Pressure injury risk closed-loop intervention method and system based on cumulative injury assessment

Abstract

The invention discloses a pressure injury risk closed-loop intervention method and system based on cumulative injury assessment. The method comprises the following steps: collecting multi-point time sequence data of a bed surface, introducing a tissue vulnerability space coefficient matrix and a self-repairing attenuation item to calculate a cumulative pressure damage dose PTD, calculating a cumulative shear damage dose STD by combining a shear direction weighting function, and fusing to obtain a comprehensive cumulative damage index CDI; on the basis of CDI distribution, a micro turning control sequence is generated through multi-target constraint optimization to drive the partition controllable supporting units to execute; after execution, data are collected again to calculate the mitigation effect index, when the mitigation effect index does not reach the standard, the control sequence is updated, execution is performed again, and closed-loop intervention based on effect verification is formed. According to the method, accumulated damage modeling and closed-loop verification are deeply coupled, the technical span from passive alarm to active intervention-verification-re-optimization is realized, the bedsore prevention effect is remarkably improved, and the nursing burden is reduced.

Description

Technical Field

[0001] This invention belongs to the field of medical and nursing equipment and intelligent sensing technology, specifically relating to a closed-loop intervention method and system for pressure injury risk based on cumulative damage assessment, which can be applied to scenarios such as pressure ulcer prevention, rehabilitation nursing, and intensive care for long-term bedridden patients. Background Technology

[0002] Pressure ulcers (pressure injuries) pose a serious health threat to patients who are bedridden for extended periods. Statistics show that the incidence of pressure ulcers in hospitalized patients is 5%–15%, and the incidence in intensive care unit patients is as high as 30% or more. Pressure ulcers not only cause pain, prolong hospital stays, and increase medical costs, but in severe cases, they can also lead to life-threatening complications such as sepsis.

[0003] The existing techniques for preventing pressure ulcers mainly have the following problems: 1. Oversimplification of evaluation indicators Current smart mattresses mostly use "current pressure value" or "pressure-time integral" as risk assessment indicators. However, pressure ulcer formation is the result of multiple factors, including: continuous pressure leading to local ischemia; shear force damaging microvessels and lymphatic vessels; insufficient micromovement leading to the accumulation of metabolic waste in tissues; and microenvironmental factors such as local temperature and humidity. Relying solely on pressure monitoring cannot accurately assess the true risk of injury, resulting in high false alarm and false negative rates.

[0004] 2. Lack of individualized assessment Existing systems typically use fixed thresholds (such as "pressure > 32 mmHg for 2 hours"), without considering individual patient differences. In reality, the tissue tolerance of diabetic patients, elderly patients, and malnourished patients is significantly lower than that of healthy individuals, and using a uniform standard can lead to high-risk patients not receiving timely intervention.

[0005] 3. Passive alarms lack closed-loop verification. Existing technologies mostly operate on a passive "measurement-judgment-alarm" model, notifying nurses to manually turn the patient when pressure exceeds the limit. However, this approach has the following problems: delayed nurse response (especially at night); inconsistent quality of manual turning; inability to verify whether turning truly relieves pressure in high-risk areas; and lack of traceable records of intervention effectiveness.

[0006] Some smart mattresses have an automatic alternating air inflation function, but most of them use a fixed cycle (such as airing once every 10 minutes) and do not dynamically adjust according to actual needs, and there is no mechanism to verify the effectiveness.

[0007] 4. Ignoring the cumulative damage effect Tissue damage is a cumulative process. Both short-term high pressure and long-term moderate pressure can lead to pressure ulcers, and human tissue has a certain self-repairing ability. Current technology lacks quantitative modeling of "damage accumulation" and "natural recovery," making it impossible to accurately predict the time window for damage occurrence.

[0008] 5. Lack of predictive early warning Existing systems mostly only issue alarms after "pressure has exceeded the limit," which is a delayed response. They lack the ability to identify "early signs of injury" and cannot intervene in advance. To address these issues, there is an urgent need for a pressure ulcer prevention system that can comprehensively assess cumulative damage from multiple factors, support individualized threshold learning, and has the capabilities for proactive intervention and closed-loop verification. Summary of the Invention

[0009] Technical problems to be solved This invention aims to overcome the aforementioned deficiencies of the prior art and provide a closed-loop intervention method and system for stress injury risk based on cumulative damage assessment, solving the following technical problems: How to establish a cumulative damage assessment model that better reflects the pathological mechanism of pressure ulcers, taking into account multiple factors such as pressure, shear, tissue differences, and self-repair; How to achieve adaptive learning of individualized risk thresholds and dynamically adjust them based on patient characteristics and clinical feedback; How to construct a closed-loop mechanism for proactive intervention and effect verification to ensure that each intervention achieves the expected relief effect; How to achieve predictive warning sign identification and intervene in advance before damage occurs; How to ensure data consistency and system robustness in a multi-sensor fusion environment. Technical solution

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: Methodological aspects: This paper presents a method for pressure ulcer risk assessment and closed-loop decompression intervention, comprising three core steps: data collection, cumulative damage calculation, and closed-loop intervention control. (1) Multi-source data acquisition and consistency verification (S1) Collect time-series data from multiple points on the bed surface, including at least pressure signals, and optionally shear, volumetric motion, and temperature signals. Perform time synchronization and consistency checks on the multi-source data, and downweight or remove abnormal data.

[0011] (2) Cumulative damage modeling (S2-S3) Introducing a three-layer innovation to establish a cumulative damage model: Spatial dimension: Construct a tissue vulnerability coefficient matrix K_tissue(x,y), assigning higher weights (1.2-1.6) to high-risk areas such as the sacrum, coccyx, and heels, and lower weights (0.6-0.8) to areas with abundant muscle, to characterize the tolerance differences of different parts of the human body; Time dimension: A self-repair attenuation term exp(-λ·Δτ) is introduced, where λ is the tissue self-repair rate parameter and Δτ is the time interval since the last effective decompression. This term characterizes the physiological mechanism by which damage in human tissue naturally resolves after decompression. Multifactor fusion: The cumulative pressure damage dose (PTD) and cumulative shear damage dose (STD) are calculated separately, and combined with the microcirculation perfusion index (I_perfusion, which is estimated by the pressure recovery rate and temperature change rate after depressurization), the weighted fusion is used to obtain the comprehensive cumulative damage index (CDI).

[0012] Mathematical expression:

[0013]

[0014]

[0015] (3) Closed-loop intervention control (S4-S6) High-risk areas are identified based on the spatial distribution of CDI, and micro-turning control sequences are generated through multi-objective constraint optimization (objective: pressure reduction in high-risk areas ≥60%; constraints: turning angle ≤15°, new pressure in non-high-risk areas <40mmHg, etc.).

[0016] Core innovation: After intervention, data is collected again to calculate the mitigation effect index. If the target is not met, the control strategy is updated and re-implemented, forming a closed loop of "intervention-verification-re-optimization" to ensure that each intervention achieves the expected effect. If the target is still not met after consecutive attempts, the system will upgrade to a manual turning alarm.

[0017] (4) Individualized threshold self-learning (optional enhancement) Baseline risk correction coefficients are calculated based on patient age, disease (diabetes, vascular disease), and activity level. Skin pressure recovery rate (γ_recovery) is continuously monitored; a decrease in this index indicates reduced tissue elasticity. Nurses' skin assessment feedback (no abnormalities / stage I erythema / stage II ulceration) is received, and individualized thresholds are dynamically adjusted using a supervised learning algorithm.

[0018] (5) Predictive precursor identification (optional enhancement) Extract time-series features such as pressure fluctuation rate, temperature gradient trend, pressure relief recovery rate, duration of continuous inactivity, and CDI change rate. Use gradient boosting decision tree or neural network to train a prediction model, output the risk probability of skin damage in the next 1-3 hours, and issue early warnings and trigger interventions based on probability classification.

[0019] System level: A pressure ulcer risk assessment and closed-loop decompression intervention system is provided, including: Sensing components: array of pressure, shear, body motion, and temperature sensors; Zoned controllable support unit array: 20-100 independently controllable fluid units (airbags / liquidbags), each unit covering 50-200 cm²; Controller: Executes the above methods to realize functions such as cumulative damage calculation, closed-loop control, threshold learning, and precursor prediction; Storage: Records traceable data such as the evolution of CDI throughout time, intervention records, and clinical feedback; Human-computer interaction module: Displays CDI heatmap and early warning signals, receives nurse feedback, and outputs intervention suggestions for beneficial effects. Compared with the prior art, the present invention has the following significant advantages: 1. Cumulative damage modeling is more consistent with pathological mechanisms. Introducing an organizational vulnerability space coefficient and using different assessment standards for different parts is more scientific than a uniform threshold. Introducing a self-repair attenuation term reflects the physiological recovery ability of human tissues and avoids false alarms caused by the monotonous accumulation of damage assessment. By integrating multiple factors such as pressure, shear, and microcirculation, the accuracy of assessment is significantly improved.

[0020] Expected results: Compared with the traditional method that only uses pressure integrals, the CDI index of this invention has a correlation with the occurrence of pressure ulcers that is increased by more than 40%, and the false positive rate is reduced by more than 50%.

[0021] 2. Closed-loop validation ensures the effectiveness of the intervention. Unlike the traditional passive mode of "timed turning over" or "waiting for nurse's response after alarm", this invention constructs a closed loop of "intervention-verification-re-optimization". The relief effect is verified in real time after each intervention. If the target is not met, the strategy is automatically adjusted and tried again to ensure that the pressure in high-risk areas is truly reduced. Record traceable evidence of intervention implementation to provide objective basis for medical quality assessment and dispute resolution.

[0022] Expected results: Success rate of automatic intervention >85%, reduction of cases requiring manual intervention by more than 60%, and significant reduction in nursing workload.

[0023] 3. Individualized threshold self-learning enhances adaptability Adjust the baseline threshold according to patient characteristics (age, disease, activity level) to avoid a "one-size-fits-all" approach; Continuously monitor skin recovery rate and dynamically track changes in tissue condition; By combining clinical feedback with closed-loop learning, the threshold is automatically adjusted as the patient's condition evolves.

[0024] Expected results: The underreporting rate for high-risk patients (diabetic, elderly) will be reduced by 70%, and the false alarm rate for ordinary patients will be reduced by 50%.

[0025] 4. Predictive early warning signs enable proactive intervention. Extract latent features such as pressure fluctuation rate, temperature gradient, and recovery rate to capture early signs of damage; Predict the risk of pressure ulcers 1-3 hours in advance to allow sufficient intervention window; • Tiered early warning (green / yellow / orange / red) guides nursing priority.

[0026] Expected outcome: More than 80% of pressure ulcer events are detected within 2 hours before they occur, and early intervention reduces the incidence rate by more than 65%.

[0027] 5. Multi-source data consistency verification improves robustness. Time synchronization and consistency checks were performed on multi-channel data of pressure, shear, volume motion, and temperature. When abnormal data is detected, it is automatically downgraded or removed to avoid misjudgment by the system due to a single sensor failure. Output data quality alarms to guide maintenance personnel in troubleshooting.

[0028] Expected results: The system can still operate normally even when a single-channel sensor fails, improving reliability by 40%.

[0029] 6. Multi-objective optimization control balances safety and comfort. Optimization objective: Maximize the decrease in CDI in high-risk areas; Constraints: Limit the turning angle to ≤15° (to prevent the patient from rolling off), the rate of height change to ≤5mm / s (to prevent waking the patient), and the new pressure in non-high-risk areas to <40mmHg (to prevent the creation of new high-pressure points); It supports multiple control strategies: unilateral tilt, local depressurization, and alternating fluctuations.

[0030] Expected results: Micro-turning relieves pressure in high-risk areas, while improving patient comfort scores by 30% and reducing sleep interruptions by 50%.

[0031] 7. System cost and maintainability advantages The approach focuses on methodological innovation, with moderate requirements for sensor hardware, and supports low-cost solutions that use only pressure sensors. Data consistency verification and anomaly handling mechanisms reduce maintenance costs; Traceable data records support remote diagnostics and preventative maintenance.

[0032] In summary, this invention achieves a technological leap from "passive alarm" to "active intervention-validation-re-optimization" by deeply coupling cumulative damage modeling with closed-loop validation. It is significantly superior to existing technologies in terms of accuracy of pressure ulcer prevention, effectiveness of intervention, individualized adaptability, and system robustness, and has important clinical application value and market promotion prospects. Attached Figure Description

[0033] Figure 1 The overall flowchart of the method of this invention shows the complete closed-loop process of S1~S6, 100 method; Z_high high-risk area set; CDI cumulative damage index; Figure 2 A schematic diagram illustrating the cumulative damage modeling principle, showing the tissue vulnerability spatial coefficient matrix K_tissue and the self-healing decay curve; 200 K_tissue matrix; 210 self-healing decay curve; λ self-healing rate parameter; Δτ time since the last effective pressure relief. Figure 3 Example of a CDI heatmap, showing the spatial distribution of cumulative injury index in different parts of the human body: 300 high-risk area (sacrum and coccyx); 310 medium-risk area (scapula); 320 medium-high-risk area (heel). Figure 4 The closed-loop intervention control flowchart shows the iterative process of "400 Identify high-risk areas - 410 Optimize control strategies - 420 Implement intervention - 430 Verify; 440 Success record; 450 Further optimization; 460 Manual intervention"; Figure 5 The flowchart of individualized threshold self-learning illustrates the dynamic threshold adjustment mechanism based on patient characteristics and clinical feedback: 500 Patient profile; 510 Recovery rate; 520 Nursing feedback; 530 Threshold learning; 540 Individualized threshold output. Figure 6 The architecture diagram of the pressure ulcer early warning model shows 600 feature extraction; 610 prediction model; 620 risk probability output; and 630 graded early warning strategy. Figure 7 System overall architecture diagram, including the connection relationships of 700 sensing components; 710 controller; 720 human-machine interaction; 730 execution module; 740 storage module; and 750 cloud platform (optional). Figure 8The diagram illustrates a micro-tilting control strategy, showing (a) unilateral tilting; (b) localized pressure relief; and (c) alternating wave timing control. The diagram also shows the support unit action sequences for these three control modes. Detailed Implementation

[0034] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0035] Example 1: Reference Figure 1-8 A closed-loop intervention system based on pressure sensing.

[0036] System components: This embodiment provides an economical pressure ulcer prevention mattress system suitable for general wards and nursing homes.

[0037] 1. Sensing components A resistive thin-film pressure sensor array is used, positioned beneath the upper layer of foam in the mattress. Sensor specifications: Quantity: 64 (8×8 array), measurement range: 0~200 mmHg, accuracy: ±5 mmHg, sampling frequency: 1 Hz, spatial resolution: the coverage area of ​​a single sensor is approximately 150 cm².

[0038] 2. Partition-controlled support unit array It uses 20 independently controlled airbag units, divided into 5 zones: head zone: 2 airbags, shoulder zone: 4 airbags, back zone: 6 airbags, sacral and coccygeal zone: 4 airbags (key protection), and leg zone: 4 airbags.

[0039] Each airbag is equipped with a miniature solenoid valve and pressure sensor, allowing for independent inflation / deflation. The internal pressure adjustment range is 10-80 mmHg, and the height adjustment range is 0-50 mm.

[0040] 3. Control System Main control chip: STM32F407 (ARM Cortex-M4, 168MHz), data storage: 32GB eMMC flash memory, communication interface: WiFi, Bluetooth, RS485, display terminal: 10.1-inch touch screen (bedside) + mobile APP (nurse station).

[0041] Method implementation: Step S1: Data Acquisition The controller collects data from 64 pressure sensors at a frequency of 1Hz and records it as P(i,j,t), where i=1-8, j=1-8 are the sensor array indices, and t is the timestamp.

[0042] This embodiment only uses pressure data; shear, body motion, and temperature are optional (not configured).

[0043] Step S2: Calculate the cumulative pressure injury dose (PTD) (1) Construct the spatial coefficient matrix K_tissue of organizational vulnerability Based on human anatomy, 64 measurement points were mapped to different body parts: K_tissue matrix (8×8): [0.8, 0.8, 0.8, 0.8, 0.8, 0.8, 0.8, 0.8] ← Head (Low Risk); [1.1, 1.1, 1.2, 1.2, 1.2, 1.2, 1.1, 1.1] ← Scapula (Medium-high risk); [0.7, 0.7, 0.8, 0.8, 0.8, 0.8, 0.7, 0.7] ← Upper back (low risk); [0.7, 0.7, 0.7, 0.7, 0.7, 0.7, 0.7, 0.7] ← Mid-back (low risk); [1.4, 1.5, 1.6, 1.6, 1.6, 1.6, 1.5, 1.4] ← Sacrococcygeal region (highest risk); [1.3, 1.3, 1.4, 1.4, 1.4, 1.4, 1.3, 1.3] ← Hips / Thighs (High Risk); [0.8, 0.8, 0.9, 0.9, 0.9, 0.9, 0.8, 0.8] ← Lower leg (medium risk); [1.2, 1.2, 1.3, 1.3, 1.3, 1.3, 1.2, 1.2] ← Heel (high risk); (2) Calculate the discrete form of PTD Since the sampling frequency is 1Hz, discrete summation is used instead of integration: PTD(i,j,t_n) = Σ[k=0 to n] P(i,j,t_k) · K_tissue(i,j) · e^(-λ·Δτ_k) · Δt in: Δt = 1 second (sampling interval)

[0044] Δτ_k = t_k - t_last_relief, where t_last_relief is the most recent effective pressure relief time. Effective pressure relief identification: When the pressure at a certain point drops from P_high to P_low, and (P_high - P_low) / P_high>0.5, it is recorded as an effective pressure relief event, and Δτ at that point is reset to 0.

[0045] Actual calculation example: Assume that at a point in the sacrococcygeal region (i=5, j=4), K_tissue = 1.6, and the pressure history is as follows: Time t, Pressure P (mmHg), Elapsed since last depressurization Δτ (min), Attenuation factor e^(-λΔτ), PTD increment 0 45 0 1.000 72.0 60s 48 1 0.9995 76.7 120s 50 2 0.9990 79.9 180s 15 (Depressurization) - - - 240s 18 1 (Reset) 0.9995 28.8 PTD cumulative value = 72.0 + 76.7 + 79.9 + 28.8 = 257.4 mmHg·min Step S3: Calculate CDI This embodiment only uses pressure data; the STD term is missing, and I_perfusion is calculated through the depressurization response. Pressure relief response characteristics: When a pressure relief event is detected, record the maximum pressure P_max before pressure relief and the pressure P_10min 10 minutes after pressure relief, and calculate: γ_recovery = (P_max - P_10min) / P_max; If γ_recovery > 0.7, it indicates good tissue recovery, and I_perfusion should be 1.0 (normal). If γ_recovery < 0.5, it indicates poor tissue recovery, and I_perfusion is set to 0.6 (impaired perfusion).

[0046] Simplified CDI formula: CDI = PTD · (2.0 - I_perfusion); When I_perfusion = 1.0, CDI = PTD; When I_perfusion = 0.6, CDI = 1.4·PTD (damage aggravation).

[0047] Step S4: Identify priority intervention areas and generate control sequences (1) Set threshold This embodiment uses a fixed baseline threshold: Threshold_baseline = 800 mmHg·min For elderly patients (≥75 years old) and diabetic patients, the threshold is reduced by 30%: Threshold_personal = 560 mmHg·min (2) Identify high-risk areas Traverse the 64 measurement points and find the set of points Z_high where CDI>Threshold_personal.

[0048] The Z_high was mapped to five airbag regions (head, scapula, back, sacrum, and legs) to determine the areas requiring intervention.

[0049] Example: If the average CDI value of the four measurement points in the sacrococcygeal region is 950 mmHg·min > 800, it is determined to be a priority intervention area.

[0050] (3) Generate control strategy Strategy Type 1: Local Pressure Relief For high-risk cases in the sacrococcygeal region: 4 airbags in the sacrococcygeal region are deflated to 10 mmHg (minimum pressure), and 6 airbags in the back are inflated to 60 mmHg (improved support). Duration: 10 minutes.

[0051] Strategy Type 2: Unilateral Tilt If the scapula and sacrococcygeal region are both at high risk: deflate the left airbag (2 scapulae + 2 sacrococcygeal) to 20 mmHg, and inflate the right airbag to 70 mmHg, forming an approximately 10° rightward tilt, lasting for 8 minutes.

[0052] Strategy Type 3: Alternating Fluctuations If multiple areas are high-risk: Phase 1 (5 minutes): Deflat the sacrococcygeal airbag and inflate the back airbag; Phase 2 (5 minutes): Inflate the sacral and coccygeal airbags and deflate the back airbags, repeat 3 times.

[0053] Multi-objective optimization implementation (simplified version): The "local pressure relief" strategy is preferred. If there are more than 2 high-risk areas, the "unilateral tilt" strategy is selected. If there are more than 3 high-risk areas, the "alternating fluctuation" strategy is selected.

[0054] Constraint checks: Single height change <50mm (met), tilt angle verified by geometric calculation <15° (met), additional pressure in non-high-risk areas <40mmHg (verified by sensor prediction). Step S5: Execute and verify (1) Execute control sequence The controller sends commands to 20 solenoid valves to inflate and deflate the air according to a strategy. The inflation speed is approximately 5 mm / s (meeting comfort requirements).

[0055] (2) Verification time window After execution, wait for T_verify = 1 minute to allow the pressure distribution to stabilize.

[0056] (3) Re-collect data Collect data from 64 pressure sensors again, P_new(i,j,t), and recalculate PTD_new and CDI_new.

[0057] Step S6: Calculate the mitigation effect and make a decision (1) Calculate the remission effect index For priority intervention areas Z_high (e.g., 4 measurement points in the sacrococcygeal region): Relief_score = Σ[CDI_old(i,j) - CDI_new(i,j)] for (i,j) ∈ Z_high Target value: Relief_score > 0.6 · Σ CDI_old (i.e., CDI decrease ≥ 60%) (2) Decision-making logic Scenario A: Meets the standard If the Relief_score reaches the target, record the intervention as successful and return to step S1 to continue monitoring.

[0058] Scenario B: Failed to meet the standard and number of attempts < 3 Update the control strategy: If "partial pressure relief" was used before, switch to "single-sided tilt". If "single-sided tilt 10°" was used before, increase the angle to 15° and return to step S4 to execute again.

[0059] Situation C: Failed to meet the standard 3 times consecutively The system will issue an upgraded alarm for manual repositioning: The bedside screen will display: "Warning: Sacrococcygeal pressure continues to exceed the limit, automatic intervention has failed, please manually reposition the patient immediately!" The nurse station app will push a notification with the patient's bed number, high-risk area, and suggested position (left lateral decubitus 30°). This will be recorded as a dangerous event and written to the log. Implementation effect verification: The system described in this embodiment was deployed in the neurosurgery ward of a tertiary hospital, and 30 long-term bedridden patients (average age 68 years, bedridden time >14 days) were monitored for 28 days, compared with a traditional alternating air mattress (control group of 30 patients): The improvement of the traditional mattress by the present invention system is indicated. The incidence of pressure ulcers was 3.3% (1 case) compared to 16.7% (5 cases), representing an 80% reduction. The frequency of stage I erythema decreased from 12 to 38 times, a reduction of 68%. The frequency of manual turning over per day decreased by 68% from 2.1 to 6.5 times. The number of nighttime interventions by nurses decreased by 89% from 0.3 to 2.8 per day. Patient's sleep quality score (VAS) improved by 41% from 7.2 to 5.1. Typical Case: Patient Zhang, male, 72 years old, with a 10-year history of diabetes, was bedridden due to post-stroke surgery. On the 3rd day of using this system, the sacrococcygeal CDI reached 720 mmHg·min (threshold 560). The system automatically implemented the "local pressure relief" strategy, and after verification, the CDI decreased to 280 mmHg·min (a decrease of 61%), meeting the target. On the 7th day, intervention was triggered again. The first attempt resulted in a 52% decrease, which did not meet the target. The system switched to the "unilateral tilt" strategy, and the second attempt resulted in a 68% decrease, meeting the target. During the 28-day monitoring period, the patient did not develop any pressure ulcers, and the nurse intervened only once at night.

[0060] Example 2: An advanced system integrating shear sensing and precursor prediction System components: This embodiment provides a high-end pressure ulcer prevention mattress system suitable for ICU intensive care and high-risk patients.

[0061] 1. Sensing components (1) Pressure sensing: Capacitive flexible pressure sensor array is used: Quantity: 256 (16×16 array), measurement range: 0~300 mmHg, accuracy: ±2 mmHg, sampling frequency: 5 Hz.

[0062] (2) Shear sensing, using triaxial force sensors: quantity: 64 (8×8 array, spatially aligned with pressure sensors), shear force measurement range: 0~50 N, resolution: shear components in X and Y directions, sampling frequency: 5 Hz.

[0063] (3) Body motion sensing, using accelerometers: number: 4 (placed on the chest, waist, hips and legs respectively), detection micro-motion amplitude: 0.01~10 m / s², sampling frequency: 10 Hz.

[0064] (4) Temperature sensing, using an infrared thermal imaging array: resolution: 32×32 pixels, temperature measurement range: 25~45°C, accuracy: ±0.5°C, sampling frequency: 0.2 Hz (one frame every 5 seconds).

[0065] 2. Partition-controlled support unit array It employs 60 independently controlled gas-liquid mixing bladder units; the gas bladder layer provides rapid response (adjustment speed 5 mm / s); the liquid bladder layer provides stable support (avoiding swaying); the internal pressure adjustment range is 5~100 mmHg; and the height adjustment range is 0~80 mm. 3. Control System Main control chip: NVIDIA Jetson Nano (AI acceleration), data storage: 256GB SSD, communication interfaces: WiFi 6, 5G, Ethernet, display terminal: 15.6-inch touch screen + cloud care platform.

[0066] Method implementation: Step S1: Multi-source data acquisition and consistency verification (1) Data Acquisition The controller synchronously collects four types of data: Pressure: P(i,j,t), 256 points, 5Hz; Shear: S_x(i,j,t), Sy(i,j,t), 64 points, 5Hz; Volumetric motion: A_k(t), 4 points, 10Hz; Temperature: T(m,n,t), 1024 points, 0.2Hz (2) Time synchronization All sensors use a unified timestamp t (UTC millisecond precision). Data from different sampling frequencies are interpolated and aligned to a 5Hz reference.

[0067] (3) Consistency check Verification Rule 1: Pressure-body movement consistency The peak motion time t_motion (acceleration > 0.5 m / s²) should be detected, and the displacement of the center of gravity of the pressure distribution should occur within t_motion ± 2 seconds.

[0068] Calculate the coordinates of the pressure center: x_center = Σ[i·P(i,j)] / ΣP(i,j) y_center = Σ[j·P(i,j)] / ΣP(i,j) If body movement occurs but the displacement of the pressure center is less than 5cm, it is judged as data inconsistency, and the weight of the body movement channel is reduced by 50%.

[0069] Verification Rule 2: Temperature-Pressure Consistency In high-pressure areas (P>50mmHg), a temperature rise (ΔT>0.5°C) should occur after 5 minutes.

[0070] Calculate spatial overlap: Overlap = |Z_highP ∩ Z_highT| / |Z_highP| Z_highP represents the high-pressure region, and Z_highT represents the high-temperature region.

[0071] If Overlap < 0.6, the temperature sensor is considered faulty, and the temperature channel data is discarded.

[0072] Verification Rule 3: Consistency of Shear-Volume Direction When the body motion direction θ_motion is detected (through vector synthesis from multiple acceleration sensors), the principal shear force direction θ_shear should have an angle of less than 30° with it.

[0073] If the included angle is greater than 60°, the shear sensor is deemed abnormal, and the weight of the shear channel is reduced by 70%.

[0074] Step S2: Calculate the cumulative pressure injury dose (PTD) Same as Example 1, but with a finer K_tissue matrix (16×16) and a higher sampling frequency (5Hz).

[0075] Step S3: Calculate the cumulative shear damage dose (STD) and CDI. (1) Calculate the shear vector magnitude For each measurement point (i,j): S(i,j,t) = √[S_x(i,j,t)² + S_y(i,j,t)²] Direction angle: θ(i,j,t) = arctan[S_y / S_x] (2) Determine the shear direction weighting function α(θ) Based on the direction of human skin texture (Langer lines): Back: Skin texture mainly follows the direction of the spine (θ_skin = 90°); Buttocks: Skin texture mainly follows the direction of the gluteal crease (θ_skin = 0°); definition: α(θ) = 1.0 - 0.4·cos²(θ - θ_skin) When the shear direction is perpendicular to the skin texture, α = 1.0 (most susceptible to damage). When the shearing direction is parallel to the skin texture, α = 0.6 (less damage).

[0076] (3) Calculate STD STD(i,j,t_n) = Σ[k=0 to n] S(i,j,t_k) · α(θ_k) · K_tissue(i,j) ·e^(-λ·Δτ_k) · Δt (4) Calculate the microcirculation perfusion index I_perfusion Combining the characteristics of the pressure relief response: γ_recovery = (P_max - P_10min) / P_max (pressure recovery rate) ΔT_recovery = T_10min - T_depressurization time (temperature recovery characteristic) I_perfusion = 0.6·γ_recovery + 0.4·(ΔT_recovery / 2.0) Normalize to the interval [0, 1].

[0077] (5) Integration of CDI CDI = 0.55·PTD + 0.30·STD + 0.15·(1 - I_perfusion) Among them, the weighting coefficient (Obtained through training with clinical data).

[0078] Steps S4-S6: Closed-loop intervention control Similar to Example 1, but with the following optimizations: (1) More refined multi-objective optimization The optimal control sequence is solved using a genetic algorithm: Objective function:

[0079] Constraints: CDI_new <Threshold_personal for (i,j) ∈ Z_high

[0080] θ_tilt ≤ 15° dH / dt ≤ 5 mm / s (rate of height change) |P(i,j) - P(i±1,j±1)|<20 mmHg (pressure difference between adjacent units) Controlled variables: Target internal pressure of 60 airbags, P_bag(k), k=1~60 (2) Introduce predictive precursor identification (before step S4) Feature extraction: Extract within a sliding window of T_w = 30 minutes: Pressure volatility: σ_P = std(P[t-1800 : t]), updated every 5 minutes. Temperature gradient: ΔT_trend = (T_current - T_baseline) / T_baseline Recovery rate: R_recovery = moving average of γ_recovery Duration without physical movement: T_static = time since the last physical movement event CDI change rate: dCDI / dt ≈ (CDI_current - CDI_10min ago) / 600 Predictive model: Using XGBoost gradient boosting tree: Input features: [σ_P, ΔT_trend, R_recovery, T_static, dCDI / dt, time period identifier (nighttime=1)] Output: P_risk (probability of pressure ulcer occurrence in the next 2 hours) Training set: 100 patients, 2000 hours of monitoring data, including 150 pressure ulcer events. Model hyperparameters: Tree depth: 6 levels Learning rate: 0.05 Number of trees: 300 AUC validation set: 0.89 Early warning response: P_risk < 20%: Normal monitoring (CDI calculated every 10 minutes) 20% ≤ P_risk < 50%: Increase monitoring frequency (calculate CDI every 5 minutes), and send a yellow alert via the nurse's app. 50% ≤ P_risk < 80%: Immediately trigger closed-loop intervention in step S4, without waiting for CDI to exceed the limit. P_risk ≥ 80%: Red alert, triggering both automatic and manual intervention notifications. (3) Individualized threshold self-learning Initial threshold calculation: Based on patient information: Age: 75 years old → r_age = 0.75 Diabetes: Yes → r_diabetes = 0.7 Braden rating: 12 points → r_mobility = 0.85 BMI: 18 (underweight) → r_nutrition = 0.9 R_base = 0.75 × 0.7 × 0.85 × 0.9 = 0.40 Threshold_personal = 800 × 0.40 = 320 mmHg·min Dynamic adjustment: Week 1: Nurses performed daily skin assessments. On the 3rd day, stage I erythema was found on the sacrococcygeal region, at which time CDI = 280 mmHg·min < 320.

[0081] The system has determined that the threshold is too high; adjustments are needed. Threshold_personal = 320 × 0.85 = 272 mmHg·min (a decrease of 15%) Week 2: Continuous monitoring of γ_recovery revealed a decrease from an initial 0.72 to 0.58, indicating reduced tissue elasticity.

[0082] f(γ_recovery) = 0.58 / 0.72 = 0.81 Threshold_personal = 272 × 0.81 = 220 mmHg·min (further reduction) Week 3: The nurse reported that the skin condition was good, with no new erythema and the CDI peak did not exceed 250.

[0083] The system's judgment can be appropriately relaxed: Threshold_personal = 220 × 1.10 = 242 mmHg·min (increase by 10%) Final stable threshold: 242 mmHg·min (a 70% reduction from the initial value, reflecting the characteristics of high-risk patients) Implementation effect verification: The system described in this embodiment was deployed in the ICU of a tertiary hospital, and 20 high-risk patients (mean Braden score 10.5, bedridden time > 21 days) were monitored for 28 days. The results were compared with the standard manual turning protocol (control group of 20 patients, turned every 2 hours): Indicators of the improvement in manual turning over by the present invention Incidence of pressure ulcers: 5% (1 case), 30% (6 cases), a decrease of 83%. The frequency of stage I erythema occurrence decreased from 8 to 45 times, a reduction of 82%. The frequency of manual turning over decreased from 0.9 times per day to 12 times per day, a reduction of 93%. Warning accuracy (AUC) 0.91 N / A New feature Nursing workload (hours / day): 0.6 3.5 Reduced by 83% Patient comfort score improved by 31% from 8.1 to 6.2. Typical Case: Patient Li, female, 82 years old, with diabetes and heart failure, Braden score of 9 (very high risk). On the 5th day of using this system, the prodromal prediction module detected: σ_P continuously increasing (0.8→1.2→1.8 mmHg), γ_recovery decreasing to 0.52, T_static reaching 40 minutes, outputting P_risk = 76% (orange warning). At this time, CDI = 210 mmHg·min, which had not yet exceeded the individualized threshold of 242, but the system triggered closed-loop intervention in advance, implementing the "sacral and coccygeal local pressure relief + unilateral tilt of 10°" combined strategy. After verification, CDI decreased to 92 mmHg·min. Two hours later, the nurse checked and found that the skin of the sacrum and coccyx was slightly red but did not reach the stage I standard, avoiding a pressure ulcer event. During the 28-day monitoring period, the patient triggered warnings 13 times, automatic intervention was successful 11 ​​times, manual turning was performed 2 times, and no pressure ulcers occurred.

Claims

1. A closed-loop intervention method for pressure injury risk based on cumulative injury assessment, executed by a processor, applied to a nursing mattress system with zoned controllable support units, characterized in that, Includes the following steps: S1 collects time-series data from multiple points on the bed surface. The time-series data includes at least pressure-related signals and optionally shear-related signals, body motion-related signals, and / or temperature-related signals. S2 Based on the pressure-related signal, the cumulative pressure damage dose PTD(x,y,t) is calculated according to the spatial location (x,y) and time t. The calculation of PTD introduces the tissue vulnerability spatial coefficient matrix K_tissue(x,y) to characterize the tolerance differences of different body parts, and introduces a self-repair attenuation term to characterize the natural regression process of tissue damage after pressure relief. S3 When a shear-related signal is present, the cumulative shear damage dose STD(x,y,t) is calculated according to spatial location and time. The calculation of STD introduces a shear direction weighting function α(θ) to characterize the anisotropic properties of the skin. The PTD and STD are weighted and fused to obtain the comprehensive cumulative damage index CDI(x,y,t). S4 determines the priority intervention area based on the spatial distribution of the CDI, generates micro-turning and / or alternating decompression control sequences through multi-objective constraint optimization, and drives the partitioned controllable support unit to execute them; Within the verification time window after the S5 control sequence is executed, the time series data is collected again and the PTD, STD and / or CDI are recalculated to obtain the cumulative damage index distribution after the intervention. S6 Calculate the relief effect index of the priority intervention area. When the relief effect index does not reach the preset target and the number of attempts does not exceed the upper limit, update the control sequence and return to step S4 to re-execute, so as to form a closed-loop intervention based on effect verification. If the target is not met after several attempts, an alarm will be issued to upgrade the manual turnaround.

2. The closed-loop intervention method for pressure injury risk based on cumulative damage assessment according to claim 1, characterized in that, The cumulative pressure injury dose PTD(x,y,t) in step S2 satisfies the following continuous form or its equivalent discrete form: ； in: P(x,y,τ) is the pressure value at position (x,y) at time τ; K_tissue(x,y) is the spatial coefficient matrix of tissue vulnerability, with higher weights assigned to bony prominence areas and areas with poor blood supply. The values ​​range as follows: sacrococcygeal region: 1.4-1.6, high-risk area; heel: 1.2-1.4, medium-high risk; scapula: 1.0-1.2, medium risk; back muscle area: 0.6-0.8, low risk. λ is the tissue self-repair rate parameter, and its value range is... ; Δτ is the time interval since the last effective pressure relief event, which is identified by detecting a pressure drop exceeding 50%.

3. The closed-loop intervention method for pressure injury risk based on cumulative damage assessment according to claim 1, characterized in that, The cumulative shear damage dose STD(x,y,t) in step S3 satisfies the following continuous form or its equivalent discrete form: ； in: S(x,y,τ) is the magnitude of the shear strength or shear vector; α(θ) is the shear direction weight function, where θ is the direction angle of the shear vector. α(θ) has a maximum value of 1.0 in the direction perpendicular to the skin texture and a value of 0.6 to 0.8 in the parallel direction.

4. The closed-loop intervention method for pressure injury risk based on cumulative damage assessment according to claim 1, characterized in that, The Comprehensive Cumulative Damage Index (CDI) mentioned in step S3 must satisfy at least one of the following fusion forms: (a) Two-factor fusion: ; (b) Three-factor fusion: ; in, The weighting parameter satisfies the normalization condition; I_perfusion is the microcirculation perfusion index, which is derived from at least one of the following decompression response characteristics: Pressure recovery rate after depressurization: γ_recovery = (P_max - P_10min) / P_max; Local temperature change rate after depressurization; The weighted combination of the two mentioned above.

5. A closed-loop intervention method for stress injury risk based on cumulative damage assessment according to claim 1, characterized in that, It also includes an individualized risk threshold adaptive learning step: (A) Obtain patient individual characteristic parameters, which include at least age, disease factors (diabetes, peripheral vascular disease), activity level score and / or body mass index (BMI), and calculate the baseline risk correction factor R_base; (B) Continuously monitor the patient's skin pressure recovery rate index γ_recovery. When this index continues to decline, it is determined that the tissue elasticity is reduced. (C) Receive feedback from clinical nursing staff on the graded assessment of the patient's skin condition, which includes at least: no abnormalities, stage I erythema, and stage II superficial ulcers; (D) Dynamically adjust individualized risk thresholds using supervised learning algorithms: When the assessment feedback is "Stage I erythema or more severe" and the current CDI value is below the threshold, reduce the threshold by 20% to 40%; When the assessment feedback is "no abnormality" and the current CDI value is close to or exceeds the threshold, the threshold is increased by 5% to 15%. (E) Output individualized risk threshold: Threshold_personal = Threshold_baseline × R_base × f(γ_recovery) × η_feedback, where f(γ_recovery) is an adjustment function based on recovery rate and η_feedback is a learning coefficient based on clinical feedback.

6. The closed-loop intervention method for pressure injury risk based on cumulative damage assessment according to claim 1, characterized in that, It also includes steps for predicting early signs of pressure ulcers: (P1) Extract multidimensional physiological time-series features within a sliding time window T_w = 20~40 minutes. The multidimensional physiological time-series features include at least: pressure fluctuation rate σ_P(t), local temperature change rate ΔT_trend, pressure recovery rate after decompression R_recovery, continuous non-movement duration T_static, and cumulative damage index change rate dCDI / dt. (P2) A training dataset was constructed based on historical patient monitoring data and pressure ulcer occurrence records. Gradient boosting decision tree, random forest or neural network were used to train the pressure ulcer precursor prediction model. (P3) Input the multidimensional physiological time-series features extracted in real time into the prediction model and output the probability of skin damage events occurring within the future time window T_pred = 1~3 hours, P_risk. (P4) Output graded early warning signals based on the risk probability: P_risk < 20% is green and safe, 20% ≤ P_risk < 50% is yellow and caution is required, 50% ≤ P_risk < 80% is orange and triggers the closed-loop intervention in step S4 in advance, and P_risk ≥ 80% is red and critical and immediately requests manual intervention.

7. The closed-loop intervention method for stress injury risk based on cumulative damage assessment according to claim 1, characterized in that, Step S4, generating the control sequence, includes solving the following multi-objective constrained optimization problem: Objective function: Minimize the sum of squared pressures in the priority intervention region Z_high, or maximize the decrease in the cumulative damage index; Constraints: Pressure reduction in priority intervention areas ≥ 60%; Peak pressure increase in non-priority intervention areas < 40 mmHg; Body turning angle ≤ 15°; Height change rate of a single support unit ≤ 5 mm / s; Pressure difference between adjacent support units < 20 mmHg; The control sequence includes one or a combination of the following types: unilateral tilt type, which elevates the left or right side of the body by 5° to 15°; local decompression type, which decompresses only a single high-risk area such as the sacrum, coccyx, or heel; and fluctuating alternation type, inflating and deflating multiple areas sequentially according to a time sequence, with phases staggered to suppress shear peaks.

8. A closed-loop intervention method for stress injury risk based on cumulative damage assessment according to claim 1, characterized in that, Step S1 also includes a multi-source data consistency verification step: Synchronize and align time-series data from different acquisition channels using timestamps; Calculate the consistency index between the data of each channel. The consistency index includes the time deviation between the pressure peak time and the body motion event time, the spatial overlap between the temperature rise region and the high pressure region, and the angular deviation between the shear direction and the body motion direction. When the consistency index exceeds the preset threshold, it is determined that the corresponding channel data is abnormal, and one of the following measures is taken: reduce the weight coefficient of the channel data in the CDI fusion calculation, temporarily remove the channel data and trigger sensor self-test, or output data quality alarm to the nursing terminal.

9. A closed-loop intervention system for stress injury risk based on cumulative damage assessment, characterized in that, include: The sensing component includes a sensor array distributed within the mattress body, the sensor array including at least a pressure acquisition channel, and optionally including a shear acquisition channel, a body motion acquisition channel and / or a temperature acquisition channel; A partitioned controllable support unit array includes 20-100 fluid units that can independently control internal pressure or support height. The fluid units are air bladders, liquid bladders, or gas-liquid mixture bladders, and each unit covers an area of ​​50-200 cm². The controller, communicatively connected to the sensing components and the partitioned controllable support unit array, is configured to perform the method according to any one of claims 1 to 8; The memory is used to record the spatiotemporal distribution evolution data of CDI throughout the entire time course, the control parameters and relief effect scores of each automatic intervention, clinical assessment feedback records, and complete monitoring data snapshots when pressure ulcer events occur; The human-computer interaction module includes a bedside display terminal and / or a mobile nursing terminal, which is used to display CDI heat maps and graded early warning signals in real time, receive skin assessment feedback input from nursing staff, and output manual turning upgrade alarms and suggested positions; The controller is also configured to: perform time synchronization and consistency verification on data from different acquisition channels, and dynamically adjust the data weight of each channel or remove abnormal channel data when consistency is not met.

Images