A method and system for determining emergency level based on deep learning
By building an emergency level determination system based on deep learning, utilizing patient historical data and imaging data, and combining it with image separation technology, we have solved the problems of comprehensive inspection and rapid classification in the emergency department, and achieved efficient emergency level determination and treatment.
Patent Information
- Application Number
- CN202411249896.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-06
- Publication Date
- 2025-09-30
- Estimated Expiration
- 2044-09-06
AI Technical Summary
Emergency departments lack comprehensive examinations, rescue times are long, and rescue efficiency is low. Machine learning models fail to adapt to the characteristics of medical data when processing medical data, resulting in large fluctuations in results and an inability to quickly classify patients' emergency levels and provide targeted treatment.
By acquiring patients' historical diagnostic data, weak electrophysiological parameters, and medical imaging data, a deep learning-based emergency level determination model is constructed. CNN or RNN convolutional neural network models are used to fuse and process emergency grading features. Image separation technology is combined to separate lesion areas to assist doctors in emergency treatment.
It reduces rescue time, improves rescue efficiency, accurately classifies patients' emergency levels, reduces fluctuations in model processing results, and improves the utilization level of machine learning models.
Smart Images
Figure CN119132568B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of medical emergency technology, and in particular relates to a method and system for determining emergency levels based on deep learning. Background Art
[0002] The emergency department is a hub for critically ill patients. Due to the wide variety of conditions and the tight timelines for emergency care, it carries the heaviest workload and is the gateway for all emergency patients admitted to the hospital. Emergency department diagnosis presents three key characteristics: First, time is of the essence. For cases such as severe burns, sudden heart attacks, cerebral hemorrhages, or brain surgical injuries, failure to provide timely, targeted emergency care can lead to missed opportunities for treatment, resulting in death or lifelong disability. Second, the sheer volume of patients is a significant burden. Due to limited testing resources in the emergency department—primarily limited equipment and staff—each patient requires unique testing equipment. For example, dyspnea could be caused by the novel coronavirus, advanced lung cancer, or a surgical injury. Failure to distinguish the underlying cause can lead to misdiagnosis and delays in treatment. Therefore, providing emergency department physicians with timely and accurate diagnostic and treatment data is a crucial component of emergency care and crucial to the success of saving lives.
[0003] In addition, in the existing technology, when witnesses or parties find that a patient has fallen into shock or has lost blood and is unconscious, they do not know the cause of the illness. At this time, the emergency department receives a call and only prepares a few relevant medicines. As a result, it is not uncommon for patients to suddenly die due to lack of a specific medicine while getting the medicine.
[0004] It can be seen that the existing emergency department patient monitoring and emergency care system is still very backward. The emergency department has not fully coordinated to retrieve and utilize the patient's previous diagnostic data. Due to various reasons, the rescue time is prolonged, and the rescue time must be every second. Therefore, the existing technology still has problems such as the inability to conduct comprehensive inspections, too long rescue times, low rescue efficiency, and inability to assist doctors in quickly classifying patients' emergency levels and conducting targeted treatments for corresponding symptoms. At the same time, the processing methods for patient examination films are relatively simple. For the black background areas and target areas that appear in medical images, or even the target areas and severe lesion areas, the lesion areas cannot be well separated from the target areas and black background areas. In addition, the use of machine learning and deep learning models to process medical data is already very mature, but when using neural network models, the characteristics of medical data cannot be taken into account and the model parameters are not adaptively adjusted, resulting in large fluctuations in the model's processing results of medical data. Such problems have always plagued the emergency department's use of machine learning models. Summary of the Invention
[0005] To solve the above technical problems, the present invention proposes a method and system for determining emergency level based on deep learning.
[0006] In a first aspect of the present invention, a method for determining emergency level based on deep learning is provided, wherein the method comprises:
[0007] S1. Obtain the patient's historical diagnostic data, first weak electrophysiological parameter, and first medical imaging data, and obtain the first emergency condition severity level set by the corresponding doctor at this time;
[0008] S2. Fusing the historical diagnostic data, the first weak electrophysiological parameter, and the first medical imaging data to obtain a first emergency classification feature;
[0009] S3. Constructing an emergency level determination model based on the first emergency classification characteristics and the first emergency condition severity level;
[0010] S4. Receive real-time diagnostic data, a second weak electrophysiological parameter, and second medical imaging data of the target to be treated for emergency, fuse the real-time diagnostic data, the second weak electrophysiological parameter, and the second medical imaging data to obtain a second emergency grading feature, use the emergency level determination model to process the second emergency grading feature to generate a second emergency condition severity level, and assist doctors in performing emergency treatment based on the second emergency condition severity level.
[0011] Furthermore, the historical diagnostic data is the patient's historical hospitalization diagnostic data, including blood pressure, heart rate, respiratory rate, and special diagnostic parameters of each examination during hospitalization; the special diagnostic parameters are determined by the patient's characteristic symptoms, and the historical diagnostic data constitutes parameter N;
[0012] The first weak electrophysiological parameter or the second weak electrophysiological parameter is a cardiac electrophysiological parameter, a cerebral electrophysiological parameter, or a myoelectrical electrophysiological parameter.
[0013] Furthermore, the first medical image data or the second medical image data is acquired using an X-ray, a CT scan, or an MRI.
[0014] Furthermore, the emergency level determination model adopts a deep learning model, and the deep learning model is a CNN convolutional neural network model or a RNN convolutional neural network model.
[0015] Furthermore, before fusing to obtain the first emergency grading feature or the second emergency grading feature, the separated first lesion image data or the second lesion image data is subjected to computational compression to obtain a first lesion compression vector or a second lesion compression vector P(M) i .
[0016] Furthermore, the real vector P(M) i and blood pressure, heart rate, respiratory rate and special diagnostic parameter characteristics N, the EEG parameter B f The first emergency classification feature or the second emergency classification feature is obtained by fusion.
[0017] A deep learning-based emergency level determination system is also provided, comprising an emergency data acquisition terminal and an emergency level determination terminal, characterized in that:
[0018] The emergency data acquisition terminal includes an emergency data collection module, an electrophysiological parameter detection module, and a medical image acquisition module. The emergency level determination terminal includes a hospital main control module, an emergency level determination model construction module, and an emergency level determination module, wherein:
[0019] The emergency data collection module is connected to the hospital main control module to obtain the patient's historical diagnostic data stored therein or the real-time diagnostic data of the emergency target, and is connected to the electrophysiological parameter detection module to obtain the first weak electrophysiological parameter or the second weak electrophysiological parameter, and is connected to the medical image acquisition module to obtain the first medical image data or the second medical image data;
[0020] The electrophysiological parameter detection module is used to collect the first weak electrophysiological parameter or the second weak electrophysiological parameter of the patient and is connected to the emergency data collection module;
[0021] The medical image acquisition module is configured to acquire the first medical image data or the second medical image data using an X-ray film, a CT film, or an MRI;
[0022] The hospital main control module is used to store the first emergency condition severity level corresponding to the patient's historical diagnosis data;
[0023] The emergency level determination model construction module is configured to fuse the historical diagnostic data, the first weak electrophysiological parameter, and the first medical imaging data to obtain a first emergency level classification feature, and to construct an emergency level determination model based on the first emergency level classification feature and the first emergency condition severity level;
[0024] The emergency level determination module is connected to the hospital main control module and assists doctors in emergency treatment based on the severity level of the second emergency condition.
[0025] The present invention reduces rescue time and improves rescue efficiency by utilizing comprehensive inspection data. The emergency level classification obtained through the model assists doctors in quickly classifying patients into emergency levels and conducting targeted treatments for corresponding symptoms. At the same time, for the films taken during patient examinations, the black background areas and target areas, and even the target areas and serious lesion areas that appear in the medical images, the image separation technology is used to separate the lesion areas from the target areas and the black background areas. In addition, when utilizing the neural network model, the model parameters are adaptively adjusted taking into account the characteristics of medical data at different orders of magnitude, so that the fluctuation of the model's processing results for medical data is greatly reduced, thereby improving the level of utilization of machine learning models by emergency departments.
[0026] More embodiments and improved effects of the present invention will be further introduced in conjunction with the drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 is a flow chart of the method for determining emergency level based on deep learning of the present invention;
[0028] Figure 2 2. It is a schematic diagram of an emergency level determination terminal in the emergency level determination system based on deep learning of the present invention;
[0029] Figure 3 2. It is a schematic diagram of an emergency data acquisition terminal in the emergency level determination system based on deep learning of the present invention;
[0030] Figure 4 Schematic diagram of image separation required for obtaining features of the lesion area in the present invention;
[0031] Figure 5 This is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0032] The invention is further described below with reference to the accompanying drawings and specific implementation methods.
[0033] To solve the above technical problems, the present invention proposes a method and system for determining emergency level based on deep learning.
[0034] In a first aspect of the present invention, a method for determining emergency level based on deep learning is provided, wherein the method comprises:
[0035] S1. Obtain the patient's historical diagnostic data, first weak electrophysiological parameter, and first medical imaging data, and obtain the first emergency condition severity level set by the corresponding doctor at this time;
[0036] S2. Fusing the historical diagnostic data, the first weak electrophysiological parameter, and the first medical imaging data to obtain a first emergency classification feature;
[0037] S3. Constructing an emergency level determination model based on the first emergency classification characteristics and the first emergency condition severity level;
[0038] S4. Receive real-time diagnostic data, a second weak electrophysiological parameter, and second medical imaging data of the target to be treated for emergency, fuse the real-time diagnostic data, the second weak electrophysiological parameter, and the second medical imaging data to obtain a second emergency grading feature, use the emergency level determination model to process the second emergency grading feature to generate a second emergency condition severity level, and assist doctors in performing emergency treatment based on the second emergency condition severity level.
[0039] Furthermore, the historical diagnostic data is the patient's historical hospitalization diagnostic data, including blood pressure, heart rate, respiratory rate, and special diagnostic parameters of each examination during hospitalization; the special diagnostic parameters are determined by the patient's characteristic symptoms, and the historical diagnostic data constitutes parameter N;
[0040] In this embodiment, if the patient has chronic lung problems, the special diagnostic parameters are physiological parameters related to the lungs, including total white blood cell count, white blood cell classification, total red blood cell count, hemoglobin, C-reactive protein, platelets, etc. In this embodiment, any one of them can be selected as a special diagnostic parameter.
[0041] The first weak electrophysiological parameter or the second weak electrophysiological parameter is a cardiac electrophysiological parameter, a cerebral electrophysiological parameter, or a myoelectrical electrophysiological parameter.
[0042] In this embodiment, the electroencephalographic physiological parameters are collected using a wireless electroencephalographic acquisition device, the g-tec device.
[0043] Furthermore, the first medical image data or the second medical image data is acquired using an X-ray, a CT scan, or an MRI.
[0044] Furthermore, the emergency level determination model adopts a deep learning model, and the deep learning model is a CNN convolutional neural network model or a RNN convolutional neural network model.
[0045] Furthermore, the calculation formula of the electroencephalographic physiological parameters is as follows:
[0046]
[0047] Where B fis the electroencephalographic physiological parameter, k is the electroencephalographic characteristic coefficient, the value range is 0.05-0.10, T is the time period for collecting the patient's electroencephalogram, n is the number of electroencephalographic points collected, S1, S2, ..., S n It is the EEG signal amplitude measured at the corresponding EEG point within a period of time, in μV.
[0048] In this embodiment, the value of n is the number of points set for collection on the wireless EEG collection device, with an upper limit of 64.
[0049] Furthermore, before using the emergency level determination model, the collected images need to be processed. The image processing includes image separation. The lesion samples are separated from the background by binarization. The improved Os algorithm is used. The inter-class variance S between the lesion and the background is defined as:
[0050]
[0051] Where a is the total average grayscale of the image, n is the image sampling rate, T n is the gray value of the sampled lesion image, M n To sample the grayscale values of the lesion-related image, the grayscale values are acquired based on the sampling rate, p0 is the ratio of the number of pixels belonging to the target to the entire image, p1 is the ratio of the number of pixels belonging to the background to the entire image, a0 is the average grayscale of the target, a0 is the average grayscale of the background, and the first medical image data or the second medical image data is processed to obtain the first lesion image data or the second lesion image data.
[0052] The features obtained after image separation in this embodiment are Figure 4 Characteristics of some areas of the disease are shown in .
[0053] Furthermore, before fusing to obtain the first emergency grading feature or the second emergency grading feature, the separated first lesion image data or the second lesion image data is subjected to computational compression to obtain a first lesion compression vector or a second lesion compression vector P(M) i , and its calculation formula is:
[0054]
[0055] Wherein, the first lesion image data or the second lesion image data M after separation is i Compress to another real vector P(M) i In the example, the eigenvalue range is between -1 and 1, and the sum of all eigenvalues is 0, K represents the Kth dimension, T i Represents the image feature of the i-th lesion.
[0056] Furthermore, the real vector P(M) iand blood pressure, heart rate, respiratory rate and special diagnostic parameter characteristics N, the EEG parameter B f The first emergency classification feature or the second emergency classification feature is obtained by fusion.
[0057] In this embodiment, the first emergency classification feature or the second emergency classification feature obtained by fusion is (P(M) i , N, B f ).
[0058] Furthermore, the activation function calculation formula of the improved CNN convolutional neural network model is as follows:
[0059]
[0060] X is the first emergency classification feature or the second emergency classification feature, H(S) is the output activation function value, f , S, P(M) i > indicates taking EEG parameter B f , parameter characteristics N and real vector P(M) i The minimum order of magnitude in the fusion feature, |B f , N, P(M) i | indicates the EEG parameter B f , parameter characteristics N and real vector P(M) i The maximum magnitude among fused features.
[0061] The CNN convolutional neural network model, using an improved activation function, concluded that emergency levels can be classified into four categories: Level A (critical), Level B (critical), Level C (emergency), and Level D (non-emergency). The CNN convolutional neural network model's method for classifying these four emergency levels is similar to that used in CNN models in the field and will not be repeated here.
[0062] A deep learning-based emergency level determination system is also provided, comprising an emergency data acquisition terminal and an emergency level determination terminal, characterized in that:
[0063] The emergency data acquisition terminal includes an emergency data collection module, an electrophysiological parameter detection module, and a medical image acquisition module. The emergency level determination terminal includes a hospital main control module, an emergency level determination model construction module, and an emergency level determination module, wherein:
[0064] The emergency data collection module is connected to the hospital main control module to obtain the patient's historical diagnostic data stored therein or the real-time diagnostic data of the emergency target, and is connected to the electrophysiological parameter detection module to obtain the first weak electrophysiological parameter or the second weak electrophysiological parameter, and is connected to the medical image acquisition module to obtain the first medical image data or the second medical image data;
[0065] The electrophysiological parameter detection module is used to collect the first weak electrophysiological parameter or the second weak electrophysiological parameter of the patient and is connected to the emergency data collection module;
[0066] The medical image acquisition module is configured to acquire the first medical image data or the second medical image data using an X-ray film, a CT film, or an MRI;
[0067] The hospital main control module is used to store the first emergency condition severity level corresponding to the patient's historical diagnosis data;
[0068] The emergency level determination model construction module is configured to fuse the historical diagnostic data, the first weak electrophysiological parameter, and the first medical imaging data to obtain a first emergency level classification feature, and to construct an emergency level determination model based on the first emergency level classification feature and the first emergency condition severity level;
[0069] The emergency level determination module is connected to the hospital main control module and assists doctors in emergency treatment based on the severity level of the second emergency condition.
[0070] The present invention reduces rescue time and improves rescue efficiency by utilizing comprehensive inspection data. The emergency level classification obtained through the model assists doctors in quickly classifying patients into emergency levels and conducting targeted treatments for corresponding symptoms. At the same time, for the films taken during patient examinations, the black background areas and target areas, and even the target areas and serious lesion areas that appear in the medical images, the image separation technology is used to separate the lesion areas from the target areas and the black background areas. In addition, when utilizing the neural network model, the model parameters are adaptively adjusted taking into account the characteristics of medical data at different orders of magnitude, so that the fluctuation of the model's processing results for medical data is greatly reduced, thereby improving the level of utilization of machine learning models by emergency departments.
[0071] The control device may include a processor 401 and a memory 402 storing computer program instructions. Specifically, the processor 401 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the present invention.
[0072] Memory 402 may include a large capacity memory for data or instructions. By way of example and not limitation, memory 402 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 402 may include removable or non-removable (or fixed) media. Where appropriate, memory 402 may be inside or outside the data processing device. In a specific embodiment, memory 402 is a non-volatile solid-state memory. In a specific embodiment, memory 402 includes a read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or a flash memory, or a combination of two or more of these.
[0073] The processor 401 reads and executes computer program instructions stored in the memory 402 to implement any one of the methods for analyzing storage tank detection data in the above embodiments.
[0074] In one example, the electronic device may further include a communication interface 403 and a bus 410. Figure 3 As shown, the processor 401, memory 402, and communication interface 403 are connected and communicate with each other via a bus 410. The communication interface 403 is mainly used to implement communication between various modules, devices, units, and / or equipment in the embodiment of the present invention.
[0075] Bus 410 comprises hardware, software or both, couples the parts of electronic equipment to each other.For example, and not limitation, bus can comprise accelerated graphics port (AGP) or other graphics bus, enhanced industry standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industry standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations.In suitable cases, bus 410 can comprise one or more buses.Although the embodiment of the present invention describes and shows specific bus, the present invention considers any suitable bus or interconnection.
[0076] Of course, it can be understood that each embodiment of the present invention can achieve one of the effects alone, and a combination of multiple embodiments of the present invention can achieve all of the above effects, but it is not required that every embodiment of the present invention achieve all of the above advantages and effects, because each embodiment of the present invention can constitute a separate technical solution and make one or more contributions to the existing technology.
[0077] For any module structure not specifically defined in this invention, the prior art shall prevail. The prior art mentioned in the aforementioned background and specific embodiments of this invention may be considered as part of this invention and used to understand the meaning of certain technical features or parameters. The scope of protection of this invention shall be based on the actual content of the claims.
Claims
1. A method for determining emergency level based on deep learning, characterized in that: The method comprises: S1. Obtain the patient's historical diagnostic data, first weak electrophysiological parameter, and first medical imaging data, and obtain the first emergency condition severity level set by the corresponding doctor at this time; S2. Fusing the historical diagnostic data, the first weak electrophysiological parameter, and the first medical imaging data to obtain a first emergency classification feature; S3. Constructing an emergency level determination model based on the first emergency classification characteristics and the first emergency condition severity level; The emergency level determination model adopts a deep learning model, which is a CNN convolutional neural network model. The improved activation function calculation formula of the CNN convolutional neural network model is as follows: X is the emergency classification feature, H(X) is the output activation function value, Indicates EEG parameters , parameter characteristics N and real vector The minimum order of magnitude in the fusion feature, Indicates EEG parameters , parameter characteristics N and real vector The maximum magnitude of the fused features; S4. Receive real-time diagnostic data, a second weak electrophysiological parameter, and second medical imaging data of the patient to be treated for emergency treatment, fuse the real-time diagnostic data, the second weak electrophysiological parameter, and the second medical imaging data to obtain a second emergency classification feature, process the second emergency classification feature using the emergency level determination model to generate a second emergency condition severity level, and assist a physician in performing emergency treatment based on the second emergency condition severity level. Before using the emergency level determination model, the medical image data is processed; the improved Os algorithm is used, and the inter-class variance S between the lesion and the background is defined as: Where a is the total average grayscale of the image, n is the image sampling rate, is the gray value of the j-th sampled lesion image, is the grayscale value of the jth sampled lesion-related image, and the grayscale value is obtained based on the sampling rate. p0 is the ratio of the number of pixels belonging to the target to the entire image, p1 is the ratio of the number of pixels belonging to the background to the entire image, a0 is the average grayscale of the target, and a1 is the average grayscale of the background. The processed medical imaging data is used to construct the emergency level determination model.
2. The method for determining emergency level based on deep learning according to claim 1, characterized in that: The historical diagnostic data is the patient's historical hospitalization diagnostic data, including blood pressure, heart rate, respiratory rate, and special diagnostic parameters for each examination during hospitalization; the special diagnostic parameters are determined by the patient's characteristic symptoms, and the historical diagnostic data constitutes parameter feature N; Weak electrophysiological parameters are brain electrophysiological parameters; The calculation formula of the electroencephalographic physiological parameters is as follows: Where, is the electroencephalographic physiological parameter, k is the electroencephalographic characteristic coefficient, and its value range is 0.05-0.
10. is the time period for collecting the patient's EEG, m is the number of EEG points collected, is the EEG signal amplitude measured at the g-th EEG point over a period of time, in units of .
3. The method for determining emergency level based on deep learning according to claim 1, characterized in that: Before fusing the emergency classification features, the separated lesion image data is compressed to obtain a real vector .
4. A deep learning-based emergency level determination system, comprising an emergency data acquisition terminal and an emergency level determination terminal, characterized in that: The emergency data acquisition terminal includes an emergency data collection module, an electrophysiological parameter detection module, and a medical image acquisition module. The emergency level determination terminal includes a hospital main control module, an emergency level determination model construction module, and an emergency level determination module, wherein: The emergency data collection module is connected to the hospital main control module to obtain the patient's historical diagnostic data stored therein or the real-time diagnostic data of the emergency target, and is connected to the electrophysiological parameter detection module to obtain the first weak electrophysiological parameter or the second weak electrophysiological parameter, and is connected to the medical image acquisition module to obtain the first medical image data or the second medical image data; The electrophysiological parameter detection module is used to collect the first weak electrophysiological parameter or the second weak electrophysiological parameter of the patient and is connected to the emergency data collection module; The medical image acquisition module is configured to acquire the first medical image data or the second medical image data using an X-ray film, a CT film, or a nuclear magnetic resonance imaging (MRI); The hospital main control module is used to store the first emergency condition severity level corresponding to the patient's historical diagnosis data; The emergency level determination model construction module is configured to fuse the historical diagnostic data, the first weak electrophysiological parameter, and the first medical imaging data to obtain a first emergency level classification feature, and to construct an emergency level determination model based on the first emergency level classification feature and the first emergency condition severity level; The emergency level determination model adopts a deep learning model, which is a CNN convolutional neural network model. The improved activation function calculation formula of the CNN convolutional neural network model is as follows: X is the emergency classification feature, H(X) is the output activation function value, Indicates EEG parameters , parameter characteristics N and real vector The minimum order of magnitude in the fusion feature, Indicates EEG parameters , parameter characteristics N and real vector The maximum magnitude of the fused features; Before using the emergency level determination model, the medical image data is processed; the improved Os algorithm is used, and the inter-class variance S between the lesion and the background is defined as: Where a is the total average grayscale of the image, n is the image sampling rate, is the gray value of the j-th sampled lesion image, is the grayscale value of the jth sampled lesion-related image, and the grayscale value is obtained based on the sampling rate. p0 is the ratio of the number of pixels belonging to the target to the entire image, p1 is the ratio of the number of pixels belonging to the background to the entire image, a0 is the average grayscale of the target, and a1 is the average grayscale of the background. The processed medical imaging data is used to construct the emergency level determination model; The emergency level determination module is connected to the hospital main control module and assists doctors in emergency treatment based on the second emergency severity level obtained by the emergency level determination model.