Iomt-based pressure ulcer prevention medical decision support system and method
An IoMT-based system uses sensors and AI to automatically adjust mattress position, addressing the inefficiencies of manual patient repositioning and reducing the risk of pressure ulcers while minimizing caregiver workload.
Patent Information
- Application Number
- KR1020230072268
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2023-06-05
- Publication Date
- 2026-11-06
- Estimated Expiration
- 2043-06-05
AI Technical Summary
Existing methods for preventing pressure ulcers, such as manually changing the position of immobile patients, are labor-intensive and inefficient, particularly for patients with high body weight, leading to increased workload for nurses and caregivers.
An IoMT-based medical decision support system that collects and analyzes monitoring data from a pressure ulcer prevention mattress, using sensors to automatically adjust the mattress position and posture based on AI-driven deep learning models, reducing the need for manual labor.
The system effectively prevents pressure ulcers by periodically changing the patient's position without human intervention, reducing the workload on medical personnel and ensuring patient comfort.
Smart Images

Figure 112023061898113-PAT00002_ABST
Abstract
Description
Technology Field
[0001] The present disclosure relates to an IoMT-based medical decision support system and method for preventing pressure ulcers, and specifically, to an IoMT-based medical decision support system and method that collects monitoring data regarding pressure control, motion control, and motion detection from a pressure ulcer prevention mattress and remotely controls changes in posture based on the monitoring data. Background Technology
[0003] Unless otherwise indicated in this specification, the contents described in this section are not prior art for the claims of this application, and are not to be recognized as prior art simply because they are included in this section.
[0004] A Medical Decision Support System (MDSS) is a computer-based system that assists medical professionals in making decisions. MDSS analyzes and interprets medical information, patient data, and medical knowledge to support and guide medical professionals in their decision-making. MDSS can be used for various purposes in the medical field. For example, it can be utilized as a system to aid diagnosis by analyzing various medical records and images, a system to prevent errors and improve safety during drug prescription, or a system to update decision-making by providing clinical guidelines and the latest research results. MDSS plays an auxiliary role in the decision-making of medical professionals and can help improve the accuracy and efficiency of medical diagnosis and treatment.
[0005] Meanwhile, bedsores are a condition in which the skin necroses due to body pressure caused by patients lying in the same position while unable to move. Bedsores exhibit a very high incidence rate, reaching 18 percent in critically ill patients, 28 percent in long-term hospitalized patients, and 29 percent in home healthcare settings. Korea is aging at a very rapid pace, and chronic degenerative diseases associated with this aging are also on the rise. Consequently, there is a very high demand for technologies to prevent the occurrence of bedsores.
[0006] Conventionally, in the case of hospitalized patients, nurses or caregivers have manually moved immobile patients to change their positions, as shown in Fig. 1. Since changing the patient's position must be performed periodically, this increases the workload for nurses and caregivers. Furthermore, for patients whose body weight exceeds a certain level, periodic position changes require heavy labor from nurses and caregivers, which is cited as a cause of nurses' early resignation. Prior art literature
[0008] 1. Korean Patent Publication No. 10-2022-0080936 (June 15, 2022) 2. Korean Patent Registration No. 10-1174025 (August 8, 2012) The problem to be solved
[0009] The IoMT-based medical decision support system and method for preventing bedsores according to the embodiment collects and analyzes monitoring data including pressure, motion information, and patient movement information of the bedsore prevention mattress through a sensor attached to the bedsore prevention mattress, and periodically controls the bedsore prevention mattress according to the results of the monitoring data analysis. Through this, the posture of a patient who is unable to move can be automatically changed without labor input, and in the case of an inpatient, the patient's body position can be changed by controlling the bedsore mattress in the ward from the nursing station.
[0010] In the embodiment, the necessary posture and position change cycle for the patient are calculated through an artificial intelligence-based deep learning model, and the pressure ulcer mattress can be controlled according to the calculated result.
[0011] In addition, in the embodiment, through the analysis of pressure ulcer mattress monitoring data, information on the condition of the pressure ulcer mattress, information on the patient's posture, and information on whether the patient has fallen can be transmitted to the administrator terminal and caregiver terminal of the nursing station. means of solving the problem
[0013] The artificial intelligence and IoMT-based medical decision support system for pressure ulcer prevention according to the embodiment
[0014] A pressure ulcer prevention mattress that collects monitoring data including pressure, tilt, operating status, and operating speed from sensors and patient status information, and transmits the collected monitoring data to a server; a server that receives monitoring data from the pressure ulcer prevention mattress, periodically controls the shape and operation of each pressure ulcer prevention mattress, and checks the status of each pressure ulcer prevention mattress after control; and a nursing station that collects and monitors patient status information from the pressure ulcer prevention mattress and remotely controls the pressure ulcer prevention mattress according to the patient status information; wherein the patient status information includes patient monitoring images generated by the pressure ulcer prevention mattress and bio-information obtained from external sensors. Effects of the invention
[0016] The IoMT-based medical decision support system and method for preventing pressure ulcers described above can periodically change the position of immobile patients by remotely controlling the pressure, tilt, and motion of the pressure ulcer prevention mattress using monitoring data collected from the mattress and patient information. Through this, pressure ulcers can be prevented and the labor of medical personnel can be significantly reduced.
[0017] In addition, in the embodiment, through the analysis of monitoring data collected from the pressure ulcer prevention mattress, it is possible to predict whether the mattress is functioning normally and identify information on changes in the patient's position and condition, thereby making patient management easier.
[0018] In addition, in the embodiment, the pressure ulcer prevention mattress is automatically controlled according to the predicted patient condition, allowing a patient with limited mobility to lie down in a more comfortable state.
[0019] The effects of the present invention are not limited to the effects described above, and should be understood to include all effects that can be inferred from the configuration of the invention described in the detailed description of the invention or the claims. Brief explanation of the drawing
[0021] Figure 1 is a diagram showing the process of changing a patient's position by a conventional nurse. FIG. 2 is a diagram showing the configuration of an IoMT-based medical decision support system for preventing pressure ulcers according to an embodiment. FIG. 3 is a diagram showing the data processing configuration of a pressure ulcer prevention mattress according to an embodiment. FIG. 4 is a diagram showing the data processing configuration of a server according to an embodiment. FIG. 5 is a signal flowchart of an IoMT-based medical decision support system for pressure ulcer prevention according to an embodiment. Specific details for implementing the invention
[0022] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components regardless of drawing symbols are assigned the same reference number, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not inherently possess distinct meanings or roles. Furthermore, in describing embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.
[0023] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.
[0024] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.
[0025] In this application, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.
[0026] In this specification, the term "part" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more hardware, and two or more units may be realized by one hardware.
[0027] Some of the operations or functions described herein as being performed by a terminal, device, or device may instead be performed by a server connected to said terminal, device, or device. Likewise, some of the operations or functions described as being performed by a server may also be performed by a terminal, device, or device connected to said server.
[0028] Hereinafter, the present invention will be described in detail with reference to the attached drawings.
[0029] Figure 2 is a diagram showing the configuration of an IoMT-based medical decision support system for preventing bedsores according to an embodiment.
[0030] Referring to FIG. 2, an IoMT-based medical decision support system for preventing bedsores according to an embodiment may be configured to include a bedsore prevention mattress (100), a server (200), and a nursing station (300). In the embodiment, the bedsore prevention mattress (100) collects patient condition information and bedsore prevention mattress monitoring information. In the embodiment, the patient condition information includes a patient monitoring image, and the bedsore prevention mattress monitoring information may include pressure, tilt, operating status, operating speed, current, voltage, frequency, etc., for each area of the mattress collected from each sensor configured on the mattress. In the embodiment, the bedsore prevention mattress (100) transmits the collected monitoring data and patient condition information to the server (200). In the embodiment, the bedsore prevention mattress (100) transmits the collected patient condition information to the nursing station (300).
[0031] The server (200) receives monitoring data and patient status information from the anti-bedsore mattress, periodically controls the shape and operation of each anti-bedsore mattress, and checks the status of each anti-bedsore mattress after control. In the embodiment, after checking the mattress status, the server (200) determines the mattress setting hardness control matching accuracy, setting operation cycle matching accuracy, setting operation range matching accuracy, and setting operation speed matching accuracy.
[0032] In addition, in the embodiment, the server (200) calculates the probability of a patient falling based on patient condition information and pressure ulcer prevention mattress monitoring information, and detects an abnormal condition. In addition, the server (200) can determine whether each mattress is malfunctioning by analyzing monitoring data collected from the pressure ulcer prevention mattress.
[0033] In the embodiment, the nursing station (300) collects and monitors patient condition information from the anti-bedsore mattress (100) and remotely controls the anti-bedsore mattress according to the patient condition information.
[0034] In the embodiments, patient condition information may include patient monitoring images generated by the anti-bedsore mattress and biometric information obtained from external sensors. In the embodiments, biometric information may include heart rate, blood pressure, temperature, pulse, etc. The artificial intelligence and IoMT-based anti-bedsore medical decision support system and method according to the embodiments collect and analyze monitoring data including pressure, motion information of the anti-bedsore mattress, and patient movement information through sensors attached to the anti-bedsore mattress, and periodically control the anti-bedsore mattress according to the results of the monitoring data analysis. Through this, the posture of a patient unable to move can be automatically changed without labor input, and in the case of an inpatient, the patient's body position can be changed by controlling the anti-bedsore mattress in the ward from the nursing station.
[0035] In the embodiment, the necessary posture and position change cycle for the patient are calculated through an artificial intelligence-based deep learning model, and the pressure ulcer mattress can be controlled according to the calculated result.
[0036] In addition, in the embodiment, through the analysis of pressure ulcer mattress monitoring data, information on the condition of the pressure ulcer mattress, information on the patient's posture, and information on whether the patient has fallen can be transmitted to the administrator terminal and caregiver terminal of the nursing station.
[0037] Figure 3 is a diagram showing the data processing configuration of a pressure ulcer prevention mattress according to an embodiment.
[0038] Referring to FIG. 3, a pressure sore prevention mattress according to an embodiment may be configured to include a sensor unit (110), a communication unit (130), and a control unit (150). The term “unit” as used herein should be interpreted to include software, hardware, or a combination thereof, depending on the context in which the term is used. For example, software may be machine language, firmware, embedded code, and application software. As another example, hardware may be a circuit, processor, computer, integrated circuit, integrated circuit core, sensor, MEMS (Micro-Electro-Mechanical System), passive device, or a combination thereof.
[0039] The sensor unit (110) collects sensing data from sensors installed on the pressure sore prevention mattress. In the embodiment, the sensors installed on the pressure sore prevention mattress may include, but are not limited to, motion detection sensors, pressure detection sensors, speed detection sensors, current, voltage, frequency monitoring sensors, tilt detection sensors, and cameras that photograph the patient to generate patient condition information as an image.
[0040] The communication unit (130) collects pressure ulcer prevention mattress monitoring data sensed from each sensor from the sensor unit (110), transmits it to the server, and transmits patient status information to the nursing station.
[0041] The control unit (150) receives a control signal from a server or a nursing station and controls the pressure sore prevention mattress according to the received control signal. In addition, in the embodiment, the control unit (150) can directly control the mattress according to patient condition information to change the patient's body position or posture. For example, if the control unit (150) detects an abnormal condition, such as when the probability of falling exceeds a certain level or when the patient maintains a breathing difficulty posture for more than a certain period of time, it can directly control the pressure sore prevention mattress to change the patient's body position or posture. In the embodiment, when an abnormal condition is detected, the control unit (150) notifies the nursing station. In addition, in the embodiment, if periodic motion of the patient is detected, the control unit (150) recognizes the periodic motion as an uncomfortable pattern and can control the pressure sore prevention mattress according to control information matched to the motion. For example, if the control unit (150) periodically detects a motion of the patient attempting to lift their leg, it can perform control to change the angle of the lower body part of the mattress. Additionally, the control unit (150) can perform control to increase the inclination of the upper body portion of the mattress when it periodically recognizes a motion in which the patient attempts to lift their upper body.
[0042] Figure 4 is a diagram showing the data processing configuration of a server according to an embodiment.
[0043] Referring to FIG. 4, the server (200) according to the embodiment may be configured to include a collection unit (210), a preprocessing unit (220), a deep learning unit (230), a generation unit (240), a detection unit (250), a judgment unit (260), and a feedback unit (270).
[0044] The collection unit (210) collects monitoring information and patient condition information from a plurality of pressure ulcer prevention mattresses. Additionally, the collection unit (210) may collect a training data set of a deep learning model. In an embodiment, the deep learning model may include a mattress control model, an abnormal state detection model, and a fault determination model. The training data set of the mattress control model may include sensor data, user feedback, control algorithms, system status information, etc. The training data set of the abnormal state detection model may include abnormal state information in the image state of the patient and mattress monitoring data for each abnormal state. The training data set of the fault determination model may include normal monitoring data and fault diagnosis history information, etc.
[0045] Additionally, in the embodiment, the collection unit (210) collects monitoring data of the anti-bedsore mattress through a thing board-based snap-in application for early IoT data collection processing. In the embodiment, the thing board is an open source IoT platform that collects and manages sensor data. In the embodiment, the collection unit (210) uses the thing board to collect and store monitoring data, which is sensor data of the anti-bedsore mattress.
[0046] Snap-in applications are extension features added to the Thing Board platform that perform specific functions or provide custom functions. In the case of a pressure ulcer prevention mattress, the snap-in application can be responsible for collecting, processing, and monitoring data from the mattress. In the embodiment, the collection unit (110) can collect various monitoring data generated from the pressure ulcer prevention mattress using a Thing Board-based snap-in application. The monitoring data may include the pressure and temperature of the mattress, the position and movement of the user's body, the time of mattress use, etc. In the embodiment, the collected data is stored on the Thing Board platform and can be integrated with other systems for real-time monitoring or analysis as needed. In the embodiment, the efficiency of the pressure ulcer prevention mattress can be evaluated based on the monitoring data, and the patient's health condition and preventive measures can be taken.
[0047] The preprocessing unit (220) preprocesses the collected training data to remove data that has bias or differentiation among the collected artificial intelligence learning training data sets. In the embodiment, the preprocessing unit (120) preprocesses the collected data to process it into a form suitable for learning an artificial intelligence model. For example, the preprocessing unit (220) can perform processes such as noise removal, outlier removal, and missing value processing. In addition, the preprocessing unit (120) can prevent the model from learning unnecessary patterns by normalizing the data, removing outliers, or adjusting the scale of the data through data preprocessing.
[0048] The deep learning unit (230) trains a deep learning neural network with a training data set to implement a deep learning model including a mattress control model, an abnormal state detection model, and a failure determination model. In the embodiment, the mattress control model is a deep learning model that controls the anti-bedsore mattress according to anti-bedsore monitoring data and patient state information to adjust the patient's body position and controls the anti-bedsore mattress by identifying control patterns received from the server and nursing station. The abnormal state detection model is a deep learning model that detects the patient's abnormal state according to patient state information and monitoring data obtained from the anti-bedsore mattress. In the embodiment, the patient's abnormal state may include cases where the probability of falling exceeds a certain level, where a breathing difficulty posture is maintained above a certain level, or where the probability of bedsore occurrence exceeds a certain level. The failure determination model is a deep learning model that analyzes the anti-bedsore mattress monitoring data to identify whether the anti-bedsore mattress is malfunctioning and the malfunctioning parts.
[0049] In the embodiments, the deep learning neural network may include at least one of a Deep Neural Network (DNN), a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), and a Bidirectional Recurrent Deep Neural Network (BRDNN), but is not limited thereto.
[0050] The generating unit (240) generates a pressure sore prevention mattress control signal based on the last change time of the pressure sore prevention mattress and patient condition information. In the embodiment, the pressure sore prevention mattress control signal may include signals that control the tilt, movement, and movement speed of each part of the mattress. In the embodiment, the generating unit (240) can automatically control the pressure sore prevention mattress by inputting the last control time and patient condition information into the mattress control model. To this end, in the embodiment, the generating unit (240) identifies the recent change event of the mattress by querying stored data in the system to determine the last change time of the mattress. The recent change event according to the embodiment includes the last time when the configuration, settings, or state of the mattress was changed. In addition, since mattress control must be adjusted according to the patient's condition, the generating unit (240) collects patient condition information from various sensors. For example, the generating unit (240) determines the patient's current condition by measuring the patient's heart rate, respiratory rate, body temperature, movement, etc.
[0051] Subsequently, the generating unit (240) generates a pressure sore prevention mattress control signal. In the embodiment, the pressure sore prevention mattress control signal of the mattress is generated based on the time of the last change and patient condition information. To this end, the control algorithm utilizes a previously trained model or follows predefined rules. For example, the control algorithm generates control signals such as adjusting pressure on a specific location of the patient or automatically adjusting the position of the mattress to distribute pressure when the patient moves. In the embodiment, the generated control signal is transmitted to the mattress to actually control the pressure sore prevention mattress. The mattress allows for appropriate pressure distribution to prevent pressure sores while keeping the patient comfortable, and enables safe changes in the patient's body position.
[0052] In the embodiment, the generating unit (240) receives status information from the pressure ulcer prevention mattress after transmitting a control signal and checks whether the control signal is reflected. In the embodiment, the status information includes the hardness, operation cycle, and operation speed of the mattress. In the embodiment, the generating unit (240) calculates the mattress's set hardness control matching accuracy, set operation cycle matching accuracy, set operation range matching accuracy, and set operation speed matching accuracy as the ratio of the mattress's hardness, operation cycle, and operation speed included in the control signal to the mattress's hardness, operation cycle, and operation speed that is actually controlled. In the embodiment, if the calculated set hardness control matching accuracy, set operation cycle matching accuracy, set operation range matching accuracy, and set operation speed matching accuracy are below a certain level, the generating unit (240) determines that it is a remote control error and can notify the administrator and nursing station of this.
[0053] The detection unit (250) monitors patient status information to detect abnormal conditions in the patient. In the embodiment, abnormal conditions may include a patient's risk of falling exceeding a certain level, a breathing difficulty posture, etc. In the embodiment, the detection unit (250) may detect abnormal conditions by analyzing images of the patient taken from the patient status information.
[0054] In the embodiment, the detection unit (250) can detect an abnormal condition of a patient through an image recognition process using artificial intelligence machine learning. Artificial intelligence image recognition is a process in which a machine recognizes objects and understands scenes from photographs just like a human, and it corresponds to one of the computer vision technologies. In the embodiment, for image recognition, a data processing process is performed to classify and detect patient objects included in the patient's captured image, and to identify and segment objects at the pixel level. In the embodiment, the detection unit (250) performs an out-of-distribution detection process to handle patterns that have not been learned, in addition to noise response. Out-of-distribution detection is the identification of whether an image input to the artificial intelligence is a learned probability distribution data or not. In the embodiment, through out-of-distribution detection, images that are difficult for the artificial neural network to judge are filtered out or exceptions are processed, thereby increasing stability and reliability. In the embodiment, to detect non-learning distribution data, the probability value indicating how confident the deep learning judgment is is calibrated, or non-learning distribution data is generated and learned using a Generative Adversarial Network (GAN) to improve detection accuracy.
[0055] In addition, in the embodiment, to reduce the model size while maintaining image recognition accuracy, a lightweight deep learning technique that simplifies computation is used to finalize the body shape box. In the embodiment, for image recognition, data lightweighting is enabled by modifying convolution filters in a Convolutional Neural Network (CNN) to reduce computational dimensions, pruning to delete weights of the neural network that have little impact, and quantization to simplify computation by reducing floating-point numbers of weight values. Furthermore, in the embodiment, the output of a pre-trained large neural network is mimicked by a small neural network to simplify computation while maintaining accuracy.
[0056] In addition, in the embodiment, the detection unit (250) can input patient status information into an abnormal state detection model to obtain patient abnormal state information.
[0057] Additionally, the detection unit (250) collects patient condition information using various sensors to detect abnormal conditions. For example, the detection unit (250) monitors the patient's movement using an accelerometer sensor and detects breathing sounds using a breathing noise sensor. Additionally, it collects biosignal data such as the patient's heart rate, oxygen saturation, and body temperature. Subsequently, the detection unit (250) evaluates the patient's condition by analyzing the collected sensor data. For example, the accelerometer data is analyzed to detect falls, and the breathing noise data is used to identify abnormal breathing patterns, including postures of difficulty breathing. Furthermore, the detection unit (250) can identify abnormal conditions by comparing the collected biosignal data with normal ranges. In the embodiment, the detection unit (250) uses an abnormal condition detection algorithm and an abnormal condition detection model to perform abnormal condition detection based on condition analysis. In the embodiment, the detection unit (250) evaluates the patient's condition using the collected data through the abnormal condition detection algorithm and the abnormal condition detection model, and determines whether there is an abnormal condition by comparing it with predefined criteria. In the embodiment, the detection unit (250) generates an abnormal condition notification to take appropriate action when an abnormal condition is detected. To do this, it may send a warning message or send a warning signal to medical staff. Additionally, in the embodiment, the detection unit (250) performs a patient fall prediction by analyzing patient condition information and monitoring data of the pressure ulcer prevention mattress, and if the probability of a fall is calculated to be above a certain level through the fall prediction, it transmits the fall prediction result to the nursing station and generates a pressure ulcer prevention mattress control signal for fall prevention.
[0058] In addition, in the embodiment, the sensing unit (250) identifies patient characteristic information including user sensitivity, activity level, and patient posture through patient condition information and pressure ulcer prevention mattress monitoring information, and inputs the patient characteristic information into a fall prediction model to obtain the possibility of a fall. Subsequently, the patient characteristic information is input into a pressure ulcer occurrence prediction model to obtain the possibility of pressure ulcer occurrence.
[0059] For example, the sensing unit (250) identifies user sensitivity, activity level, and patient posture. In an embodiment, the sensing unit (250) analyzes patient condition information and monitoring information of a pressure ulcer prevention mattress to identify characteristics related to the user's sensitivity, activity level, and patient posture. In an embodiment, the sensing unit (250) can infer the user's sensitivity and activity level by analyzing the patient's movement patterns, changes in posture, pressure distribution requirements, etc. Additionally, the sensing unit (250) constructs patient characteristic information based on the user's sensitivity, activity level, and characteristics related to the patient's posture. In an embodiment, the patient characteristic information may include the patient's movement ability, muscle strength, and risk factors for pressure ulcer development. Subsequently, the sensing unit (250) calculates the patient's fall probability through a trained fall prediction model to predict the possibility of a fall. In an embodiment, the fall prediction model receives the patient's characteristic information as input and evaluates the patient's fall probability. For example, a machine learning algorithm can be used to predict the possibility of a fall based on the patient's characteristic information, previous fall history, and the patient's movement patterns. Additionally, the sensing unit (250) evaluates the likelihood of developing a pressure ulcer by inputting patient characteristic information into a trained pressure ulcer development prediction model to predict the likelihood of developing a pressure ulcer. In the embodiment, the pressure ulcer development prediction model can predict the likelihood of developing a pressure ulcer by using information such as the patient's posture, pressure distribution requirements, pressure points, etc.
[0060] The judgment unit (260) determines whether a mattress is faulty by analyzing monitoring data of the pressure sore prevention mattress. In the embodiment, the judgment unit (260) collects monitoring data of the mattress through sensors or IoT devices of the pressure sore prevention mattress. The mattress monitoring data is information that measures various parameters and conditions of the mattress. Subsequently, the collected data is preprocessed and processed into a form suitable for analysis. In this step, tasks such as noise removal, outlier processing, and data normalization can be performed. The preprocessed data is used as input to a fault determination model. In the embodiment, the judgment unit (260) evaluates the condition of a new mattress using the fault determination model. For example, the judgment unit (260) inputs the mattress monitoring data into the model, and the model predicts whether a fault exists based on the input data. For example, the model can determine that the mattress is faulty if the pressure distribution deviates from the normal range or if a specific fault pattern appears. Subsequently, the judgment unit (260) notifies the fault determination result. This can be done by sending notifications to mattress managers, maintenance personnel, and nursing stations, or by sending warning signals to the system.
[0061] The feedback unit (270) evaluates the learned artificial neural network model and the deep learning model. In an embodiment, the feedback unit (270) can evaluate the artificial neural network model through at least one of accuracy, precision, and recall. Accuracy is an indicator that measures how closely the result predicted by the artificial neural network model matches the actual result. Precision is an indicator that measures the ratio of actual positive results among the results predicted as positive. Recall is an indicator that measures the ratio of actual positive results predicted as positive by the model. In an embodiment, the feedback unit (270) calculates the accuracy, precision, and recall of the artificial neural network model and evaluates the artificial neural network model based on at least one of the calculated indicators.
[0062] In the embodiment, the feedback unit (270) can measure the accuracy of the artificial neural network model using an evaluation dataset. The evaluation dataset consists of data that the model has not used for training and is used to objectively evaluate the performance of the model. In the embodiment, the feedback unit (270) runs the artificial neural network model using the evaluation dataset and compares the predicted value of the artificial neural network model for each input data with the actual correct value of the corresponding data. Subsequently, the accuracy of the model's prediction can be measured through the comparison results. For example, in the feedback unit (270), the accuracy can be calculated as the ratio of the data correctly predicted by the model out of the total data.
[0063] Additionally, the feedback unit (270) can calculate an F1 score, which is an indicator representing the balance of precision and recall calculated as the harmonic mean of precision and recall, evaluate the artificial neural network model based on the calculated F1 score, generate an AUC-ROC curve, which is an indicator visualizing the performance of the classification model as a graph, and evaluate the artificial neural network model based on the generated AUC-ROC curve. In the embodiment, the feedback unit (270) can evaluate the model's performance as being better as the area under the ROC curve (AUC) is closer to 1.
[0064] Additionally, the feedback unit (270) can evaluate the interpretability of the artificial neural network model. In the embodiment, the feedback unit (270) evaluates the interpretability of the artificial neural network model through SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) methods. SHAP (SHapley Additive exPlanations) is a library that provides interpretations of the results predicted by the model, and the feedback unit (270) extracts SHAP values from the library. In the embodiment, the feedback unit (270) can predict how much the feature information input into the model influenced the model prediction by extracting SHAP values.
[0065] The LIME (Local Interpretable Model-agnostic Explanations) method is a method for explaining the model's predictions for individual samples. In the embodiment, the feedback unit (270) approximates the sample into an interpretable model through the LIME method and calculates the importance of each characteristic information. Additionally, the feedback unit (270) can estimate the influence of each characteristic variable by analyzing the model's internal weights and bias values.
[0066] The feedback unit (270) performs improvement work when the fairness of the artificial neural network model is low or discriminative. In the embodiment, if the data of a specific group is insufficient by more than a certain level, the feedback unit (270) collects additional data representing the specific group and performs a data preprocessing process. In the embodiment, the feedback unit (270) performs a data preprocessing process including data normalization, outlier removal, and data scaling to prevent the model from learning unnecessary patterns. Additionally, in the embodiment, the feedback unit (270) can prevent discriminativeness or guarantee fairness by adding specific conditions to the model learning algorithm.
[0067] In the embodiment, the feedback unit (270) evaluates the performance of the model by comparing the model's predicted results with the actual results through confusion matrix analysis to ensure fairness. A confusion matrix is a matrix used to evaluate the classification performance of a model in supervised learning. The confusion matrix displays the classification results by comparing the results predicted by the model with the actual results. In the embodiment, the feedback unit (270) can evaluate the performance of the model by calculating the accuracy and misclassification rate for each class through confusion matrix analysis.
[0068] In addition, in the embodiment, the feedback unit (270) enables the distribution of data to be verified through visualization analysis of the training data. For example, in the case of image data, the diversity and fairness of the data can be evaluated by visualizing image samples for each class.
[0069] Additionally, the feedback unit (270) performs bias verification of the artificial neural network model. In the embodiment, the feedback unit (270) checks whether the model is biased toward a specific class or attribute through bias verification of the training information. To this end, the feedback unit (270) compares the number of samples for each class or evaluates the classification performance for each class.
[0070] Additionally, the feedback unit (270) verifies the fairness and diversity of the training data and improves the artificial neural network model through fairness verification and the calculation of evaluation indicators. In the embodiment, fairness verification is to check whether the artificial neural network model shows differentiation regarding specific attributes such as gender, race, and age in relation to the training information. In the embodiment, the feedback unit (270) can verify whether there is differentiation regarding specific attributes by comparing the number of samples for each attribute or by evaluating the classification performance for each attribute.
[0071] Additionally, the feedback unit (270) calculates various metrics to evaluate the performance of the artificial neural network model. For example, the performance of the model can be evaluated by calculating metrics such as accuracy, precision, recall, and F1-score. At this time, the fairness and diversity of the model can be evaluated by calculating metrics for each class.
[0072] Additionally, the feedback unit (270) collects feedback on problems that occur while the artificial neural network model is used in a real environment, and continuously improves the artificial neural network model by reflecting the collected feedback in the artificial neural network model.
[0073] Below, a medical decision support method for preventing bedsores based on artificial intelligence and IoMT will be described in turn. Since the operation (function) of the medical decision support method for preventing bedsores based on artificial intelligence and IoMT according to the embodiment is essentially the same as the function of the system, descriptions that overlap with FIGS. 2 to 4 will be omitted.
[0074] Figure 5 is a signal flow diagram of an artificial intelligence and IoMT-based medical decision support system for preventing bedsores according to an embodiment.
[0075] Referring to FIG. 5, in step S100, monitoring data including pressure, tilt, operating status, and operating speed, and patient status information are collected from sensors in the anti-bedsore mattress, and the collected monitoring data is transmitted to a server. In step S200, the server receives monitoring data from the anti-bedsore mattress, periodically controls the shape and operation of each anti-bedsore mattress, and checks the status of each anti-bedsore mattress after control. In step S300, the nursing station collects and monitors patient status information from the anti-bedsore mattress, and remotely controls the anti-bedsore mattress according to the patient status information. In the embodiment, the patient status information may include patient monitoring images generated by the anti-bedsore mattress and biometric information obtained from external sensors. In the embodiment, the biometric information may include heart rate, blood pressure, temperature, pulse, etc.
[0076] The artificial intelligence and IoMT-based medical decision support system and method for preventing pressure ulcers described above can periodically change the position of immobile patients by remotely controlling the pressure, tilt, and motion of the pressure ulcer prevention mattress using monitoring data collected from the mattress and patient information. Through this, pressure ulcers can be prevented and the labor of medical personnel can be significantly reduced.
[0077] In addition, in the embodiment, through the analysis of monitoring data collected from the pressure ulcer prevention mattress, it is possible to predict whether the mattress is functioning normally and identify information on changes in the patient's position and condition, thereby making patient management easier.
[0078] In addition, in the embodiment, the pressure ulcer prevention mattress is automatically controlled according to the predicted patient condition, allowing a patient with limited mobility to lie down in a more comfortable state.
[0079] The disclosed content is merely illustrative and can be modified and implemented in various ways by a person skilled in the art without departing from the gist of the claim in the patent claims; therefore, the scope of protection of the disclosed content is not limited to the specific embodiments described above.
Claims
Claim 1 A pressure ulcer prevention mattress that collects monitoring data including pressure, tilt, operating status, and operating speed from sensors and patient status information, and transmits the collected monitoring data to a server; a server that receives monitoring data from the pressure ulcer prevention mattress, periodically controls the shape and operation of each pressure ulcer prevention mattress, and checks the status of each pressure ulcer prevention mattress after control; and a nursing station that collects and monitors patient status information from the pressure ulcer prevention mattress and remotely controls the pressure ulcer prevention mattress according to the patient status information; wherein the patient status information includes patient monitoring images generated by the pressure ulcer prevention mattress and bio-information obtained from external sensors, and the pressure ulcer prevention mattress includes a sensor unit comprising a motion detection sensor, a pressure detection sensor, a speed detection sensor, current, voltage, and frequency monitoring sensors, and a camera that photographs the patient to generate patient status information as an image; and a communication unit that collects pressure ulcer prevention mattress monitoring data from the sensor unit and transmits it to a server, and transmits patient status information to a nursing station. An IoMT-based medical decision support system for preventing bedsores, characterized by including a control unit that receives a control signal from a server or nursing station and controls a bedsore prevention mattress according to the received control signal. Claim 2 delete Claim 3 The IoMT-based medical decision support system for preventing bedsores according to claim 1, wherein the server comprises: a collection unit that collects monitoring information from a plurality of bedsore prevention mattresses; a generation unit that generates a bedsore prevention mattress control signal according to the last change time of the bedsore prevention mattresses and patient condition information; a detection unit that detects abnormal conditions including patient falls and respiratory distress postures by monitoring patient condition information; and a judgment unit that determines whether there is a failure for each mattress through the analysis of monitoring data of the bedsore prevention mattresses. Claim 4 In claim 3, the server is characterized by implementing a deep learning model including a matrix control model, an abnormal state detection model, and a failure determination model by training a deep learning neural network including at least one of a Deep Neural Network (DNN), a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), and a Bidirectional Recurrent Deep Neural Network (BRDNN) on a training dataset. Claim 5 In paragraph 3, the above-mentioned sensing unit performs a fall prediction of a patient through the analysis of patient condition information and monitoring data of a pressure ulcer prevention mattress, and if the probability of a fall is calculated to be above a certain level through the fall prediction, transmits the fall prediction result to a nursing station and generates a pressure ulcer prevention mattress control signal for fall prevention, characterized in that it is an IoMT-based pressure ulcer prevention medical decision support system. Claim 6 In claim 5, the above-mentioned sensing unit; is characterized by identifying patient characteristic information including user sensitivity, activity level, and patient posture through patient condition information and pressure ulcer prevention mattress monitoring information, inputting the patient characteristic information into a fall prediction model to obtain the probability of a fall, and inputting the patient characteristic information into a pressure ulcer occurrence prediction model to obtain the probability of a pressure ulcer occurrence, in an IoMT-based pressure ulcer prevention medical decision support system. Claim 7 delete
Citation Information
Patent Citations
Pressure ulcer patient managing system using artificial inteeligence and oprating method thereof
KR1020200135052A
A system for controlling the Anti-pearl mattress using wearable device
KR1020210109889A
Fall prevention system and fall prevention method using dual camera and infrared camera
KR102205964B1
