Lung monitoring method and system based on electrical impedance tomography images and respiratory sound images

Abstract

The present invention relates to an operating method of a lung monitoring system operated by at least one processor, the operating method comprising the steps of: acquiring a plurality of electrical impedance tomography (EIT) images of the chest of a patient and a plurality of respiratory sound intensity images generated on the basis of respiratory sound signals acquired from the chest of the patient; mapping, on the basis of a plurality of pieces of respiratory interval information acquired by binning a respiratory signal of the patient, an EIT image and a respiratory sound intensity image corresponding to each respiratory interval to each other; and inputting, into a respiratory anomaly prediction model, at least one pair of the EIT image and the respiratory sound intensity image included in each respiratory interval, to acquire a prediction result for respiratory anomalies of the patient.

Description

Method and system for lung monitoring based on electrical impedance tomography and breath sound imaging

[0001] The present disclosure relates to an artificial intelligence-based lung monitoring technology.

[0002] As the aging population accelerates, the importance of lung diseases is increasing, and when lung diseases progress to a severe stage, the rate of mechanical ventilation use is higher compared to other diseases. In such cases, it is crucial to maintain appropriate ventilation by monitoring the lungs in real time during mechanical ventilation.

[0003] For example, patients with acute respiratory distress syndrome (ARDS) require mechanical ventilation due to hypoxia caused by inflammatory lung damage from various causes. In this case, if the optimal respiratory volume is not maintained, ventilator-induced lung injury (VILI) may occur. Therefore, to minimize the occurrence of VILI, a lung protective ventilation (LPV) strategy is required to set the optimal positive end-expiratory pressure (PEEP) and tidal volume.

[0004] To this end, Electrical Impedance Tomography (EIT) technology has recently been actively used for lung-protective ventilation monitoring. EIT is a technology that applies an electric current by attaching electrodes around the patient's chest and visualizes the distribution of electrical impedance within the body through the measured potential difference. While EIT offers the advantage of enabling non-invasive, real-time lung ventilation monitoring without the risk of radiation exposure, it has limitations in that accurate diagnosis of lung diseases is difficult due to lower resolution compared to CT or X-ray. Furthermore, because EIT visualizes lung cross-sections based on changes in electrical impedance, it is difficult to accurately identify minute airflow patterns and changes within the lungs, and it is challenging to obtain precise information regarding ventilation volume and distribution.

[0005] Meanwhile, a patient's lung sounds (or respiratory sounds) are also an important indicator for diagnosing and predicting lung diseases. However, lung sounds are unstructured data in which the characteristics of each respiratory disease are not quantified, and if medical professionals directly auscultate a patient's lung sounds, diagnostic results may vary depending on the practitioner, or there may be risks of infection for the medical staff. To address this, non-contact lung sound measurement devices are used; however, conventional devices cannot measure lung sounds from multiple chest regions, so their accuracy may be lower compared to actual auscultation.

[0006] Therefore, a method is required to monitor a patient's lungs based on the patient's EIT images and breath sound data acquired from multiple chest regions, and to accurately predict signs of respiratory abnormalities.

[0007] The present disclosure provides a lung monitoring method and system based on EIT images and breath sound images.

[0008] The present disclosure provides a method and system for segmenting a continuous respiratory signal into multiple respiratory cycles, extracting and synchronizing EIT images and respiratory sound intensity images corresponding to each respiratory segment, and using the synchronized images to predict signs of respiratory abnormalities based on artificial intelligence and monitor the lungs.

[0009] According to one embodiment, a method of operation of a lung monitoring system operated by at least one processor comprises: acquiring a plurality of electrical impedance tomography (EIT) images of a patient’s chest and a plurality of breath sound intensity images generated based on breath sound signals acquired from the patient’s chest; mapping the EIT images and breath sound intensity images corresponding to each breath segment based on a plurality of breath segment information acquired by binning the patient’s breath signal; and inputting at least one pair of EIT images and breath sound intensity images included in each breath segment into a breath abnormality prediction model to obtain a breath abnormality prediction result of the patient.

[0010] The step of acquiring the plurality of breath sound intensity images may include: acquiring breath sound intensity by measurement location from a multi-channel breath sound signal measured from at least one chest region of the patient; blending images representing the distribution of breath sound intensity by measurement location to generate a breath sound intensity image for a specific point in time; and repeatedly generating the breath sound intensity image for a specific point in time according to a set time to generate the plurality of breath sound intensity images.

[0011] The above breath sound intensity may be a value corresponding to at least one of a peak value, effective value, minimum-maximum average, peak vector, rise-fall time, and amplitude for the signal obtained in the time domain of the multi-channel breath sound signal, or a value corresponding to at least one of an amplitude, phase, and period for the signal obtained in the frequency domain of the multi-channel breath sound signal.

[0012] The mapping step described above may include the step of extracting a plurality of breathing segments corresponding to a preset amplitude range or phase range from the breathing signal.

[0013] The mapping step may further include a step of extracting a plurality of abnormal breathing segments from the breathing signal by analyzing at least one of the breathing cycle, inspiratory time fraction, and breathing depth of the breathing signal.

[0014] Prior to the step of obtaining the above-mentioned respiratory abnormality prediction result, the method may include the step of generating a training dataset for training the respiratory abnormality prediction model using at least one pair of EIT images and respiratory sound intensity images, and the step of training the respiratory abnormality prediction model using the training dataset.

[0015] The above training dataset may include at least one of a first training dataset generated by concatenating at least one EIT image and at least one breath sound intensity image, a second training dataset generated by performing a domain transform on at least one EIT image or at least one breath sound intensity image and combining at least one EIT image or at least one breath sound intensity image with the domain-transformed image, and a third training dataset generated by combining a patient's biosignal with the first training dataset or the second training dataset.

[0016] After the step of obtaining the above-mentioned respiratory abnormality prediction result, the method further includes the step of providing at least one pair of EIT images and respiratory sound intensity images included in each of the above-mentioned respiratory segments, and the above-mentioned respiratory abnormality prediction result to an external terminal, wherein the providing step may include the step of providing the area related to the above-mentioned respiratory abnormality prediction result in the EIT image or respiratory sound intensity image by distinguishing it according to importance.

[0017] The above-mentioned step may further include the step of receiving and providing information regarding biosignals pre-set by the user from a ventilator linked to the patient.

[0018] According to one embodiment, a lung monitoring system operated by at least one processor comprises: an image processing unit that acquires a plurality of electrical impedance tomography (EIT) images of a patient’s chest and a plurality of breath sound intensity images generated based on breath sound signals acquired from the patient’s chest; a synchronization unit that maps the EIT images and breath sound intensity images corresponding to each breath segment based on a plurality of breath segment information acquired by binning the patient’s breath signal; and a breath abnormality prediction unit that inputs at least one pair of EIT images and breath sound intensity images included in each breath segment into a breath abnormality prediction model to obtain a breath abnormality prediction result of the patient.

[0019] The image processing unit can obtain breath sound intensity by measurement location from a multi-channel breath sound signal measured from at least one chest region of the patient, blend an image showing the distribution of breath sound intensity by measurement location into one to generate a breath sound intensity image for a specific point in time, and repeatedly generate the breath sound intensity image for the specific point in time according to a set time to generate the plurality of breath sound intensity images.

[0020] The above breath sound intensity may be a value corresponding to at least one of a peak value, effective value, minimum-maximum average, peak vector, rise-fall time, and amplitude for the signal obtained in the time domain of the multi-channel breath sound signal, or a value corresponding to at least one of an amplitude, phase, and period for the signal obtained in the frequency domain of the multi-channel breath sound signal.

[0021] The above system may include a breathing cycle segmentation unit that extracts a plurality of breathing segments corresponding to a preset amplitude range or phase range from the breathing signal.

[0022] The above-mentioned breathing cycle segmentation unit can further extract a plurality of abnormal breathing segments from the breathing signal by analyzing at least one of the breathing cycle, inhalation time fraction, and breathing depth of the breathing signal.

[0023] The above-described respiratory abnormality prediction unit may include a training data generation unit that generates a training dataset for training the respiratory abnormality prediction model using at least one pair of EIT images and respiratory sound intensity images, and a training unit that trains the respiratory abnormality prediction model using the training dataset.

[0024] The above training dataset may include at least one of a first training dataset generated by concatenating at least one EIT image and at least one breath sound intensity image, a second training dataset generated by performing a domain transform on at least one EIT image or at least one breath sound intensity image and combining at least one EIT image or at least one breath sound intensity image with the domain-transformed image, and a third training dataset generated by combining a patient's biosignal with the first training dataset or the second training dataset.

[0025] The apparatus further includes at least one pair of EIT images and breath sound intensity images included in each of the above-mentioned breathing intervals, and an information providing unit that provides the breath abnormality prediction results to an external terminal, wherein the information providing unit may display and provide areas related to the breath abnormality prediction results in the EIT images or breath sound intensity images according to importance.

[0026] The above information providing unit can receive and further provide information regarding biosignals pre-set by the user from a ventilator linked to the patient.

[0027] According to the present disclosure, since longitudinal analysis is possible by combining the patient's EIT images and breath sound intensity images, signs of respiratory abnormalities can be predicted more accurately compared to single data.

[0028] According to the present disclosure, by obtaining standardized cohort data regarding the characteristics of lung disease and signs of respiratory abnormalities, it can be utilized in various medical fields.

[0029] According to the present disclosure, since conventional EIT systems can be utilized, they can be extended and applied to various medical fields.

[0030] According to the present disclosure, by providing lung management information to medical personnel, including results of predicting signs of respiratory abnormalities, synchronized EIT images, and breath sound intensity images, a lung protective ventilation (LPV) strategy suitable for the patient can be implemented.

[0031] FIG. 1 is a conceptual diagram of a lung monitoring system according to one embodiment.

[0032] FIG. 2 is a configuration diagram of a lung monitoring system according to one embodiment.

[0033] FIG. 3 is a diagram illustrating a method for generating a breath sound intensity image according to one embodiment.

[0034] FIG. 4 is a drawing illustrating a breathing section according to one embodiment.

[0035] FIG. 5 is a diagram illustrating synchronized EIT images and breath sound intensity images according to one embodiment.

[0036] FIG. 6 is a diagram illustrating a respiratory abnormality sign prediction unit according to one embodiment.

[0037] FIG. 7 is a diagram illustrating a method for generating a training dataset according to one embodiment.

[0038] FIG. 8 is a flowchart of a lung monitoring method according to one embodiment.

[0039] FIG. 9 is a hardware configuration diagram of a lung monitoring system according to one embodiment.

[0040] Embodiments of the present disclosure are described below with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present disclosure may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0041] In the present disclosure, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "…part," "…unit," and "…module" as used in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.

[0042] In the present disclosure, the devices are composed of hardware including at least one processor, a memory device, a communication device, etc., and a program that is executed in combination with the hardware is stored in a designated location. The hardware has a configuration and performance capable of executing the method of the present invention. The program includes instructions that implement the method of operation of the present invention described with reference to the drawings, and executes the present invention in combination with hardware such as a processor and a memory device.

[0043] In the present disclosure, "transmission or provision" may include not only direct transmission or provision but also indirect transmission or provision through another device or by using an alternative route.

[0044] Expressions described in the singular in this disclosure may be interpreted as singular or plural unless explicit expressions such as "one" or "single" are used.

[0045] In the present disclosure, the same reference numerals refer to the same components regardless of the drawings, and "and / or" includes each of the mentioned components and all combinations of one or more.

[0046] In this disclosure, 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. Such terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of this disclosure, the first component may be named the second component, and similarly, the second component may be named the first component.

[0047] In the flowchart described with reference to the drawings in this disclosure, the order of operations may be changed, several operations may be merged or some operations may be divided, and certain operations may not be performed.

[0048] In the present disclosure, an artificial intelligence model is an artificial intelligence model that learns at least one task and may be implemented as a computer program executed on a computing device. The computer program is stored on a non-transitory storage media and includes instructions described to execute the operation of the present disclosure by a processor. The computer program may be downloaded over a network or sold in the form of a product.

[0049] FIG. 1 is a conceptual diagram of a lung monitoring system according to one embodiment.

[0050] Referring to FIG. 1, a lung monitoring system based on electrical impedance tomography images and breath sound images (simply put, lung monitoring system) (10) is a computing device operated by at least one processor, comprising a memory for storing instructions and a processor for executing instructions, and the processor performs the operation of the present disclosure by executing instructions included in a computer program.

[0051] The lung monitoring system (10) can perform binning on the input respiratory signal. That is, the lung monitoring system (10) can divide a continuous respiratory signal into multiple respiratory bins according to specific conditions. At this time, the respiratory signal may be a signal that measures physical changes resulting from the patient's breathing movements. For example, the respiratory signal may be a signal in the form of a sinusoidal wave representing changes occurring in the movement of the chest or abdomen, air pressure, etc., during breathing.

[0052] The lung monitoring system (10) can perform synchronization between an Electrical Impedance Tomography (EIT) image and a breath sound intensity image based on acquired breathing interval information. At this time, the synchronized EIT image and breath sound intensity image can be input into a breathing abnormality prediction model (430).

[0053] Here, the EIT image is an image of the lung cross-section formed by visualizing impedance information measured from multiple electrodes attached to the patient's chest, and is a time-series data representing the ventilation distribution of the lungs as color changes by region. The EIT system can generate EIT images based on the measured impedance information using methods such as back projection, iterative reconstruction, and deep learning. As EIT technology and EIT imaging methods are already known, a detailed description is omitted in this disclosure.

[0054] Additionally, the breath sound intensity image may be time-series data representing breath sound intensity according to measurement location as a change in brightness, utilizing breath sound signals measured in multiple thoracic regions of the patient. In this case, the breath sound signal may be a signal measured from sounds generated in the patient's airway or lungs.

[0055] That is, the lung monitoring system (10) can receive multiple EIT images and multiple breath sound intensity images generated at regular time intervals from the chest of the same patient, and can perform synchronization by extracting the EIT image and breath sound intensity image corresponding to each breath segment based on the breath segment information obtained from the breath signal.

[0056] The lung monitoring system (10) can predict a patient's respiratory abnormality using a respiratory abnormality prediction model (430) trained to predict respiratory abnormality signs, including lung disease, from synchronized EIT images and breath sound intensity images. At this time, the respiratory abnormality prediction model (430) is an artificial intelligence model capable of learning at least one task and can be implemented in the form of software or a program executed on a computing device.

[0057] The respiratory abnormality prediction model (430) may be learned and installed by a separate learning device, or the lung monitoring system (10) may generate learning data and train the respiratory abnormality prediction model (430) based on the learning data.

[0058] The lung monitoring system (10) can provide synchronized EIT images and breath sound intensity images, along with the results of predicting signs of respiratory abnormalities, to a system connected to the hospital or an external terminal via a network. At this time, the external terminal may be a smartphone, a PC, etc., but is not necessarily limited thereto, and may be any device capable of outputting the results of predicting signs of respiratory abnormalities.

[0059] Therefore, the lung monitoring system (10) can perform a longitudinal analysis by combining the patient's EIT image and breath sound intensity image, so it can predict signs of respiratory abnormalities more accurately compared to when using single data.

[0060] FIG. 2 is a configuration diagram of a lung monitoring system according to one embodiment.

[0061] FIG. 3 is a diagram illustrating a method for generating a breath sound intensity image according to one embodiment.

[0062] FIG. 4 is a drawing illustrating a breathing section according to one embodiment.

[0063] FIG. 5 is a diagram illustrating synchronized EIT images and breath sound intensity images according to one embodiment.

[0064] Referring to FIG. 2, the lung monitoring system (10) includes an image processing unit (100), a respiratory cycle segmentation unit (200), a synchronization unit (300), a respiratory abnormality prediction unit (400), and an information provision unit (500).

[0065] In the present disclosure, the image processing unit (100), the respiratory cycle segmentation unit (200), the synchronization unit (300), the respiratory abnormality prediction unit (400), and the information provision unit (500) are referred to as a computing device operated by at least one processor. Here, the image processing unit (100), the respiratory cycle segmentation unit (200), the synchronization unit (300), the respiratory abnormality prediction unit (400), and the information provision unit (500) may be implemented in a single computing device or distributed across separate computing devices. When distributed across separate computing devices, the image processing unit (100), the respiratory cycle segmentation unit (200), the synchronization unit (300), the respiratory abnormality prediction unit (400), and the information provision unit (500) may communicate with each other through a communication interface. For example, the synchronization unit (300) may remotely receive data via a network from the image processing unit (100) or the respiratory sound segmentation unit (200) implemented in a separate computing device.

[0066] The image processing unit (100) may include an EIT image processing unit (110) that processes the EIT image of the patient received and a breath sound intensity image processing unit (120) that generates a breath sound intensity image from a breath sound signal.

[0067] The EIT image processing unit (110) can receive EIT images from an external EIT system connected via a network, and the EIT image processing unit (110) can perform preprocessing operations on the received EIT images, such as improving resolution or removing noise.

[0068] The operation of the breath sound intensity image processing unit (120) is explained below with reference to FIG. 3.

[0069] Referring to FIG. 3, the breath sound intensity image processing unit (120) can receive a breath sound signal from the breath sound measurement system (20). At this time, the breath sound measurement system (20) may be a multi-channel breath sound measurement system capable of measuring multi-channel breath sound signals in multiple chest regions of a patient, and may perform a preprocessing process to remove noise such as heart sounds other than breath sounds from the measured breath sound signal.

[0070] For example, the breath sound measurement system (20) can perform preprocessing such as resampling, normalization, and noise reduction on the breath sound signal, and the breath sound measurement system (20) can also perform preprocessing using other known methods.

[0071] The breath sound measurement system (20) can measure breath sound signals from multiple chest regions by using multiple wired or wireless patches attached to the anterior and posterior sides of the patient's chest. At this time, the location where the patches are attached and the number of patches may vary depending on the judgment of the medical staff.

[0072] The breath sound intensity image processing unit (120) can convert a continuous form of breath sound signal measured from the breath sound measurement system (20) into a breath sound intensity image. For example, the breath sound intensity image processing unit (120) can obtain breath sound intensity using at least one of a peak value, valid value, minimum-maximum average, peak vector, rise-fall time, amplitude obtained in the time domain of the breath sound signal, or an amplitude, phase, and period obtained in the frequency domain.

[0073] At this time, the breath sound intensity image processing unit (120) can convert the breath sound signal in the time domain into a time-frequency domain spectrogram through a Fourier transform to obtain the breath sound intensity in the frequency domain. In addition, the breath sound intensity image processing unit (120) can convert the spectrogram into a Mel-spectrogram by applying a Mel-filter Back to the spectrogram.

[0074] The breath sound intensity image processing unit (120) can obtain breath sound intensity by analyzing breath sound signals measured from a plurality of patches and can express the obtained breath sound intensity in the form of a two-dimensional distribution by measurement location. The breath sound intensity image processing unit (120) can generate a single breath sound intensity image for a specific point in time by blending the breath sound intensity by measurement location into a single image.

[0075] Afterwards, the breath sound intensity image processing unit (120) can repeatedly generate a single breath sound intensity image for each measurement time using the method described above, and finally generate multiple breath sound intensity images in a time series.

[0076] Referring again to FIG. 2, the image processing unit (100) can transmit the EIT image and the breath sound intensity image obtained from the chest of the same patient to the synchronization unit (300). At this time, each image may be data obtained from the patient's chest in real time, or data received from a database that stores each image.

[0077] The respiratory cycle segmentation unit (200) can receive a patient's respiratory signal, and the received respiratory signal may be continuous data appearing in the form of a periodic waveform. Accordingly, the respiratory cycle segmentation unit (200) can perform segmentation on the respiratory signal to obtain multiple respiratory segment information for synchronizing the EIT image and the respiratory sound intensity image. The respiratory cycle segmentation unit (200) can map the EIT image and the respiratory sound intensity image corresponding to each respiratory segment to perform synchronization for each segment.

[0078] The breathing cycle segmentation unit (200) can segment the breathing signal according to a pre-set threshold condition. For example, the breathing cycle segmentation unit (200) can extract segments corresponding to a pre-set amplitude range and phase range from the entire breathing signal.

[0079] According to another embodiment, the breathing cycle segmentation unit (200) can determine an abnormal breathing signal and extract the corresponding segment. For example, the breathing cycle segmentation unit (200) can determine signs of abnormal breathing by analyzing the breathing cycle, inhalation time fraction, and breathing depth from the breathing signal, and can segment the breathing signal based on the results of each breathing signal analysis.

[0080] When the breathing cycle segmentation unit (200) analyzes a breathing signal based on the breathing cycle, the breathing cycle segmentation unit (200) can calculate the breathing cycle based on the time interval between consecutive maximum amplitude points. Additionally, the breathing cycle segmentation unit (200) can determine a segment satisfying a pre-set breathing cycle condition as an abnormal breathing signal and extract it as a breathing segment.

[0081] When the respiratory cycle segmentation unit (200) analyzes a respiratory signal based on the inhalation time fraction, the respiratory cycle segmentation unit (200) can calculate the inhalation time based on the time interval between consecutive maximum amplitude points and minimum amplitude points, and thereby calculate the ratio of inhalation time to the respiratory cycle. Additionally, the respiratory cycle segmentation unit (200) can determine a segment satisfying a pre-set inhalation time fraction condition as an abnormal respiratory signal and extract it as a respiratory segment.

[0082] When the breathing cycle segmentation unit (200) analyzes a breathing signal based on breathing depth, the breathing cycle segmentation unit (200) can calculate the breathing depth based on the difference between the maximum amplitude value and the minimum amplitude value. Additionally, the breathing cycle segmentation unit (200) can determine a segment satisfying a pre-set breathing depth condition as an abnormal breathing signal and extract it as a breathing segment.

[0083] FIG. 4 is a drawing illustrating a breathing section according to one embodiment.

[0084] Referring to FIG. 4, the breathing cycle segmentation unit (200) can extract segments corresponding to a preset amplitude range and phase range from the entire breathing signal.

[0085] For example, the breathing cycle segmentation unit (200) can extract segments having a positive slope value to extract inhalation segments from the entire breathing signal. Additionally, the breathing cycle segmentation unit (200) can extract segments corresponding to a specific amplitude range.

[0086] The breathing cycle segmentation unit (200) may extract a segment that satisfies both the amplitude range and the slope range. For example, the breathing cycle segmentation unit (200) may extract a segment corresponding to a specific amplitude range from a segment having a negative slope value.

[0087] Referring again to FIG. 2, the breathing cycle segmentation unit (200) can transmit breathing segment information obtained for the entire breathing signal to the synchronization unit (300).

[0088] The synchronization unit (300) can extract a plurality of EIT images and a plurality of breath sound intensity images corresponding to each breath segment based on breath segment information.

[0089] Afterward, the synchronization unit (300) can sort the images within each breathing interval according to a specific criterion. For example, the synchronization unit (300) can sort the images within each breathing interval in chronological order.

[0090] Through this process, the synchronization unit (300) can perform synchronization between the EIT image and the breath sound intensity image for each breathing interval.

[0091] FIG. 5 is a diagram illustrating synchronized EIT images and breath sound intensity images according to one embodiment.

[0092] Referring to FIG. 5, the breathing cycle segmentation unit (200) can extract a segment having a positive slope value to extract only the inhalation segment from the entire breathing signal.

[0093] Afterwards, the extracted breathing interval information can be transmitted to the synchronization unit (300), and the synchronization unit (300) can perform synchronization by extracting and aligning multiple EIT images and multiple breathing sound intensity images corresponding to each breathing interval.

[0094] Referring again to FIG. 2, the respiratory abnormality prediction unit (400) can receive synchronized EIT images and respiratory sound intensity images for each respiratory interval from the synchronization unit (300), and can input the images into a respiratory abnormality prediction model (430) trained to predict respiratory abnormalities including lung disease.

[0095] The respiratory abnormality prediction model (430) can extract features related to respiratory abnormalities from the input EIT image and respiratory sound intensity image, respectively, and predict respiratory abnormalities based on the features. For example, the respiratory abnormality prediction model (430) can predict respiratory abnormalities such as atelectasis, pneumonia, acute respiratory distress syndrome (ARDS), pulmonary fibrosis, and emphysema.

[0096] The respiratory abnormality prediction unit (400) can provide information on which part of the input EIT image and respiratory sound intensity image the corresponding respiratory abnormality prediction was based on. For example, the respiratory abnormality prediction unit (400) can provide a visual representation by overlaying the area related to the respiratory abnormality prediction result in the form of a heat map on the EIT image and respiratory sound intensity image.

[0097] The information providing unit (500) can transmit the results of the prediction of abnormal breathing signs and the synchronized EIT video and breathing sound intensity video to an external terminal (20), and the external terminal (20) can provide the received data through a user interface screen. For example, the information providing unit (500) can simultaneously provide the synchronized EIT video and breathing sound intensity video by breathing sound interval through a split screen.

[0098] Meanwhile, the information providing unit (500) may provide biosignals received from an external monitoring device, such as a ventilator, in conjunction with the device. For example, the information providing unit (500) may provide additional biosignals including tidal volume (Vt), pulse rate (PR), inhaled oxygen concentration (FiO2), positive end expiratory pressure (PEEP), transcutaneous oxygen saturation (SpO2), and arterial oxygen partial pressure (PaO2). In addition, the information providing unit (500) may receive and provide information on biosignals that have been pre-set by medical staff.

[0099] That is, based on lung management information including synchronized EIT images and breath sound intensity images, predicted results of respiratory abnormalities, and biosignals received from the information provider (500), the medical staff can perform lung protective ventilation (LPV) strategies to minimize the occurrence of ventilator-induced lung injury (VILI).

[0100] FIG. 6 is a diagram illustrating a respiratory abnormality sign prediction unit according to one embodiment.

[0101] FIG. 7 is a diagram illustrating a method for generating a training dataset according to one embodiment.

[0102] Referring to FIG. 6, the respiratory abnormality prediction unit (400) may include a training data generation unit (410) and a learning unit (420) that trains a respiratory abnormality prediction model (430) using the generated training data. At this time, the training data generation unit (410) and the learning unit (420) may not necessarily be included in the respiratory abnormality prediction unit (400), but for the convenience of explanation, the description assumes that the training data generation unit (410) and the learning unit (420) are included in the respiratory abnormality prediction unit (300).

[0103] Referring to FIG. 7, the learning data generation unit (410) can generate first to third learning datasets by combining EIT images and breath sound intensity images in various forms.

[0104] The training data generation unit (410) can generate a first training dataset in the form of concatenating at least one EIT image and at least one breath sound intensity image.

[0105] The training data generation unit (410) can generate a domain transform image for at least one EIT image or at least one breath sound intensity image. For example, the training data generation unit (410) can generate a domain transform image using a known method such as a wavelet transform. The training data generation unit (410) can also generate a second training dataset by combining at least one EIT image, a breath sound intensity image, and domain transform images.

[0106] Additionally, the learning data generation unit (410) can generate a third learning dataset by combining a first learning dataset or a second learning dataset with a specific biosignal received from a monitoring system.

[0107] Referring again to FIG. 6, the learning unit (420) can train a respiratory abnormality prediction model (430) using various types of learning datasets generated by the learning data generation unit (410).

[0108] The respiratory abnormality prediction model (430) can learn the relationship between the input video and the respiratory abnormality through a training dataset. That is, the respiratory abnormality prediction model (430) may be an artificial intelligence model trained to predict respiratory abnormalities from the input videos. At this time, a specific biosignal received from a monitoring system may also be input to the respiratory abnormality prediction model (430).

[0109] Finally, the respiratory abnormality prediction model (430) can predict respiratory abnormalities such as atelectasis, pneumonia, acute respiratory distress syndrome (ARDS), and pulmonary fibrosis based on the input EIT image and breath sound intensity image.

[0110] The respiratory abnormality prediction model (430) can visually represent in the EIT image or respiratory sound intensity image which part of the input image was used to predict the respiratory abnormality. That is, the respiratory abnormality prediction model (430) can be Explainable Artificial Intelligence (XAI), and methods such as Class Activating Map (CAM) can be applied for this purpose.

[0111] FIG. 8 is a flowchart of a lung monitoring method according to one embodiment.

[0112] Referring to FIG. 8, the lung monitoring system (10) generates a breath sound intensity image based on a breath sound signal and segments the breath signal into multiple breath cycles according to a preset condition (S110). For example, the lung monitoring system (10) can extract segments corresponding to a preset amplitude range and slope range from the entire breath signal.

[0113] The lung monitoring system (10) extracts an EIT image and a breath sound intensity image corresponding to each breathing interval and performs synchronization between the EIT image and the breath sound intensity image for each breathing interval (S120). For example, the lung monitoring system (10) can perform synchronization by arranging the images extracted within each breathing interval in chronological order.

[0114] The lung monitoring system (10) inputs synchronized EIT images and breath sound intensity images into a trained respiratory abnormality prediction model to obtain a respiratory abnormality prediction result and provides lung management information to medical staff (S130). At this time, the respiratory abnormality prediction model (430) may be an artificial intelligence model trained to predict respiratory abnormalities from synchronized EIT images and breath sound intensity images. Additionally, the lung management information may include synchronized EIT images and breath sound intensity images, a respiratory abnormality prediction result, and a biosignal.

[0115] FIG. 9 is a hardware configuration diagram of a lung monitoring system according to one embodiment.

[0116] Referring to FIG. 9, the lung monitoring system (10) is a computing device operated by one or more processors (11), and may include a processor (11), a memory (13) for loading a computer program executed by the processor (11), a storage (15) for storing the computer program and various data, a communication interface (17), and a bus (19) connecting them. In addition, the lung monitoring system (10) may include various additional components.

[0117] A computer program may include instructions that cause a processor (11) to perform a method / operation according to various embodiments of the present disclosure when loaded into memory (13). That is, the processor (11) may perform a method / operation according to various embodiments of the present disclosure by executing the instructions. A computer program is composed of a series of computer-readable instructions grouped by function and refers to being executed by a processor. A computer program may include instructions that perform the operation described in the present disclosure.

[0118] The processor (11) controls the overall operation of each component of the lung monitoring system (10). The processor (11) may be configured to include at least one of a CPU (Central Processing Unit), MPU (Micro Processor Unit), MCU (Micro Controller Unit), GPU (Graphic Processing Unit), or any form of processor well known in the art of the present disclosure. Additionally, the processor (11) may perform operations for at least one application or computer program for executing a method / operation according to various embodiments of the present disclosure.

[0119] Memory (13) stores various data, commands and / or information. Memory (13) may load one or more computer programs from storage (15) to execute a method / operation according to various embodiments of the present disclosure. Memory (13) may be implemented as volatile memory such as RAM, but the technical scope of the present disclosure is not limited thereto.

[0120] Storage (15) can store computer programs non-temporarily. Storage (15) may be configured to include non-volatile memory such as ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), flash memory, a hard disk, a removable disk, or any form of computer-readable recording medium well known in the art to which this disclosure belongs.

[0121] The communication interface (17) supports wired and wireless internet communication of the waste monitoring system (10). Additionally, the communication interface (17) may support various communication methods other than internet communication. To this end, the communication interface (17) may be configured to include a communication module well known in the art of the present disclosure.

[0122] The bus (19) provides communication functions between components of the lung monitoring system (10). The bus (19) can be implemented as various types of buses, such as an address bus, a data bus, and a control bus.

[0123] The embodiments of the present disclosure described above are not implemented only through devices and methods, but may also be implemented through a program that realizes a function corresponding to the configuration of the embodiments of the present disclosure or a recording medium on which such program is recorded.

[0124] Although embodiments of the present disclosure have been described in detail above, the scope of the present disclosure is not limited thereto, and various modifications and improvements by those skilled in the art using the basic concepts of the present disclosure as defined in the following claims also fall within the scope of the present disclosure.

Claims

1. A method of operation of a lung monitoring system operated by at least one processor, wherein A step of acquiring a plurality of Electrical Impedance Tomography (EIT) images of the patient's chest and a plurality of respiratory sound intensity images generated based on respiratory sound signals acquired from the patient's chest, A step of mapping an EIT image and a breath sound intensity image corresponding to each breath segment based on multiple breath segment information obtained by binning the patient's breath signal, and The method comprises the step of inputting at least one pair of EIT images and breath sound intensity images included in each of the above breathing intervals into a respiratory abnormality prediction model to obtain a result of predicting respiratory abnormalities of the patient. Method of operation.

2. In Paragraph 1, The step of acquiring the above plurality of breath sound intensity images is A step of obtaining breath sound intensity by measurement location from a multi-channel breath sound signal measured from at least one chest region of the above patient, A step of generating a breath sound intensity image for a specific point in time by blending images showing the distribution of breath sound intensity by measurement location, and A step comprising repeatedly generating a breath sound intensity image for the above specific point in time over a set period of time to generate the plurality of breath sound intensity images. Method of operation.

3. In Paragraph 2, The above breath sound intensity is A value corresponding to at least one of the peak value, effective value, minimum-maximum average, peak vector, rise-fall time, and amplitude for the signal acquired in the time domain of the multichannel breath sound signal, or A value corresponding to at least one of the amplitude, phase, and period of the signal obtained in the frequency domain of the multichannel breath sound signal, Method of operation.

4. In Paragraph 1, The above mapping step A method comprising the step of extracting a plurality of breathing segments corresponding to a preset amplitude range or phase range from the above breathing signal, Method of operation.

5. In Paragraph 1, The above mapping step The method further includes the step of extracting a plurality of abnormal breathing segments from the breathing signal by analyzing at least one of the breathing cycle, inspiratory time fraction, and breathing depth of the breathing signal. Method of operation.

6. In Paragraph 1, Prior to the step of obtaining the predicted results of the above-mentioned respiratory abnormal signs The step of generating a training dataset for training the respiratory abnormality prediction model using at least one pair of EIT images and breath sound intensity images, and A method comprising the step of training the respiratory abnormality prediction model using the above training dataset, Method of operation.

7. In Paragraph 6, The above training dataset is A first training dataset generated by concatenating at least one EIT image and at least one breath sound intensity image, A second training dataset generated by performing a domain transform on at least one EIT image or at least one breath sound intensity image and combining at least one EIT image or at least one breath sound intensity image with the domain-transformed image, and A third learning dataset comprising at least one of the first learning dataset or the second learning dataset generated by combining the patient's biosignal with the first learning dataset or the second learning dataset. Method of operation.

8. In Paragraph 1, After the step of obtaining the predicted results of the above-mentioned respiratory abnormal signs The method further includes the step of providing at least one pair of EIT images and breath sound intensity images included in each of the above breathing intervals, and the above breathing abnormality sign prediction results to an external terminal. The steps provided above A step comprising providing a region related to the predicted respiratory abnormality sign result in the EIT image or respiratory sound intensity image, distinguished according to importance. Method of operation.

9. In Paragraph 8, The steps provided above A method further comprising the step of receiving and providing information on biosignals pre-set by the user from a ventilator linked to the patient. Method of operation.

10. A lung monitoring system operated by at least one processor, An image processing unit that acquires a plurality of Electrical Impedance Tomography (EIT) images of a patient's chest and a plurality of respiratory sound intensity images generated based on a respiratory sound signal acquired from the patient's chest, A synchronization unit that maps an EIT image and a breath sound intensity image corresponding to each breath segment based on multiple breath segment information obtained by binning the patient's breath signal, and A respiratory abnormality prediction unit comprising at least one pair of EIT images and breath sound intensity images included in each of the above respiratory intervals, inputting them into a respiratory abnormality prediction model to obtain a predicted result of the patient's respiratory abnormality. Lung monitoring system.

11. In Paragraph 10, The above image processing unit Acquiring breath sound intensity by measurement location from a multi-channel breath sound signal measured from at least one chest region of the patient, blending images representing the distribution of breath sound intensity by measurement location to generate a breath sound intensity image for a specific point in time, and repeatedly generating the breath sound intensity image for the specific point in time over a set period of time to generate the plurality of breath sound intensity images. Lung monitoring system.

12. In Paragraph 11, The above breath sound intensity is A value corresponding to at least one of the peak value, effective value, minimum-maximum average, peak vector, rise-fall time, and amplitude for the signal acquired in the time domain of the multichannel breath sound signal, or A value corresponding to at least one of the amplitude, phase, and period of the signal obtained in the frequency domain of the multichannel breath sound signal, Lung monitoring system.

13. In Paragraph 10, The above system A breathing cycle segmentation unit comprising extracting a plurality of breathing segments corresponding to a preset amplitude range or phase range from the above breathing signal, Lung monitoring system.

14. In Paragraph 13, The above breathing cycle segmentation part Analyzing at least one of the breathing cycle, inspiratory time fraction, and breathing depth of the above breathing signal to further extract a plurality of abnormal breathing segments from the above breathing signal, Lung monitoring system.

15. In Paragraph 10, The above respiratory abnormality prediction unit A training data generation unit that generates a training dataset for training the respiratory abnormality prediction model using at least one pair of EIT images and respiratory sound intensity images, and A learning unit comprising a learning unit that trains the respiratory abnormality prediction model using the above-mentioned learning dataset, Lung monitoring system.

16. In Paragraph 15, The above training dataset is A first training dataset generated by concatenating at least one EIT image and at least one breath sound intensity image, A second training dataset generated by performing a domain transform on at least one EIT image or at least one breath sound intensity image and combining at least one EIT image or at least one breath sound intensity image with the domain-transformed image, and A third learning dataset comprising at least one of the first learning dataset or the second learning dataset generated by combining the patient's biosignal with the first learning dataset or the second learning dataset. Lung monitoring system.

17. In Paragraph 10, It further includes at least one pair of EIT images and breath sound intensity images included in each of the above breathing intervals, and an information providing unit that provides the above-mentioned respiratory abnormality sign prediction results to an external terminal, and The above information providing department Providing areas related to the predicted respiratory abnormality signs in the EIT image or respiratory sound intensity image, distinguished by importance. Lung monitoring system.

18. In Paragraph 17, The above information providing department Receiving and providing information on vital signs pre-set by the user from a ventilator linked to the patient, Lung monitoring system.

Images