Emergency critical value intelligent identification method and device, electronic equipment and storage medium

Abstract

The invention discloses an emergency critical value intelligent identification method and device, electronic equipment and a storage medium, and belongs to the technical field of image recognition, and the emergency critical value intelligent identification method comprises the steps: obtaining CT image data and clinical diagnosis guide data of an emergency patient; cT image data and clinical diagnosis guide data are input into the trained critical value recognition network, a critical value recognition result is obtained, and the architecture of the trained critical value recognition network comprises a visual branch network and a text branch network; and when the critical value identification result exceeds a threshold value, giving an alarm for reminding. According to the method, a large amount of manpower and material resource cost can be saved, and meanwhile the utilization efficiency of medical resources can be remarkably improved.

Description

Technical Field

[0001] This application belongs to the technical field of image recognition, and particularly relates to an intelligent recognition method, device, electronic device and storage medium for emergency critical values. Background Art

[0002] Emergency medicine is one of the most important links in the medical system, providing timely diagnosis and treatment for critically ill patients. Emergency patients usually have complex and rapidly changing conditions, and many patients are in a state of unstable vital signs. Any delay may lead to the deterioration of the condition or even death. In the emergency scenario, clinicians need to quickly process a large amount of patient diagnosis and treatment information, among which laboratory test results (such as blood, urine, biochemical indicators, etc.), vital sign monitoring data (such as electrocardiogram, blood oxygen saturation, etc.) and CT (computed tomography) image data are the key basis for judging the patient's condition. However, some abnormal test results (i.e., "critical values") may directly indicate that the patient is in a life-threatening or pathological state that requires urgent intervention.

[0003] In the emergency environment, CT (computed tomography) has become the preferred examination method for most emergency diseases. CT images can directly reflect the morphological and functional pathological changes of diseases in organs, providing important diagnostic basis for doctors. However, the diagnosis and treatment efficiency in the emergency scenario is often limited by the following factors:

[0004] Heavy doctor burden: Emergency department doctors usually face high-intensity work, and excessive fatigue may lead to misdiagnosis or missed diagnosis; Image quality and lesion clarity: The quality of CT images is affected by multiple factors such as the patient's condition and scanning conditions, which may affect the identification of lesions; Diagnostic efficiency: The emergency scenario requires rapid decision-making, but the traditional image interpretation process is time-consuming and may delay treatment. Summary of the Invention

[0005] The purpose of this application is to provide an intelligent recognition method, device, electronic device and storage medium for emergency critical values to achieve accurate and rapid recognition of critical values.

[0006] According to the first aspect of the embodiments of this application, an intelligent recognition method for emergency critical values is provided. The method may include:

[0007] Obtain CT image data and clinical diagnosis guideline data of emergency patients;

[0008] Input the CT image data and clinical diagnosis guideline data into a trained critical value recognition network to obtain a critical value recognition result. The architecture of the trained critical value recognition network includes a visual branch network and a text branch network;

[0009] When the critical value recognition result exceeds the threshold, an alarm reminder is issued.

[0010] In some alternative embodiments of the present application, CT image data and clinical diagnosis guideline data are input into a trained critical value recognition network to obtain a critical value recognition result, including:

[0011] Input the CT image data into the visual branch network to obtain CT feature markers;

[0012] Input the clinical diagnosis guideline data into the text branch network to obtain core text information.

[0013] In some alternative embodiments of the present application, the visual branch network includes: a spatio-temporal Transformer module and a decoder;

[0014] Input the CT image data into the visual branch network to obtain CT feature markers, including:

[0015] Divide the CT image data into multiple image blocks according to the coronal plane, sagittal plane, and axial plane;

[0016] Input the image blocks into the spatio-temporal Transformer module for feature extraction to generate low-dimensional CT feature markers;

[0017] Reconstruct the low-dimensional CT feature markers through the decoder to obtain CT feature markers.

[0018] In some alternative embodiments of the present application, when the CT image data is a 2D image, perform repeated slicing on the CT image data to expand a depth dimension, so as to adapt to a three-dimensional input format.

[0019] In some alternative embodiments of the present application, the text branch network includes: a BERT text encoder;

[0020] Input the clinical diagnosis guideline data into the text branch network to obtain core text information, including:

[0021] Use the BERT text encoder to convert the clinical diagnosis guideline data into a unified 512-token format, and each token is represented by a 768-dimensional vector;

[0022] Normalize all tokens and map them to a compact 512-dimensional projection layer through a linear transformation to obtain core text information.

[0023] In some alternative embodiments of the present application, the architecture of the trained critical value recognition network further includes: a vision-language fusion module;

[0024] Input the CT image data and the clinical diagnosis guideline data into the trained critical value recognition network to obtain a critical value recognition result, further including:

[0025] Fuse the CT feature markers and the core text information to obtain fused features;

[0026] Obtain a critical value recognition result based on the fused features.

[0027] In some alternative embodiments of the present application, the feature fusion process satisfies the following formula:

[0028]

[0029] wherein, is the softmax function; Q is a query component reshaped from the visual feature F v and the text feature F t ; K is a key component reshaped from the visual feature F v and the text feature F t ; V is a value component reshaped from the visual feature F v and the text feature F t ; d represents the dimension of the key; f fused is the fused feature.

[0030] According to a second aspect of the embodiments of the present application, there is provided an intelligent emergency critical value recognition device, which may include:

[0031] An acquisition module, configured to acquire CT image data and clinical diagnosis guideline data of an emergency patient;

[0032] A recognition module, configured to input the CT image data and the clinical diagnosis guideline data into a trained critical value recognition network to obtain a critical value recognition result, and the architecture of the trained critical value recognition network includes a visual branch network and a text branch network;

[0033] An alarm module, configured to issue an alarm reminder when the critical value recognition result exceeds a threshold.

[0034] According to a third aspect of the embodiments of the present application, there is provided an electronic device, which may include:

[0035] A processor;

[0036] A memory for storing instructions executable by the processor;

[0037] wherein, the processor is configured to execute instructions to implement the emergency critical value intelligent recognition method as shown in any one of the embodiments of the first aspect.

[0038] According to a fourth aspect of the embodiments of the present application, there is provided a storage medium, when the instructions in the storage medium are executed by a processor of an information processing device or a server, enabling the information processing device or the server to implement the emergency critical value intelligent recognition method as shown in any one of the embodiments of the first aspect.

[0039] The above technical solution of the present application has the following beneficial technical effects:

[0040] By combining the critical value recognition network including a visual branch network and a text branch network with clinical diagnosis guideline data, the method of the embodiment of the present application can achieve fast and accurate recognition of critical values in CT image data. This method can save a large amount of human and material costs, and at the same time can significantly improve the utilization efficiency of medical resources. Brief Description of the Drawings

[0041] Figure 1 is a schematic flowchart of an emergency critical value intelligent recognition method in an exemplary embodiment of the present application;

[0042] Figure 2 is a network framework for emergency critical value intelligent recognition based on a multi-modal vision-language model in an exemplary embodiment of the present application;

[0043] Figure 3 is a schematic structural diagram of an emergency critical value intelligent recognition device in an exemplary embodiment of the present application;

[0044] Figure 4 is a schematic structural diagram of an electronic device in an exemplary embodiment of the present application;

[0045] Figure 5 is a schematic hardware structure diagram of an electronic device in an exemplary embodiment of the present application. Detailed Embodiments

[0046] To make the objectives, technical solutions, and advantages of the present application clearer and more understandable, the present application will be further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present application. In addition, in the following description, the descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present application.

[0047] The schematic diagram of the layer structure according to the embodiment of the present application is shown in the drawings. These figures are not drawn to scale, where for the purpose of clarity, some details are enlarged and some details may be omitted. The shapes of various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are merely exemplary. In practice, there may be deviations due to manufacturing tolerances or technical limitations, and those skilled in the art can design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0048] Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

[0049] In the description of the present application, it should be noted that the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0050] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0051] The intelligent emergency critical value recognition method, device, electronic device, and storage medium provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings, through specific embodiments and their application scenarios.

[0052] As Figure 1 shown, in the first aspect of the embodiments of the present application, an intelligent emergency critical value recognition method is provided. The method may include:

[0053] S110: Obtain CT image data and clinical diagnosis guideline data of an emergency patient;

[0054] S120: Input the CT image data and clinical diagnosis guideline data into a trained critical value recognition network to obtain a critical value recognition result. The architecture of the trained critical value recognition network includes a visual branch network and a text branch network;

[0055] S130: When the critical value recognition result exceeds the threshold, send an alarm reminder.

[0056] By combining the critical value recognition network including the visual branch network and the text branch network with the clinical diagnosis guideline data, the method of this embodiment can achieve fast and accurate recognition of critical values of CT image data. This method can save a large amount of human and material costs, and at the same time can significantly improve the utilization efficiency of medical resources.

[0057] In this embodiment, the vision branch network can be a 3D Vision Transformer (3D ViT) model. As a three-dimensional vision Transformer model, 3D ViT can extract low-dimensional feature tokens from high-dimensional CT volume data and reconstruct the CT volume data through these tokens. Exemplarily, during the processing, the vision branch network divides the CT volume data into uniform image blocks of size 30×30×15 in the coronal plane, sagittal plane, and axial plane. Through this operation, the high-dimensional CT data can be transformed into a one-dimensional sequence input form adapted to the Transformer model, so as to capture local information and model global relationships through the global attention mechanism. In addition, the chunking strategy significantly reduces the computational complexity and makes the high-resolution CT image processing task feasible. These image blocks are first input into the spatial and temporal Transformer module for feature extraction to generate low-dimensional CT feature tokens, and then the 3D CT volume data is reconstructed through the decoder. In the encoder part of the model, the CT image blocks are processed to generate low-dimensional feature tokens. These feature tokens are flattened and mapped to a 512-dimensional projection space through a linear transformation, and finally the encoded CT feature tokens are generated.

[0058] In addition, the text branch network can be a pre-trained BERT text encoder, which is a Transformer-based language model pre-trained on medical imaging reports. This encoder can process 512 tokens. For reports with fewer than 512 tokens, in this embodiment, <pad>The tokens are padded to standardize the length. Exemplarily, the "Treatment Recommendations" section in the diagnostic guidelines is processed using a text encoder (i.e., the text branch network) and converted into a unified 512-token format. Each token is represented as a 768-dimensional vector. These tokens are then normalized through a summation operation and mapped to a compact 512-dimensional projection layer through a linear transformation, thereby condensing the core text information in the guidelines. As Figure 2 shown, this encoding process is visually demonstrated.

[0059] In addition, the network architecture of this embodiment may further include a Vision Language Fusion Module, which can process the fusion of text features and visual features. The specific process may include: mapping the visual features and text features into a shared embedding space. In the visual feature branch of the network, the visual semantic feature F v is directly encoded and linearly mapped from the original CT scan data through a visual encoder. In addition, the Vision Language Fusion Module adopts a Transformer-based attention mechanism to further enhance the association between visual features and text semantic features. This module consists of a Multi-head Self-attention Layer. During the fusion process, the output F v of the visual feature branch and the output F t of the text branch are respectively reshaped into query, and key-value pairs.

[0060] In some embodiments, the CT image data and clinical diagnosis guideline data are input into a trained critical value recognition network, and the critical value recognition results are obtained, including:

[0061] Inputting the CT image data into the visual branch network to obtain CT feature tokens;

[0062] Inputting the clinical diagnosis guideline data into the text branch network to obtain core text information.

[0063] In some embodiments, the visual branch network includes: a spatial and temporal Transformer module and a decoder;

[0064] Inputting the CT image data into the visual branch network to obtain CT feature tokens, including:

[0065] Dividing the CT image data into multiple image blocks according to the coronal plane, sagittal plane, and axial plane;

[0066] Inputting the image blocks into the spatial and temporal Transformer module for feature extraction to generate low-dimensional CT feature tokens;

[0067] Reconstruct the low-dimensional CT feature markers through a decoder to obtain CT feature markers.

[0068] In some embodiments, when the CT image data is a 2D image, the CT image data is subjected to repeated slicing processing to expand a depth dimension, so as to be adapted to a three-dimensional input format.

[0069] In some embodiments, the text branch network includes: a BERT text encoder;

[0070] Input the clinical diagnosis guideline data into the text branch network to obtain core text information, including:

[0071] Use the BERT text encoder to convert the clinical diagnosis guideline data into a unified 512-token format, and each token is represented by a 768-dimensional vector;

[0072] Normalize all tokens and map them to a compact 512-dimensional projection layer through a linear transformation to obtain core text information.

[0073] In some embodiments, the architecture of the trained critical value recognition network further includes: a vision-language fusion module;

[0074] Input the CT image data and the clinical diagnosis guideline data into the trained critical value recognition network to obtain a critical value recognition result, and further include:

[0075] Fuse the CT feature markers and the core text information to obtain fused features;

[0076] Identify the critical value recognition result according to the fused features.

[0077] In some embodiments, the feature fusion process satisfies the following formula:

[0078]

[0079] Among them, is the softmax function; Q is the query component reshaped from the visual feature F v and the text feature F t ; K is the key component reshaped from the visual feature F v and the text feature F t ; V is the value component reshaped from the visual feature F v and the text feature F t ; d represents the dimension of the key; f fused is the fused feature.

[0080] During the training phase, the model is trained through supervised learning. The encoded CT features and text features are optimized with the corresponding ground truth labels using the Cross-Entropy (CE) loss. The formula for the CE loss is as follows:

[0081] L CE (y i , p(y i )) = -E[y i ·log(p(y i )) + (1 - y i )·log(1 - p(y i ))]

[0082] where E represents the expected value over all samples, i.e., the average of the losses over the entire dataset. L CE is the cross-entropy loss function during the model training process, which is used to measure the difference between the model's prediction results and the ground truth labels in the current state.

[0083] During the training process, the L CE loss function guides the model to adjust its parameters to optimize the prediction results through the mechanisms of backpropagation and gradient descent. log(.) represents the logarithmic function. y i is the binary class label (0 or 1). p(yi) represents the probability value that the model's predicted output belongs to the y i label. This formula evaluates the model's performance by calculating the logarithmic difference between the predicted probability and the ground truth label. By minimizing this loss function, the model can gradually improve its prediction accuracy.

[0084] The overall training process of the network is as follows: First, set the input 3D CT data as I ∈ R C×D×H×W , C, D, H, and W respectively represent the channels, depth, height, and width of the input image. The corresponding classification label is y. In this invention, a 3D ViT model is adopted to be compatible with the input of 2D images and 3D images. For 2D images, a depth dimension can be extended by repeating slices, so as to adapt it to the three-dimensional input format. Min-Max Normalization can be used to preprocess three-dimensional CT images to standardize the input data. In addition, the three-dimensional images are uniformly adjusted to the standard size of 128×128×128. In the visual encoder, a model based on the three-dimensional vision Transformer (3D ViT) is adopted. This model consists of 12 layers of Transformer, and the patch size is 4×16×164. The output feature embedding dimension is 2048×768, representing 2048 feature tokens, and each token contains 768-dimensional features. After being processed by the 3D Spatial Pooling Perceiver, the finally generated visual feature token dimension is 256×768.

[0085] In the pre-training stage of the visual encoder, in this embodiment, BERT containing 12 layers of Transformer can be used as the text encoder, and the maximum text length it supports is 128 characters. The [CLS] tokens of the visual encoder and the text encoder are used as the global feature representation, and the feature projection for cross-modal contrast learning is completed through a linear layer. In addition, this embodiment is trained in parallel on 4 GPUs, the batch size is set to 12, the learning rate is 10-4, and the warm-up and cosine decay learning rate scheduling strategy is used. Experiments have proved that this setting can significantly improve the training effect and accelerate the convergence of the model.

[0086] The above embodiments aim to provide an innovative framework based on a multi-modal vision-language large model, which integrates patient characteristics, clinical diagnosis guidelines, and imaging data. Through training on a large-scale multi-modal dataset, this model has strong visual and text learning capabilities and excellent generalization capabilities. Experiments show that this technology has significantly improved the performance of the model in the emergency critical value recognition task, providing a new intelligent solution for critical value recognition in the emergency scenario.

[0087] In addition, a large language model (LLM) and clinical diagnostic guidelines are introduced as the prior knowledge of doctors for encoding, and the text features of clinical diagnostic guidelines are embedded into the image-text joint learning framework. By deeply fusing the encoded text feature vectors with the image feature vectors, the performance of the model in the identification of emergency critical values is significantly enhanced. The large language model effectively utilizes its understanding ability of medical knowledge, further improving the complementarity and synergy of multimodal information in complex tasks. The experimental results show that this strategy significantly improves the performance of the model in multimodal information processing, providing strong technical support for complex emergency diagnosis and treatment scenarios.

[0088] The methods of the above embodiments can achieve quantitative calculation of key indicators. Through accurate disease judgment, the system can effectively assist clinicians in making rapid decisions, saving a large amount of human and material costs for subsequent emergency treatment and significantly improving the utilization efficiency of medical resources. In addition, by organically combining 3D vision transformers (3D ViT), large language models (LLMs), and diagnostic guideline strategies, high-precision identification of emergency critical values is achieved. This method not only significantly improves the diagnostic efficiency but also greatly shortens the diagnostic time of patients, providing a more reliable auxiliary diagnostic tool for clinicians.

[0089] It should be noted that for the emergency critical value intelligent identification method provided in the embodiments of the present application, the execution subject can be an emergency critical value intelligent identification device or a control module in the emergency critical value intelligent identification device for executing the method of emergency critical value intelligent identification. In the embodiments of the present application, the method of emergency critical value intelligent identification executed by the emergency critical value intelligent identification device is taken as an example to illustrate the emergency critical value intelligent identification device provided in the embodiments of the present application.

[0090] As Figure 3 shown, in the second aspect of the embodiments of the present application, an emergency critical value intelligent identification device is provided, and the device may include:

[0091] An acquisition module 310, configured to acquire CT image data and clinical diagnostic guideline data of an emergency patient;

[0092] An identification module 320, configured to input the CT image data and clinical diagnostic guideline data into a trained critical value identification network 330 to obtain a critical value identification result, and the architecture of the trained critical value identification network includes a visual branch network and a text branch network;

[0093] An alarm module 340, configured to issue an alarm reminder when the critical value identification result exceeds a threshold.

[0094] The emergency critical value intelligent recognition device in the embodiments of this application can be a device, or a component, integrated circuit, or chip in a terminal. This device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a handheld computer, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), etc. The non-mobile electronic device can be a server, a Network Attached Storage (NAS), a personal computer (PC), a television (TV), a teller machine, or a self-service machine, etc. The embodiments of this application do not make specific limitations.

[0095] The emergency critical value intelligent recognition device in the embodiments of this application can be a device with an operating system. This operating system can be the Android operating system, can be the iOS operating system, or can also be other possible operating systems. The embodiments of this application do not make specific limitations.

[0096] The emergency critical value intelligent recognition device provided in the embodiments of this application can implement Figure 1 each process implemented by the method embodiments. To avoid repetition, it will not be elaborated here.

[0097] Optionally, as Figure 4 shown, the embodiments of this application further provide an electronic device 400, including a first processor 401, a first memory 402, a program or instruction stored on the first memory 402 and executable on the first processor 401. When the program or instruction is executed by the first processor 401, it implements each process of the above-mentioned emergency critical value intelligent recognition method embodiments and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0098] It should be noted that the electronic devices in the embodiments of this application include the above-mentioned mobile electronic devices and non-mobile electronic devices.

[0099] Figure 5 It is a schematic diagram of the hardware structure of an electronic device for implementing the embodiments of this application.

[0100] The hardware structure 500 of this electronic device includes but is not limited to: a radio frequency unit 501, a network module 502, an audio output unit 503, an input unit 504, a sensor 505, a display unit 506, a user input unit 507, an interface unit 508, a second memory 509, and a second processor 510, etc.

[0101] Those skilled in the art can understand that the hardware structure 500 of the electronic device may further include a power supply (such as a battery) for powering each component. The power supply can be logically connected to the second processor 510 through a power management system, so as to manage functions such as charging, discharging, and power consumption management through the power management system. Figure 5 The electronic device structure shown does not limit the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.

[0102] It should be understood that in the embodiments of the present application, the input unit 504 may include a Graphics Processing Unit (GPU) 5041 and a microphone 5042. The graphics processor 5041 processes the image data of static pictures or videos obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The display unit 506 may include a display panel 5061, and the display panel 5061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 507 includes a touch panel 5071 and other input devices 5072. The touch panel 5071 is also called a touch screen. The touch panel 5071 may include two parts: a touch detection device and a touch controller. The other input devices 5072 may include, but are not limited to, a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, which will not be elaborated here. The second memory 509 may be used to store software programs and various data, including but not limited to application programs and operating systems. The second processor 510 may integrate an application processor and a modem processor. Among them, the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communication. It can be understood that the above modem processor may not be integrated into the second processor 510.

[0103] The embodiments of the present application also provide a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, it implements each process of the above embodiment of the emergency critical value intelligent recognition method and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0104] Among them, the processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer-readable storage medium, such as a computer Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disc, etc.

[0105] Another embodiment of the present application provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is configured to run programs or instructions to implement each process of the above-mentioned embodiment of the emergency critical value intelligent recognition method, and can achieve the same technical effects. To avoid repetition, details are not described here again.

[0106] It should be understood that the chip mentioned in the embodiments of the present application may also be referred to as a system-on-chip, system chip, chip system, or system-on-chip, etc.

[0107] It should be noted that in this document, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or device comprising such element. In addition, it should be pointed out that the methods and devices in the embodiments of the present application are not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in a reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. Additionally, the features described with reference to certain examples may be combined in other examples.

[0108] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc) and includes several instructions for causing a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present application.

[0109] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them belong to the protection scope of the present application.< / pad>

Claims

1. An intelligent identification method for emergency critical value, characterized in that: include: Obtain CT imaging data and clinical diagnostic guideline data of emergency patients; Inputting the CT image data and the clinical diagnosis guide data into a trained critical value recognition network to obtain a critical value recognition result, wherein the architecture of the trained critical value recognition network includes a visual branch network and a text branch network; When the critical value identification result exceeds the threshold, an alarm is issued.

2. The intelligent identification method for emergency critical value according to claim 1 is characterized in that: The step of inputting the CT image data and the clinical diagnosis guideline data into a trained critical value recognition network to obtain a critical value recognition result includes: Inputting the CT image data into the visual branch network to obtain CT feature labels; The clinical diagnosis guideline data is input into the text branch network to obtain core text information.

3. The intelligent identification method for emergency critical value according to claim 2 is characterized in that: The visual branch network includes: a spatial and temporal Transformer module and a decoder; The step of inputting the CT image data into the visual branch network to obtain CT feature labels includes: Dividing the CT image data into a plurality of image blocks according to the coronal plane, the sagittal plane and the axial plane; Inputting the image block into the spatial and temporal Transformer module for feature extraction to generate low-dimensional CT feature markers; The low-dimensional CT feature marker is reconstructed through the decoder to obtain a CT feature marker.

4. The intelligent identification method for emergency critical value according to claim 3 is characterized in that: When the CT image data is a 2D image, the CT image data is repeatedly sliced ​​to expand a depth dimension so as to adapt to a three-dimensional input format.

5. The intelligent identification method for emergency critical value according to claim 2 is characterized in that: The text branch network includes: BERT text encoder; The step of inputting the clinical diagnosis guideline data into the text branch network to obtain core text information includes: Using the BERT text encoder to convert the clinical diagnosis guideline data into a unified 512-token format, each token is represented by a 768-dimensional vector; All tags are normalized and mapped to a compact 512-dimensional projection layer through linear transformation to obtain the core text information.

6. The intelligent identification method for emergency critical value according to claim 2 is characterized in that: The architecture of the trained critical value recognition network also includes: a visual language fusion module; The step of inputting the CT image data and the clinical diagnosis guideline data into a trained critical value recognition network to obtain a critical value recognition result further includes: Performing feature fusion on the CT feature mark and the core text information to obtain a fusion feature; A critical value recognition result is obtained according to the fusion feature recognition.

7. The intelligent identification method for emergency critical value according to claim 6 is characterized in that: The feature fusion process satisfies the following formula: in, is the softmax function; Q is the visual feature F v and text features F t The query component reshaped from v , and text features F t The key components reshaped from the visual features F v and text features F t The value component obtained by reshaping in ; d represents the dimension of the key; f fused For fusion features.

8. An intelligent identification device for emergency critical value, characterized in that: include: An acquisition module is used to obtain CT imaging data and clinical diagnosis guideline data of emergency patients; A recognition module, used for inputting the CT image data and the clinical diagnosis guide data into a trained critical value recognition network to obtain a critical value recognition result, wherein the architecture of the trained critical value recognition network includes a visual branch network and a text branch network; The alarm module is used to issue an alarm when the critical value identification result exceeds a threshold.

9. An electronic device, characterized in that: include: A processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method for intelligent identification of emergency critical values ​​as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the intelligent identification method for emergency critical values ​​as described in any one of claims 1-7 are implemented.

Images