesian coordinate system (X, Y, Z). Combined with the existing velocity and signal-to-noise ratio values ​​of each point, a "5D point cloud data" containing five dimensions including position (X, Y, Z), velocity, and signal-to-noise ratio is finally generated.

[0051] Within each frame, the spacing between each transmitting antenna of the radar The time-transmitted linear sawtooth wave (i.e., linear continuous frequency modulated wave) emitted a total of Each chirp sample Point, to By sampling all signals in each channel with the same number of samples, the original echo signal matrix can be obtained. Then, for each chirp in each channel... A point-range FFT can be used to obtain the distance spectrum. This is then followed by inter-frame cancellation, i.e. Filter out static clutter.

[0052] in, This indicates the number of chirps (frequency-modulated pulses) transmitted. In one frame of data acquisition, the radar transmitted a total of [number missing] chirps. The chirp signal contains one linear frequency modulation process, and multiple chirps are used to achieve Doppler estimation and range resolution. This indicates the number of sampling points acquired in each chirp, i.e., the number of sampling points of the echo signal taken by the receiver within each chirp cycle; This indicates the number of channels, the number of receiving channels in a radar system (e.g., multiple transceiver antennas). Each channel is sampled independently, reflecting spatial multipath information, which helps in achieving functions such as angle estimation or beamforming. This indicates the data after MTI (Moving Target Indication) processing, which is the signal after inter-frame subtraction (frame difference filtering). It is mainly used to suppress stationary clutter (such as echoes from the ground or stationary obstacles). This represents the distance spectrum sequence of the current frame. This represents the distance spectrum sequence of the previous frame (frame-1).

[0053] Perform for each distance unit of each channel Point velocity-dimensional FFT yields the range-velocity spectrum. Further, multi-channel incoherent accumulation is used to improve the signal-to-noise ratio, enabling the detection of even weak targets.

[0054] The target, after having its signal-to-noise ratio improved through incoherent accumulation, is processed using ordered statistical constant false alarm rate (OS-CFAR) to obtain CFAR detection results. Subsequently, angular FFTs are performed in the azimuth and elevation directions, followed by peak detection using a super-resolution algorithm to obtain the corresponding angular cells. The obtained point cloud distance and velocity information, combined with the formula for converting polar coordinates to Cartesian coordinates, yields the 5D point cloud.

[0055] Step 2: After obtaining the data information in Step 1, determine whether the monitored object is on the bed using a bed area presence detection algorithm: Using distance information and 5D point cloud information, a bed area presence detection algorithm is continuously used to determine whether a patient is in bed. If no patient is detected, pressure ulcer monitoring is paused; otherwise, if a patient is detected, position changes, pressure ulcer warnings, and monitoring of the effectiveness of position changes are initiated.

[0056] The detection algorithm for the bed area in step 2 is described in detail below: The flowchart of the detection algorithm for the bed area is as follows: Figure 4 As shown.

[0057] Step 2.1: Perform DBSCAN clustering and EKF tracing on the point cloud obtained in Step 1; Step 2.2: Based on the clustering tracking results, calculate the average z-axis value of the point cloud contained within the target, and then... Figure 5 The smoothed average value shown can be used to preliminarily determine the status of getting on and off the bed. Then, by combining the average signal-to-noise ratio of the target point cloud and the change of the target distance over time, the presence of a human body in the bed area can be determined.

[0058] Step 3: After detecting the patient on the bed in Step 2, a hierarchical human body tracking algorithm is used to combine the distance direction with the 5D point cloud information. The hierarchical human body tracking algorithm accurately identifies and tracks the human torso and limbs. Using the fixed-point torso detection strategy proposed in this invention, a 5D point cloud is filtered to obtain the point cloud belonging to the torso. Then, based on the filtered point cloud, hierarchical clustering is performed on the point clouds in the environment to identify the point clouds belonging to the limbs. Finally, based on the filtered point cloud, distance information is filtered in reverse to obtain the distance and 5D point cloud information related to the bedridden patient.

[0059] The hierarchical human body tracking algorithm in step 3 is described in detail below: like Figure 6 As shown, Figure 6 This is a flowchart of a hierarchical human body tracking algorithm.

[0060] Step 3.1: For bedridden individuals, there are often conditions in the torso such as... Figure 7 The point cloud shown is a small-range reciprocating motion, that is, a point cloud that exists for more than 10 seconds within a single distance gate or for more than 20 seconds between two adjacent distance gates (hereinafter referred to as "fixed points"). By accumulating observations for several seconds, "fixed points" can be initially screened out. Then, by filtering out point clouds with low signal-to-noise ratios, "fixed points" belonging to the torso can be detected. Step 3.2: Using the geometric center of the detected "fixed point" as the cluster center and tracking center, clustering and tracking with a certain radius can obtain the position of the torso of the bedridden human body in real time; Step 3.3: Based on the torso position, continuously track other point clouds located on the bed that can form clustered targets. When their number exceeds a set threshold, determine whether they are hands or legs based on their relative position to the torso, and then filter out other irrelevant point clouds; Step 3.4: Based on the point cloud information filtered out in the above steps, the current distance range of the bedridden human body can be deduced, thereby more accurately verifying and tracking the location of the subject's chest cavity.

[0061] In this embodiment, the fixed point refers to the specific range of the small-range reciprocating motion.

[0062] Step 4: After completing the distance orientation and 5D point cloud filtering in Step 3, perform a multi-feature fusion motion algorithm on the filtered distance orientation information and 5D point cloud information. The multi-feature fusion motion algorithm accurately identifies and detects large-amplitude and small-amplitude motions of the target (such as coughing, waving hands, looking at a mobile phone, swinging legs, etc., which do not change the orientation of the torso. Compared with large-amplitude motions, the main difference lies in whether the contact surface between the torso and the bed changes. Large-amplitude motions indicate a change, while small-amplitude motions indicate no change). The multi-feature fusion motion algorithm utilizes features such as a smooth curve of distance over time, the 3D coordinates and velocity of the 5D point cloud cluster center to which the torso belongs, the regional velocity and signal-to-noise ratio of the filtered 5D point cloud, and the number of filtered point clouds. The algorithm detects and identifies whether the target undergoes large-amplitude body movements, small-amplitude movements such as coughing, waving, looking at a phone, or swinging its legs, and invalid large-amplitude body movements such as rolling over and then returning to the original position.

[0063] The multi-feature fusion volumetric motion algorithm in step 4 is described in detail below: like Figure 8 As shown, Figure 8 The flowchart is for the multi-feature fusion dynamic algorithm.

[0064] Step 4.1: Determine the size of body movement. Statistically count the number of torso points after filtering in Step 3. When large body movement occurs, the number of torso points increases significantly; when small body movement occurs, the number of torso points does not change significantly, while the number of limb points increases slightly. When the number of torso points does not change significantly, but the number of limb points increases slightly, exceeding the threshold, it is determined to be small body movement. When the number of torso points exceeds the threshold, it is determined to be a pre-large body movement state.

[0065] Step 4.2: Calculate the distance change of the 3D coordinates of the cluster tracking center obtained in Step 3, perform temporal smoothing on the distance information filtered in Step 3, and calculate the average signal-to-noise ratio of the human point cloud after filtering in Step 3; if the changes of the above three factors all exceed the set threshold, the pre-movement state is determined to be an effective turning over (the change in the contact area between the body and the bed reaches or exceeds 20%). ,in The surface area before the body moves. If the change in body contact area with the bed does not reach or exceed 20%, it is considered an invalid turn (the change in body contact area with the bed does not reach or exceed 20%). The three thresholds involved in the above multi-feature fusion body movement judgment method are all empirical thresholds, obtained through extensive practical experiments and data analysis, statistically summarized based on typical characteristics of body movement behavior in bed. Specifically: For the threshold of the change in the three-dimensional coordinate displacement of the cluster center, the threshold is set as a percentage based on the change in the three-dimensional coordinate displacement of the cluster center before and after the body movement. That is, when the percentage of the change in the three-dimensional coordinate displacement of the cluster center to the subject's height reaches a certain threshold... If the threshold is met or exceeded, the threshold is considered passed; otherwise, it is not passed. The threshold for the change in the average amplitude ratio of range-time information is set based on the percentage change in the average amplitude of range-time information over 40 frames before and after the movement. That is, the average amplitude change is greater than A movement is considered valid only if it occurs on time; otherwise, it is considered invalid. This threshold is based on statistical analysis and effectiveness verification of actual bedside movement behavior samples. For the average signal-to-noise ratio (SNR) threshold of the target point cloud, the threshold is empirically set based on the percentage change in the average SNR of the target point cloud. That is, the change in SNR of the target point cloud before and after body movement exceeds If the movement is valid, it is considered a valid body movement; otherwise, it is considered an invalid body movement. This threshold was also obtained through empirical statistics from a large number of experimental samples.

[0066] Step 5: The statistical system manages patient status around the clock. After the judgment is completed in Step 4, the judgment data is centrally managed and recorded; The statistical system performs 24 / 7 data analysis based on the results returned from steps 2 and 4, primarily tracking total bedtime, bedtime determination timestamps, timestamps of significant body movement, timestamps of minor body movement, and timestamps of invalid significant body movement. The system generates real-time reports based on this data and displays them to medical staff.

[0067] Step 6: Pressure ulcer warning and warning of ineffective large-amplitude body movement. While performing Step 5, the statistical results are analyzed, and intelligent management and reminders are issued based on different results. From the moment a patient is first detected in bed, if there is no significant movement within two hours, a pressure ulcer warning will be sent to the mini-program client, reminding medical staff to assist the patient in turning over; thereafter, a pressure ulcer warning will be sent after each significant movement if no significant movement occurs within two hours. When ineffective, large-amplitude body movements occur, an ineffective turning warning will be sent to the mini-program client, reminding medical staff that after the assisted turning, the patient still has a significant overlap of force application area and needs to be assisted turning again according to the standard procedure. The aforementioned "ineffective, large-amplitude body movements" refer to situations where the contact area between the user's body and the bed changes drastically. The above changes did not occur; however, within one minute after the physical activity ended, the contact area returned to its pre-activity size. Within this range, no continuous and effective positional changes were achieved. Although this type of movement met the criteria for judging the range of motion, it failed to achieve the nursing goal of preventing pressure ulcers due to the restoration of body position.

[0068] This example demonstrates how non-contact real-time monitoring and intelligent assessment of the effectiveness of each positional change can be achieved through the above steps. This completely solves the shortcomings of traditional manual care and existing passive monitoring technologies, enabling truly intelligent and efficient pressure ulcer prevention.

[0069] The data management, pressure ulcer early warning, and alarm in steps 5 and 6 are described in detail below: Upon receiving the judgment results transmitted in steps 2 and 4, the corresponding results and timestamps are recorded. By identifying each judgment result and calculating the interval between each timestamp, corresponding warnings or alarms are issued, and a pressure ulcer monitoring report is generated for subsequent research.

[0070] Example 2 like Figure 9 As shown, the pressure ulcer prevention early warning system based on millimeter-wave radar provided in this embodiment includes a sensor module, a data processing and analysis module, and an alarm and interaction module; The sensor module is a millimeter-wave radar, configured to acquire radar point cloud data of the patient in a non-contact manner in real time. The data processing module is configured to process the radar point cloud data to identify the patient's large and small body movements and determine the effectiveness of the positional changes; the data processing and analysis module is integrated on the millimeter-wave radar board.

[0071] The alarm and interaction module is configured to send a warning message when it detects that the patient has not turned over for an extended period of time or that turning over is ineffective. The alarm and interaction module sends the warning message to the terminal device via a mini-program. The warning message includes at least one of the following: the time when turning over is required, feedback on the effectiveness of turning over, and the patient's historical position data.

[0072] The data processing and analysis module in this embodiment has the function of intelligent analysis and decision-making. It is used to receive raw data and execute relevant algorithms in sequence, including 5D point cloud and distance information generation: processing the raw echo signal into 5D point cloud and distance data containing information such as target position and velocity; bed area patient presence detection: determining whether the patient is in bed. If not in bed, the process may be interrupted or looped; hierarchical human body tracking algorithm: after confirming that the patient is in bed, accurately tracking the patient's torso and limbs; multi-feature fusion body movement algorithm: based on the tracking results, intelligently identifying the type of body movement and judging the effectiveness of turning over.

[0073] The system provided in this embodiment employs a non-contact sensor module, specifically a millimeter-wave radar that transmits and receives high-frequency electromagnetic waves in real time. The received echo signals contain information such as the patient's body posture, movement amplitude, and minute displacements. The millimeter-wave radar can be installed on the bedside wall to continuously monitor the patient's body movement information in a non-contact manner, thereby eliminating the discomfort and potential risk of cross-infection associated with traditional contact sensors.

[0074] The data processing and analysis module is the core of this system. It receives real-time data streams from the millimeter-wave radar module and performs in-depth processing and analysis using preset algorithms. This algorithm can accurately identify large and small body movements of the patient and determine the time interval between two consecutive large body movements. By analyzing continuous millimeter-wave radar point cloud data, this module automatically determines the validity of the patient's positional changes. Furthermore, the data processing and analysis module is integrated into the millimeter-wave radar board, enhancing the system's practicality.

[0075] Based on data processing results, the alarm and interaction module sends real-time warning messages to medical staff via a mini-program when it detects that a patient has not been turned over for an extended period or that turning attempts are ineffective, indicating a risk of pressure ulcers. The warning messages include immediate feedback on the required turning time and the effectiveness of the turning, and also provide visual charts showing the frequency and time history of patient position changes. A pressure ulcer prevention report is also generated.

[0076] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A pressure ulcer prevention warning method based on millimeter wave radar, characterized by, Includes the following steps: Step S1: Collect radar echo signals of the bed area in a non-contact manner using millimeter-wave radar, and process the echo signals to generate point cloud data containing distance and velocity information; Step S2: Based on the point cloud data, determine whether the patient is in bed; Step S3: When it is determined that the patient is in bed, the point cloud data is processed to track the patient's torso and limbs; Step S4: Based on the tracking results, identify the patient's large and small body movements using a multi-feature fusion body movement algorithm, and determine the effectiveness of the positional changes; Step S5: Based on the results of body movement recognition and effectiveness judgment, conduct pressure ulcer risk assessment and issue early warning information when a risk exists. 2.The millimeter wave radar-based pressure ulcer prevention warning method of claim 1, wherein, In step S2, a bed area presence detection algorithm is used to determine whether a patient is in bed. The algorithm includes: Clustering and tracking of point clouds; Calculate the average z-axis height, average signal-to-noise ratio, and range variation over time of the tracked target point cloud; Based on the comprehensive judgment of the above calculation results, it is determined whether there is a human body in the bed area. 3.The millimeter wave radar-based pressure ulcer prevention warning method of claim 1, wherein, The tracking step in step S3 is a hierarchical human body tracking algorithm, including: The position of the torso is determined by identifying the point cloud of "fixed points" in the torso. Based on the position of the torso, the point cloud of other objects on the bed that can form clusters is tracked, and the limbs are identified according to their relative positional relationship with the torso. Based on the identified trunk and limb point clouds, the distance gate range of the patient can be inferred. 4.The millimeter wave radar-based pressure ulcer prevention warning method of claim 1, wherein, The multi-feature fusion motion algorithm in step S4 includes: The changes in the number of point clouds on the torso and limbs were statistically analyzed to make a preliminary judgment on whether large or small body movements had occurred. Calculate the 3D coordinate changes of the trunk point cloud cluster centers, the smooth curve changes of the distance information, and the changes of the average signal-to-noise ratio of the point cloud; When the above-mentioned changes exceed the set threshold, the body position change is determined to be a valid rollover; otherwise, it is determined to be an invalid rollover. 5.The millimeter wave radar-based pressure ulcer prevention warning method of claim 1, wherein, Step S5 includes: If no significant body movement occurs within a preset time after the patient is first detected in bed, a pressure ulcer warning will be issued. When an invalid large-amplitude body movement is detected, an invalid turning warning is issued.

6. A pressure ulcer prevention warning system based on millimeter wave radar, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method described in any one of claims 1 to 5.

7. A pressure ulcer prevention warning system based on millimeter wave radar, characterized by: Includes a sensor module, a data processing and analysis module, and an alarm and interaction module; The sensor module is a millimeter-wave radar, configured to acquire radar point cloud data of the patient in a non-contact manner in real time. The data processing and analysis module is configured to process the radar point cloud data to identify the patient's large and small body movements and determine the effectiveness of the body position changes. The alarm and interaction module is configured to send a warning message when it detects that the patient has not turned over for a long time or that turning over is ineffective.

8. The millimeter-wave radar-based pressure ulcer prevention warning system of claim 7, wherein, The data processing and analysis module is integrated on the millimeter-wave radar board.

9. The millimeter-wave radar-based pressure ulcer prevention warning system of claim 7, wherein, The alarm and interaction module sends warning information to the terminal device via a mini-program. The warning information includes at least one of the following: the time when the patient needs to be turned over, feedback on the effectiveness of the turning over, and the patient's historical position data.

10. The millimeter-wave radar-based pressure ulcer prevention warning system of claim 7, wherein, The data processing and analysis module is also configured to execute a bed area presence detection algorithm to determine whether the patient is in bed.

Images