Method for dynamically adjusting body position of bedridden patient based on infrared sensing

Abstract

The invention relates to the technical field of medical care, in particular to a method for dynamically adjusting the body position of a bedridden patient based on infrared sensing, which comprises the following steps: S1, three-dimensional infrared monitoring: collecting body surface temperature distribution and body contour data of the patient in real time through a plurality of infrared thermal imaging sensors and 3D depth cameras distributed around a sickbed; an eight-channel FLIR A315 thermal imaging sensor (with the precision of + / -0.1 DEG C) and an Intel RealSense depth camera are adopted to construct a three-dimensional monitoring network, sub-millimeter level fusion of a temperature field and body data is realized through non-local mean filtering and an ICP point cloud registration algorithm, the recognition precision of a pressure concentration area is improved to 3 mm < 3 > voxel units and is improved by more than three times compared with a traditional pressure sensing mattress (with the resolution being greater than or equal to 1 cm), and the recognition precision of the pressure concentration area is improved by more than three times compared with that of a traditional pressure sensing mattress (with the resolution being greater than or equal to 1 cm). The temperature gradient monitoring sensitivity reaches 0.1 DEG C / cm, the potential stress injury risk is found 4-6 hours in advance, the risk early warning accuracy rate of data display of a certain hospital in Hainan reaches 92.3%, continuous unattended monitoring is conducted for 24 hours, and the workload of manual temperature measurement of nursing staff is reduced by 80% or above.

Description

Technical Field

[0001] The present invention relates to the field of medical care technology, and in particular to a method for dynamically adjusting the body position of bedridden patients based on infrared sensing. Background Art

[0002] Long-term bedridden patients are unable to change their positions frequently and are very prone to pressure injuries (bedsores). Traditional nursing methods have the following problems:

[0003] Manual turning is time-consuming and labor-intensive: Nurses need to manually assist patients in turning over every 2 hours, which is a large workload and can easily lead to delayed operation.

[0004] Inaccurate pressure monitoring: Existing pressure-sensing mattresses can only provide static pressure distribution data and are unable to track body position changes in real time or predict potential risk areas.

[0005] Lack of intelligent early warning mechanism: Unable to dynamically adjust the posture intervention strategy according to the patient's individual conditions (such as skin condition, illness).

[0006] Low patient comfort: Traditional turning methods may interrupt patients' sleep and affect rehabilitation effects. Therefore, a dynamic adjustment method for bedridden patients' body position based on infrared sensing is proposed. Summary of the Invention

[0007] In view of this, the present invention provides a method for dynamically adjusting the body position of bedridden patients based on infrared sensing to solve or alleviate the technical problems existing in the prior art and at least provide a beneficial option.

[0008] The technical solution of the present invention is implemented as follows: a method for dynamically adjusting the body position of a bedridden patient based on infrared sensing, comprising the following steps:

[0009] S1, three-dimensional infrared monitoring, through multiple infrared thermal imaging sensors and 3D depth cameras distributed around the bed, collects the patient's surface temperature distribution and body contour data in real time;

[0010] S2, data fusion and preprocessing, denoising the collected infrared images, fusing the infrared data with the depth data, and constructing a three-dimensional temperature field model of the patient's body surface;

[0011] S3, pressure risk assessment, calculating the pressure concentration index (PCI) based on the three-dimensional temperature field model, and using the LSTM neural network model to predict the pressure injury risk level within the next 2 hours;

[0012] S4, intelligent body positioning planning, uses reinforcement learning models to generate personalized body positioning plans based on the patient's body parameters, disease limitations, and risk assessment results;

[0013] S5, dynamic execution and feedback, controls the segmented intelligent mattress to perform posture adjustment, and performs closed-loop feedback correction based on the pressure distribution changes after adjustment.

[0014] Further preferably, in S1, 8 high-precision infrared thermal imaging sensors (resolution ≥640×480, temperature accuracy ±0.1°C) distributed around the bed are used to collect the patient's surface temperature distribution data every 5 seconds, and the 3D depth camera (FOV ≥120°, depth accuracy ≤1cm) is started simultaneously to obtain three-dimensional data of the patient's body contour and the contact area with the bed surface.

[0015] Further preferably, in S2, the infrared image is denoised (using a non-local mean filtering algorithm) to eliminate ambient temperature interference, and the 3D depth data is fused with the infrared thermal imaging data through image registration technology to construct a three-dimensional temperature field model of the patient's body surface.

[0016] Further preferably, in S3, the pressure concentration index (PCI) is calculated based on the fused data: PCI = f(T, A, t), where T is the local temperature, A is the contact area, and t is the duration. An LSTM neural network model (training data set ≥ 10,000 samples) is used to predict the pressure injury risk level (divided into levels I-V) within the next 2 hours, and infrared thermal imaging data is used to extract the skin temperature gradient (ΔT / Δx). When the local temperature gradient exceeds 0.5°C / cm, it is marked as a high-risk area. The risk assessment result is corrected in combination with the historical skin condition data in the patient's electronic medical record.

[0017] Further preferably, in said S4, based on the patient's body parameters (BMI, limb length), disease limitations (such as fracture site) and risk assessment results, a pre-trained reinforcement learning model (the reward function includes pressure dispersion and comfort score) is called to generate an optimal turning angle sequence (such as 45° on the left side → supine position → 30° on the right side), and the optimal time for each position change is calculated to ensure that the interval between two turns is dynamically adjusted between 1.5-3 hours. A biomechanical model is used to simulate the impact of position changes on the patient's musculoskeletal system to avoid excessive traction. Combined with the patient's sleep monitoring data (using infrared sensors to identify changes in respiratory rate), position adjustment is performed preferentially during light sleep.

[0018] Further preferably, in said S5, a control instruction is sent to the segmented intelligent mattress via serial communication (baud rate ≥ 115200bps) to drive the inflation / deflation of the independent air chamber to achieve ±0.5° accuracy of body position adjustment. During the execution process, the motor current is monitored in real time (the threshold is set to 1.2A) to prevent mechanical failure from causing injury to the patient. The pressure distribution change is continuously monitored within 5 minutes after the adjustment. If the risk area is not effectively alleviated (PCI decreases by <20%), a secondary adjustment strategy is triggered, and a patient position preference database is established. The personalized position plan is continuously optimized through transfer learning.

[0019] The embodiment of the present invention adopts the above technical solution, which has the following advantages:

[0020] 1. This invention uses an 8-channel FLIR A315 thermal imaging sensor (±0.1°C accuracy) and an Intel RealSense depth camera to build a 3D monitoring network. It uses non-local mean filtering and ICP point cloud registration algorithm to achieve sub-millimeter fusion of temperature field and shape data, and the recognition accuracy of pressure concentration areas is improved to 3mm. 3 The voxel unit is more than three times that of traditional pressure-sensing mattresses (resolution ≥ 1cm), and the temperature gradient monitoring sensitivity reaches 0.1℃ / cm, which can detect potential pressure injury risks 4-6 hours in advance. Data from a hospital in Hainan shows that the risk warning accuracy rate is 92.3%. 24-hour continuous unmanned monitoring reduces the workload of manual temperature measurement for nursing staff by more than 80%.

[0021] 2. The present invention constructs a two-layer LSTM neural network with 128-dimensional feature input, integrates 12 consecutive frames of PCI data with patient physiological parameters, and achieves pressure injury grade prediction through training with 10,000 clinical data cases. The F1-score for predicting pressure injuries above grade III reaches 0.92, which reduces the false alarm rate by 65% ​​compared with traditional empirical judgment. It automatically generates a risk trend curve for the next 2 hours, providing a precise time window for nursing intervention. In a pilot ward, the timeliness of intervention increased by 78%. Combined with the correction factor model of historical skin injury data, the accuracy of personalized risk assessment for patients with different physical conditions increased by 30%.

[0022] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0024] Figure 1 It is a structural diagram of the present invention;

[0025] Figure 2 It is a structural diagram of the present invention;

[0026] Figure 3 It is a structural diagram of the present invention. DETAILED DESCRIPTION

[0027] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.

[0028] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0029] like Figure 1-3 As shown, an embodiment of the present invention provides a method for dynamically adjusting the body position of a bedridden patient based on infrared sensing, comprising the following steps:

[0030] S1, three-dimensional infrared monitoring, through multiple infrared thermal imaging sensors and 3D depth cameras distributed around the bed, collects the patient's surface temperature distribution and body contour data in real time;

[0031] S2, data fusion and preprocessing, denoising the collected infrared images, fusing the infrared data with the depth data, and constructing a three-dimensional temperature field model of the patient's body surface;

[0032] S3, pressure risk assessment, calculates the pressure concentration index (PCI) based on a three-dimensional temperature field model and uses an LSTM neural network model to predict the pressure injury risk level within the next 2 hours;

[0033] S4, intelligent body positioning planning, uses reinforcement learning models to generate personalized body positioning plans based on the patient's body parameters, disease limitations, and risk assessment results;

[0034] S5, dynamic execution and feedback, controls the segmented intelligent mattress to perform posture adjustment, and performs closed-loop feedback correction based on the pressure distribution changes after adjustment.

[0035] In one embodiment, in S1, eight high-precision infrared thermal imaging sensors (resolution ≥640×480, temperature accuracy ±0.1°C) distributed around the bed are used to collect the patient's body surface temperature distribution data every 5 seconds, and a 3D depth camera (FOV ≥120°, depth accuracy ≤1cm) is started simultaneously to obtain the three-dimensional data of the patient's body contour and the contact area with the bed surface; the thermal imaging sensor is FLIR A315 model (resolution 640×480, temperature range -20°C to 150°C, accuracy ±0.1°C, response wavelength 7.5-13μm), and the 3D depth camera is Intel RealSense D455 (FOV 87°×58°×95°, depth accuracy 0.5-10mm@0.5-4m, frame rate 90fps), one thermal imaging sensor is deployed at each corner of the bed (1.2m high, 45° inclination), one is deployed in the middle of the head and foot of the bed (1.5m high, vertically downward), and 3D cameras are installed symmetrically on both sides of the bedside (1.2m apart, 1.8m high, 60° angle).

[0036] In one embodiment, in S2, the infrared image is denoised (using a non-local mean filtering algorithm) to eliminate ambient temperature interference, and the 3D depth data is fused with the infrared thermal imaging data through image registration technology to construct a three-dimensional temperature field model of the patient's body surface.

[0037] In one embodiment, in S3, the pressure concentration index (PCI) is calculated based on the fused data: PCI = f(T, A, t), where T is the local temperature, A is the contact area, and t is the duration. An LSTM neural network model (training data set ≥ 10,000 samples) is used to predict the pressure injury risk level (divided into grades I-V) within the next 2 hours. Infrared thermal imaging data is used to extract the skin temperature gradient (ΔT / Δx). When the local temperature gradient exceeds 0.5°C / cm, it is marked as a high-risk area. The risk assessment result is corrected by combining the historical skin condition data in the patient's electronic medical record.

[0038] In one embodiment, in S4, based on the patient's body parameters (BMI, limb length), disease limitations (such as fracture site) and risk assessment results, a pre-trained reinforcement learning model (the reward function includes pressure dispersion and comfort score) is called to generate an optimal turning angle sequence (such as 45° on the left side → supine position → 30° on the right side), calculate the optimal time for each body position change, ensure that the interval between two turns is dynamically adjusted between 1.5-3 hours, use a biomechanical model to simulate the impact of body position changes on the patient's musculoskeletal system to avoid excessive traction, and combine the patient's sleep monitoring data (using infrared sensors to identify changes in respiratory rate) to perform body position adjustments preferentially during light sleep.

[0039] In one embodiment, in S5, control instructions are sent to the segmented smart mattress through serial communication (baud rate ≥ 115200bps) to drive the inflation / deflation of independent air chambers to achieve ±0.5° accuracy of body position adjustment. During the execution process, the motor current is monitored in real time (the threshold is set to 1.2A) to prevent mechanical failure from causing patient injury. The pressure distribution changes are continuously monitored within 5 minutes after the adjustment. If the risk area is not effectively alleviated (PCI decreases by <20%), a secondary adjustment strategy is triggered, and a patient position preference database is established. The personalized position plan is continuously optimized through transfer learning.

[0040] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various modifications and substitutions within the technical scope disclosed in the present invention, and such modifications and substitutions are intended to be within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A method for dynamically adjusting the body position of bedridden patients based on infrared sensing, characterized by: The following steps are involved: S1, three-dimensional infrared monitoring, through multiple infrared thermal imaging sensors and 3D depth cameras distributed around the bed, collects the patient's surface temperature distribution and body contour data in real time; S2, data fusion and preprocessing, denoising the collected infrared images, fusing the infrared data with the depth data, and constructing a three-dimensional temperature field model of the patient's body surface; S3, pressure risk assessment, calculating the pressure concentration index (PCI) based on the three-dimensional temperature field model, and using the LSTM neural network model to predict the pressure injury risk level within the next 2 hours; S4, intelligent body positioning planning, uses reinforcement learning models to generate personalized body positioning plans based on the patient's body parameters, disease limitations, and risk assessment results; S5, dynamic execution and feedback, controls the segmented intelligent mattress to perform posture adjustment, and performs closed-loop feedback correction based on the pressure distribution changes after adjustment.

2. The method for dynamically adjusting the body position of a bedridden patient based on infrared sensing according to claim 1, characterized in that: In S1, eight high-precision infrared thermal imaging sensors (resolution ≥640×480, temperature accuracy ±0.1°C) distributed around the bed are used to collect the patient's surface temperature distribution data every 5 seconds, and a 3D depth camera (FOV ≥120°, depth accuracy ≤1cm) is started simultaneously to obtain three-dimensional data of the patient's body contour and the contact area with the bed surface.

3. The method for dynamically adjusting the body position of a bedridden patient based on infrared sensing according to claim 1, characterized in that: In S2, the infrared image is denoised (using a non-local mean filtering algorithm) to eliminate ambient temperature interference, and the 3D depth data is fused with the infrared thermal imaging data through image registration technology to construct a three-dimensional temperature field model of the patient's body surface.

4. The method for dynamically adjusting the body position of a bedridden patient based on infrared sensing according to claim 1, characterized in that: In S3, the pressure concentration index (PCI) is calculated based on the fused data: PCI = f(T, A, t), where T is the local temperature, A is the contact area, and t is the duration. An LSTM neural network model (training data set ≥ 10,000 samples) is used to predict the pressure injury risk level (divided into grades I-V) within the next 2 hours. Infrared thermal imaging data is used to extract the skin temperature gradient (ΔT / Δx). When the local temperature gradient exceeds 0.5°C / cm, it is marked as a high-risk area. The risk assessment result is corrected by combining the historical skin condition data in the patient's electronic medical record.

5. The method for dynamically adjusting the body position of a bedridden patient based on infrared sensing according to claim 1, characterized in that: In the S4, based on the patient's body parameters (BMI, limb length), disease limitations (such as fracture site) and risk assessment results, a pre-trained reinforcement learning model (the reward function includes pressure dispersion and comfort score) is called to generate the optimal turning angle sequence (such as 45° on the left side → supine position → 30° on the right side), calculate the best time for each position change, ensure that the interval between two turns is dynamically adjusted between 1.5-3 hours, use a biomechanical model to simulate the impact of position changes on the patient's muscles and bones to avoid excessive traction, and combine the patient's sleep monitoring data (using infrared sensors to identify changes in respiratory rate) to perform position adjustments preferentially during light sleep.

6. The method for dynamically adjusting the body position of a bedridden patient based on infrared sensing according to claim 1, characterized in that: In the S5, control instructions are sent to the segmented smart mattress via serial communication (baud rate ≥ 115200bps) to drive the inflation / deflation of independent air chambers to achieve ±0.5° accuracy in body position adjustment. During the execution process, the motor current is monitored in real time (the threshold is set to 1.2A) to prevent mechanical failure from causing patient injury. The pressure distribution changes are continuously monitored within 5 minutes after the adjustment. If the risk area is not effectively alleviated (PCI decreases by <20%), a secondary adjustment strategy is triggered to establish a patient position preference database, and the personalized position plan is continuously optimized through transfer learning.

Images