Rapid injury inspection classification method and system based on artificial intelligence

Abstract

The invention provides a rapid injury inspection classification method and system based on artificial intelligence, and relates to the technical field of data processing, and the method comprises the steps: obtaining the physiological data and image data of each wounded person on site; a multi-layer perceptron (MLP) is used for rapidly analyzing the physiological data, and physiological feature vectors in the physiological data are extracted; an efficient neural network model OfficientNet is used to extract an image feature vector in the image data; using a full connection layer to splice the physiological feature vector and the image feature vector to form a fusion feature vector; taking each data feature in the fusion feature vector as an independent input channel, and inputting the data features into a quick injury inspection classification model based on a quantum neural network (QNN); and classifying the emergency degree of the injury condition of each wounded person by using a quick injury inspection classification model based on a quantum neural network. According to the invention, the injury emergency degree of each wounded person can be automatically and rapidly classified.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a rapid damage classification method and system based on artificial intelligence. Background Technology

[0002] Triage is a crucial part of emergency rescue operations at disaster and accident sites. It involves classifying and prioritizing injured individuals based on their injuries and the urgency of their treatment, and allocating limited medical resources accordingly.

[0003] Currently, casualty triage mainly relies on manual identification and classification by emergency medical personnel.

[0004] However, manual identification is time-consuming, requiring emergency personnel to assess each injured person individually, which consumes a lot of time and can easily lead to a shortage of human resources. At disaster sites, there are many injured people and a limited number of emergency personnel, so the efficiency of manual classification cannot meet the needs on site. At the same time, manual classification may lead to classification errors or incorrect prioritization, resulting in an unreasonable allocation of medical resources. Summary of the Invention

[0005] To address the technical problems of time-consuming manual triage methods in existing technologies, which require emergency personnel to assess each injured person individually, consuming a significant amount of time and easily leading to a shortage of human resources, and given the large number of injured people and limited emergency personnel at disaster sites, manual triage cannot meet the needs of the scene. Furthermore, manual triage may lead to classification errors or incorrect prioritization, resulting in the unreasonable allocation of medical resources, this invention provides a rapid triage method and system based on artificial intelligence.

[0006] The technical solutions provided by the embodiments of the present invention are as follows: First aspect: This invention provides a rapid damage classification method based on artificial intelligence, comprising: S1: Acquire physiological and image data of each injured person at the scene; S2: Use a multilayer perceptron (MLP) to quickly analyze the physiological data and extract the physiological feature vectors from the physiological data; S3: Use the efficient neural network model EfficientNet to extract image feature vectors from the image data; S4: Using a fully connected layer, the physiological feature vector and the image feature vector are concatenated to form a fused feature vector; S5: Each data feature in the fused feature vector is used as a separate input channel and input into the fast damage classification model based on quantum neural network QNN; S6: Use a rapid injury classification model based on quantum neural networks to classify the urgency of each patient's injury.

[0007] The second aspect: This invention provides a rapid damage classification system based on artificial intelligence, comprising: processor; A memory storing computer-readable instructions, which, when executed by the processor, implement the rapid damage classification method based on artificial intelligence as described in the first aspect.

[0008] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, physiological data and image data of each injured person at the scene are used as input. Through multilayer perceptron (MLP), efficient neural network model EfficientNet, and quantum neural network (QNN), the severity of each injured person's injury is automatically and quickly classified. This gives emergency personnel more time, improves the efficiency and accuracy of patient classification, ensures that limited medical resources are allocated rationally, and helps doctors make more accurate treatment decisions based on the specific data of the injured persons during the pre-hospital emergency stage. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating a rapid injury classification method based on artificial intelligence, provided in an embodiment of the present invention; Figure 2 A schematic diagram of a rapid injury classification method based on artificial intelligence provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a rapid damage classification system based on artificial intelligence, provided as an embodiment of the present invention. Detailed Implementation

[0011] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0012] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0013] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0014] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0015] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0016] Reference manual attached Figure 1 The diagram shows a flowchart of a rapid damage classification method based on artificial intelligence provided by an embodiment of the present invention.

[0017] Reference manual attached Figure 2 The diagram shows a structural schematic of a rapid damage classification method based on artificial intelligence provided in an embodiment of the present invention.

[0018] This invention provides an artificial intelligence-based rapid damage classification method, which can be implemented by an artificial intelligence-based rapid damage classification device, which can be a terminal or a server. The processing flow of the artificial intelligence-based rapid damage classification method may include the following steps: S1: Acquire physiological data and image data of each injured person at the scene.

[0019] Optionally, physiological data include: wound data, bleeding data, activity status data, consciousness response data, respiratory data, and / or neck vascular pulsation data.

[0020] Among them, wound data refers to the external injuries of the wounded; Among them, bleeding data refers to the amount and condition of bleeding in the injured person.

[0021] Among them, activity status data refers to the injured person's mobility or physical activity.

[0022] Among them, consciousness response data is used to record the wounded's response to external stimuli, reflecting the wounded's neurological health status.

[0023] Among them, respiratory data refers to the patient's breathing condition, including breathing rate, depth, rhythm, etc.

[0024] Among them, neck vascular pulsation data refers to the pulsation of the main blood vessels (such as the carotid artery) in the neck of the injured person.

[0025] Optionally, image data can be captured by drones. This image data clearly shows the specific location of each injured person, enabling emergency personnel to accurately locate them and prioritize treatment. Through efficient image processing technology, drone-captured image data can be used to automatically analyze the injured person's external trauma, bleeding, and other conditions. For example, open wounds, burns, and fractures can be quickly identified through images, providing crucial information for injury classification.

[0026] S2: Use a multilayer perceptron (MLP) to quickly analyze physiological data and extract physiological feature vectors from the physiological data.

[0027] A Multilayer Perceptron (MLP) is a feedforward neural network composed of multiple fully connected layers, where the output of each layer serves as the input to the next. An MLP typically includes an input layer, one or more hidden layers, and an output layer. Each neuron performs a weighted summation of the input through weights and biases, followed by a nonlinear transformation using an activation function. MLPs are widely used in tasks such as classification and regression, and their advantage lies in their ability to learn complex nonlinear mappings and their ability to be trained and optimized using backpropagation algorithms.

[0028] In one possible implementation, S2 specifically includes sub-steps S201 to S203: S201: Input physiological data into the input layer of the multilayer perceptron.

[0029] S202: In the hidden layer of the multilayer perceptron, a nonlinear transformation is performed through an activation function to extract the hidden state from the physiological data.

[0030] Optionally, the hidden layers of a multilayer perceptron include a first hidden layer, a second hidden layer, and a third hidden layer.

[0031] The first and second hidden layers use the ReLU activation function. Using the ReLU activation function in the first and second hidden layers helps the network better process and learn complex physiological data features during training, and quickly extract useful patterns from the data.

[0032] The third hidden layer uses the Sigmoid activation function. The Sigmoid activation function restricts the output to the range [0, 1], making it suitable for the last layer of the network or layers that require probabilistic outputs. Here, the Sigmoid can be used to normalize the hidden states of the output, ensuring that the output conforms to certain specific ranges.

[0033] S203: In the output layer of the multilayer perceptron, the hidden states output by the hidden layers are summarized and the physiological feature vector is output.

[0034] In this invention, the advantage of using a Multilayer Perceptron (MLP) to quickly analyze physiological data and extract physiological feature vectors lies in the fact that the MLP can effectively learn and extract complex nonlinear relationships in the data through a multilayer fully connected network. It performs nonlinear transformations through activation functions, enabling the network to capture subtle features and patterns from physiological data, thereby improving classification accuracy.

[0035] S3: Use the efficient neural network model EfficientNet to extract image feature vectors from image data.

[0036] Among them, EfficientNet is a high-efficiency convolutional neural network (CNN) that optimizes the network's depth, width, and input resolution by introducing a composite scaling method, achieving higher accuracy with less computational resources. EfficientNet's core advantage lies in its ability to simultaneously increase the network's depth, width, and resolution through a reasonable scaling ratio, achieving an optimal balance between accuracy and computational efficiency.

[0037] In one possible implementation, the efficient neural network model includes an MBConv module. The MBConv module introduces an inverse residual structure and combines depthwise separable convolutions and extended convolutions to optimize the utilization of computational resources.

[0038] This invention innovatively introduces the SENet unit, an attention mechanism, into the MBConv module. The SENet unit employs the Swish activation function. The core idea of ​​the SENet unit is to adaptively weight the features of each channel to enhance the expressive power of important features and suppress unimportant features.

[0039] To further enhance the feature extraction capability of efficient neural network models, this invention innovatively adds a recursive gated convolution structure to the efficient neural network model.

[0040] Specifically, the basic structure of the recursive gated convolutional structure is as follows: ; ; in, x Indicates the input feature map, This represents a linear projection mapping used to mix channel information. p 0、 q 0 indicates that the two projected features obtained from the segmentation of the input feature map are 0. f This represents a depthwise convolution operation. This indicates element-wise multiplication. p 1 indicates interactive features. This represents the inverse linear projection mapping, used to remap interactive features back to the original input channels.

[0041] Perform multi-level mapping on the input feature map: ; in, p 0、 q 0、…、 q n-1 This represents a set of projected features obtained by segmenting from the input feature map.

[0042] Recursively perform gated convolution: ; in, p k+1 Indicates the first k +1st order interactive features, f k Indicates the first k Depth-order convolution operations, q k Indicates the first k 1st order projection features g k Indicates the use of matching the first k A linear mapping of the first-order feature dimension. α This represents the scaling factor.

[0043] ; ; in, Identity Describes the identity function. Linear Represents a linear mapping function. C k-1 Indicates the firstk -1 number of feature channels, C k Indicates the first k Number of feature channels of order, C This represents the number of feature channels in the input feature map. n This represents the total order of the recursive gated convolution. β This represents the channel reduction factor. The channel reduction factor is typically 2.

[0044] By using inverse linear projection mapping, interactive features are remapped back to the original input channels.

[0045] ; in, y Represents recursively gated convolution features. Represents the inverse linear projection mapping. p k Indicates the first k Interactive features of order.

[0046] In this invention, the recursive gated convolutional structure combines deep convolution operations and gating mechanisms. Through the interaction of multiple recursive operations and gating mechanisms, the network can better capture features at different scales, thereby improving the network's expressiveness and ability to extract fine-grained features.

[0047] S4: Using a fully connected layer, the physiological feature vector and the image feature vector are concatenated to form a fused feature vector.

[0048] In this invention, physiological data and image data are typically complementary, providing the patient's physiological state and external injuries, respectively. By concatenating the physiological feature vector with the image feature vector to form a fused feature vector, the network can utilize multimodal information to perform a more comprehensive assessment of the patient's condition, thereby improving feature representation capabilities and enhancing classification accuracy.

[0049] S5: Each data feature in the fused feature vector is used as a separate input channel and input into the fast damage classification model based on quantum neural network (QNN).

[0050] Quantum Neural Networks (QNNs) are an advanced model combining quantum computing and neural networks. They leverage the properties of quantum computing (such as superposition, entanglement, and parallel computing) to handle complex computational tasks. QNNs replace neurons in traditional neural networks with qubits and use quantum gates (such as rotation gates and controlled rotation gates) for information processing and transmission. Compared to classical neural networks, QNNs can process larger datasets in a shorter time and, through the efficiency and powerful computational capabilities of quantum computing, improve the model's performance in feature extraction, pattern recognition, and classification tasks. This makes them better suited for scenarios requiring more time for casualty classification.

[0051] S6: Use a rapid injury classification model based on quantum neural networks to classify the urgency of each patient's injury.

[0052] In one possible implementation, S6 specifically includes sub-steps S601 to S604: S601: Convert each input feature value into a qubit. ; in, x i Indicates the first i A quantum bit with data characteristics t i Indicates the first i The eigenvalues ​​of the data features are given by cosine and sin. T This indicates the transpose operation.

[0053] It's worth noting that converting each data feature into a qubit leverages the properties of quantum computing to improve the accuracy and efficiency of feature representation. This enables QNNs to understand and extract key information from physiological and image data at a higher dimension.

[0054] S602: Using a phase rotation gate, the individual qubits are aggregated, and the aggregation rotation angle is determined: ; in, θ j Indicates to the first j The aggregated rotation angle of the inputs to the quantum neurons in the hidden layer, where arg represents the phase angle. R Indicates a phase rotation operation. θ ij Indicates to the first j When the input of the first hidden layer quantum neuron is... i The rotation angle of each data feature, where argtan represents the arctangent function. n This represents the total number of data features.

[0055] It should be noted that by applying phase rotation gates, QNNs can optimize and refine feature data at the quantum level, making the data representation of each qubit more accurate, thereby improving classification accuracy.

[0056] S603: Based on the aggregated rotation angle, use a controlled rotating door to perform a reverse rotation operation to determine the hidden state: ; in, h j Indicates the first j The hidden states of quantum neurons in a hidden layer. π Represents pi (π). γ Indicates the range control parameter. σ This represents the activation function.

[0057] It should be noted that by using quantum rotating gates and controlled rotating gates, QNNs can effectively handle nonlinear relationships in the data, further enhance their ability to learn the characteristics of the wounded, and help the model better understand the wounded's condition and accurately judge its urgency.

[0058] S604: Based on the hidden state, classify the urgency of each wounded soldier's injury: ; in, P k Indicates belonging to the first k The probability of each category of injury urgency. g This represents the activation function. w jk Indicates the first j The hidden layer quantum neurons and their outputs k The weights between the output neurons of each category classification result m This represents the total number of quantum neurons in the hidden layer.

[0059] It should be noted that by introducing activation functions, QNN can increase the model's non-linear expressive power, better adapt to different injuries, and generate different probabilities for the severity of each injury. This not only accurately classifies the severity of injuries (such as critical, serious, minor, or fatal), but also ensures the rational allocation of emergency resources, prioritizing the treatment of those most in need of urgent care.

[0060] Optionally, the severity of injury categories include: critically injured, seriously injured, slightly injured, and dead.

[0061] In this invention, QNN can improve the performance of classification models through the high precision of quantum computing, reduce misdiagnosis caused by manual operation or traditional calculation methods, accurately classify the condition of the injured, and ensure that emergency personnel can make timely responses based on accurate information.

[0062] In one possible implementation, the Seagull optimization algorithm is used to optimize the model parameters of the multilayer perceptron, the efficient neural network model, and the fast damage classification model based on the quantum neural network QNN.

[0063] The Seagull Optimization Algorithm is a swarm intelligence optimization algorithm that simulates the foraging behavior of seagull flocks. This algorithm optimizes the objective function by simulating the cooperation and competition mechanisms among seagulls during their food search, using a combination of local and global search methods. The core idea of ​​the Seagull Optimization Algorithm is that seagull flocks adjust their flight routes based on the location, quality, and quantity of food to find the optimal food source.

[0064] Specifically, the fitness function of the Seagull optimization algorithm is set as the reciprocal of the cross-entropy loss function.

[0065] Initialize seagull individuals. Each seagull individual represents a feasible set of model parameters. Each seagull individual consists of multiple dimensional components, and each component represents a model parameter.

[0066] During the global search phase, collisions are avoided and the movement proceeds towards the optimal individual: ; ; ; in, Indicates the first t During the nth iteration i The location of an individual seagull after the global search phase. Indicates the first t During the nth iteration i The positions of individual seagulls after collision protection treatment A Indicates control factor. Indicates the first t During the nth iteration i The location of each individual seagull Indicates the first t During the nth iteration i The displacement of each seagull towards the optimal individual. B Indicates the search balance factor. Indicates the first t The optimal individual position in the next iteration.

[0067] In this invention, during the global search phase, the Seagull optimization algorithm effectively balances the exploration and development processes by avoiding collisions and guiding individuals towards the optimal solution. In this way, individual seagulls can avoid collisions in the solution space and gradually approach the global optimum, improving the algorithm's convergence speed and accuracy, and demonstrating stronger adaptability and efficiency, especially in complex optimization problems.

[0068] ; in, t Indicates the current iteration number. T Indicates the maximum number of iterations. f c This indicates the linearly decreasing frequency.

[0069] In this invention, the control factor gradually decreases with the increase of the number of iterations. Initially, the algorithm explores more extensively, covering a wider solution space. However, as iterations progress, it gradually shifts to more refined exploration, focusing on the local optimization of the optimal solution. This strategy helps prevent the algorithm from getting trapped in local optima too early, while improving the algorithm's convergence efficiency and global search capability, thereby enhancing the algorithm's performance in complex optimization problems.

[0070] ; in, r 1 represents a random number between 0 and 1.

[0071] In this invention, introducing a random factor to adjust the search balance factor adds randomness and diversity to the search process of the Seagull optimization algorithm. The random factor makes the search step size of each individual seagull variable in each iteration, helping to avoid overly rigid search paths. This enhances the algorithm's global search capability, preventing stagnation near local optima. Furthermore, by dynamically adjusting the search intensity, it improves the efficiency of exploring the solution space, enhancing the algorithm's adaptability, especially in complex high-dimensional optimization problems where it can better find the global optimum.

[0072] During the local search phase, a random number is generated. r 2. Based on random numbers r 2. A parallel selection is made between the spiral search strategy and the encirclement strategy, with displacement performed in a spiral motion manner: ; ; ; ; ; in, Indicates the firstt During the nth iteration i The position of an individual seagull after spiraling. x express x Directional spiral flight coefficient, y express y Directional spiral flight coefficient, z express z Directional spiral flight coefficient, r Indicates the radius of the spiral flight trajectory. θ Represents 0 to 2 π Random numbers between u , v Represents the helical constant. e Represents the natural constant.

[0073] In this invention, during the local search phase, a dynamic exploration method is introduced into the Seagull Optimization Algorithm by generating random numbers and selecting between a spiral search strategy and an encirclement strategy. Using a spiral search increases the diversity and exploratory nature of individual searches, gradually approaching the optimal solution in different directions, thus avoiding getting trapped in local optima. The encirclement strategy involves a tight local search near the optimal individual. Combining these two strategies enhances the algorithm's global search and local optimization capabilities, making the search process more flexible and effectively improving the convergence speed and accuracy of the algorithm in complex high-dimensional optimization problems. The introduction of the spiral trajectory broadens the algorithm's exploration space and reduces the risk of premature convergence.

[0074] Perform mutation operations on each individual seagull: ; in, Indicates the first t During the nth iteration i The location of the individual seagull after mutation. P r Represents a random individual. ω This represents the adaptive scaling factor.

[0075] In this invention, the mutation operation helps individuals escape the predicament of local optima and explore a wider solution space by adjusting the difference between the current optimal solution and the random individual.

[0076] ; in, ω max Indicates the maximum scaling factor. ω min This represents the minimum scaling factor.

[0077] In this invention, the intensity of the mutation operation can be dynamically adjusted throughout the optimization process. Initially, a larger scaling factor promotes extensive exploration, helping to search different regions of the solution space and avoiding early entrapment in local optima. As iterations progress, the scaling factor gradually decreases, approaching its minimum value, causing the search process to gradually shift towards refined local optimization, accelerating convergence to the global optimum.

[0078] Determine if the fitness value of the mutated position is greater than the fitness value of the original position. If yes, replace the original position with the mutated position. Otherwise, leave the original position unchanged.

[0079] Update the fitness values ​​of each individual seagull and the global best individual.

[0080] Determine if the current iteration count has reached the maximum iteration count. If yes, output the set of model parameters represented by the seagull individual with the highest fitness. Otherwise, return to continue iterating.

[0081] In this invention, the Seagull optimization algorithm has strong global search capabilities and fast convergence speed. It can optimize the parameters of complex models (such as MLP, EfficientNet, QNN), improve the classification accuracy and generalization ability of the model, and reduce training time and computational resource consumption, so as to ensure that the classification of wounded is more accurate and efficient.

[0082] In one possible implementation, the AI-based rapid damage classification method further includes: S7: Issue alert messages based on the urgency of each injured person's condition.

[0083] In this invention, by automatically generating classification results of the urgency of the injuries and conveying them to emergency personnel in real time through prompts, the efficiency of emergency decision-making can be significantly improved. Emergency personnel do not need to manually determine the priority of each injured person, but can take swift action based on the prompts provided by the system.

[0084] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, physiological data and image data of each injured person at the scene are used as input. Through multilayer perceptron (MLP), efficient neural network model EfficientNet, and quantum neural network (QNN), the severity of each injured person's injury is automatically and quickly classified. This gives emergency personnel more time, improves the efficiency and accuracy of patient classification, ensures that limited medical resources are allocated rationally, and helps doctors make more accurate treatment decisions based on the specific data of the injured persons during the pre-hospital emergency stage.

[0085] Reference manual attached Figure 3The diagram shows a structural schematic of a rapid damage classification system based on artificial intelligence provided by the present invention.

[0086] The present invention also provides an artificial intelligence-based rapid injury classification system 20, applied to the above-mentioned artificial intelligence-based rapid injury classification method, comprising: Processor 201; The memory 202 stores computer-readable instructions, which, when executed by the processor 201, implement the rapid damage classification method based on artificial intelligence as described in the method embodiment.

[0087] The AI-based rapid injury classification system 20 provided by this invention can execute the above-mentioned AI-based rapid injury classification method and achieve the same or similar technical effects. To avoid duplication, this invention will not elaborate further.

[0088] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, physiological data and image data of each injured person at the scene are used as input. Through multilayer perceptron (MLP), efficient neural network model EfficientNet, and quantum neural network (QNN), the severity of each injured person's injury is automatically and quickly classified. This gives emergency personnel more time, improves the efficiency and accuracy of patient classification, ensures that limited medical resources are allocated rationally, and helps doctors make more accurate treatment decisions based on the specific data of the injured persons during the pre-hospital emergency stage.

[0089] It should be understood that the processor in the embodiments of the present invention can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0090] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0091] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0092] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0093] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0094] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0095] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0096] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0097] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0098] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0099] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0100] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] This invention provides a computer-readable storage medium storing a computer program thereon, characterized in that, when executed by a processor, the program implements the rapid damage classification method based on artificial intelligence as described in the method embodiment.

[0102] The present invention provides a computer-readable storage medium that can implement the steps and effects of the artificial intelligence-based rapid damage classification method in the above-described method embodiments. To avoid repetition, the present invention will not repeat the details.

[0103] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, physiological data and image data of each injured person at the scene are used as input. Through multilayer perceptron (MLP), efficient neural network model EfficientNet, and quantum neural network (QNN), the severity of each injured person's injury is automatically and quickly classified. This gives emergency personnel more time, improves the efficiency and accuracy of injured person classification, and ensures that limited medical resources are allocated rationally.

[0104] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0105] The following points need to be explained: (1) The accompanying drawings of the embodiments of the present invention only involve the structures involved in the embodiments of the present invention. Other structures can refer to the general design.

[0106] (2) For clarity, the thickness of layers or regions is enlarged or reduced in the drawings used to describe embodiments of the invention, i.e., these drawings are not drawn to scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being “above” or “below” another element, the element may be “directly” located “above” or “below” the other element or there may be intermediate elements.

[0107] (3) Where there is no conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0108] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An artificial intelligence-based rapid triage method for injuries, characterized by, The method comprises the following steps: S1: acquiring physiological data and image data of each wounded person on the scene; S2: using a multi-layer perception machine (MLP) to quickly analyze the physiological data and extract a physiological feature vector from the physiological data; S3: using an efficient neural network model (EfficientNet) to extract an image feature vector from the image data; S4: using a fully connected layer to splice the physiological feature vector and the image feature vector to form a fusion feature vector; S5: taking each data feature in the fusion feature vector as a separate input channel and inputting the data feature into a quantum neural network (QNN)-based rapid triage classification model; S6: using the quantum neural network-based rapid triage classification model to classify the emergency degree of each wounded person.

2. The artificial intelligence-based triage method according to claim 1, wherein, The physiological data includes wound data, bleeding data, activity state data, consciousness reaction data, breathing data, and / or neck blood vessel pulsation data. 3.The artificial intelligence-based quick triage classification method according to claim 1, wherein, The step S2 specifically comprises the following steps: S201: inputting the physiological data into an input layer of the multi-layer perception machine; S202: performing nonlinear transformation on the hidden state in the physiological data through an activation function in a hidden layer of the multi-layer perception machine; S203: summarizing the hidden state output by the hidden layer in an output layer of the multi-layer perception machine and outputting the physiological feature vector.

4. The artificial intelligence-based fast triage classification method according to claim 3, characterized in that, The hidden layer of the multi-layer perception machine comprises a first hidden layer, a second hidden layer, and a third hidden layer; The first hidden layer and the second hidden layer adopt a ReLU activation function; The third hidden layer adopts a Sigmoid activation function. 5.The artificial intelligence-based quick triage classification method according to claim 1, wherein, The efficient neural network model comprises an MBConv module; The MBConv module introduces an SENet unit of an attention mechanism, and the SENet unit adopts a Swish activation function; The efficient neural network model increases a recursive gate convolution structure.

6. The artificial intelligence-based fast triage classification method according to claim 1, wherein, The step S6 specifically comprises the following steps: S601: converting each feature value into a quantum bit; ； in, x i Indicates the first i A quantum bit with data characteristics t i Indicates the first i The eigenvalues ​​of the data features are given by cosine and sin. T Indicates the transpose operation; S602: using a phase rotation gate to aggregate each quantum bit and determining an aggregation rotation angle: ； in, θ j Indicates to the first j The aggregated rotation angle of the inputs to the quantum neurons in the hidden layer, where arg represents the phase angle. R Indicates a phase rotation operation. θ ij Indicates to the first j When the input of the first hidden layer quantum neuron is... i The rotation angle of each data feature, where argtan represents the arctangent function. n Indicates the total number of data features; S603: using a controlled rotation gate to perform an inverse rotation operation based on the aggregation rotation angle and determining a hidden state: ； wherein, h j denotes the hidden state of the j th hidden layer quantum neuron, π denotes the mathematical constant pi, γ denotes a range control parameter, σ denotes an activation function; S604: classifying the emergency degree of each wounded person based on the hidden state. ； in, P k Indicates belonging to the first k The probability of each category of injury urgency. g This represents the activation function. w jk Indicates the first j The hidden layer quantum neurons and their outputs k The weights between the output neurons of each category classification result m This represents the total number of quantum neurons in the hidden layer.

7. The artificial intelligence-based fast triage classification method according to claim 1, wherein, The emergency degree of the wounded person includes a critical wounded person category, a severely wounded person category, a slightly wounded person category, and a death category.

8. The artificial intelligence-based fast triage classification method according to claim 1, wherein, The sea gull optimization algorithm is used to optimize the model parameters of the multi-layer perception machine, the efficient neural network model, and the quantum neural network (QNN)-based rapid triage classification model. 9.The artificial intelligence-based fast triage method according to claim 1, wherein, The method further comprises the following step: S7: issuing a prompt message according to the emergency degree of each wounded person.

10. An artificial intelligence-based rapid triage system for injuries, characterized in that, The method comprises the following steps: a processor; a memory having computer readable instructions stored thereon, wherein the computer readable instructions are executed by the processor to implement the artificial intelligence-based rapid triage classification method according to any one of claims 1 to 9.

Images