Difficulty airway evaluation method based on large audio and video feature model
Through a large-scale model evaluation method based on audio and video features, combined with the multidimensionality of audio and video data, the problem of inaccurate evaluation in the prior art is solved, and more accurate and difficult airway evaluation and safety improvement of tracheal intubation are achieved. .
Patent Information
- Application Number
- CN202411087793.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-09
AI Technical Summary
The existing difficult airway evaluation method has a single dimension, which fails to fully utilize the multidimensionality of audio and video data, resulting in inaccurate evaluation and safety risks.
Using a large model based on audio and video features, the airway state characteristics are collected in real time, and data transmission and processing are used to communicate with high-speed and low-delay data networks, and the large model is constructed, and the large model is trained and iteratively optimized, combining audio and The spatial and temporal mixing sequences of video feature data are evaluated.
A more accurate difficult airway assessment is achieved, which improves the safety and success rate of tracheal intubation, reduces the volatility and artificial error of the evaluation results, and provides multi-parameter evaluation ability.
Smart Images

Figure CN119008005A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of airway assessment, and in particular to a difficult airway assessment method based on a large model of audio and video features. Background Art
[0002] The assessment of difficult airway plays an important role in the analysis of intubation feasibility and safety of intubation. If the assessment of difficult airway is inaccurate, there will be safety risks. Existing difficult airway assessment methods include those based on subjective experience assessment, audio analysis assessment, and video analysis assessment, but their assessment dimensions are relatively single and do not fully utilize the multidimensionality of data.
[0003] With the development of artificial intelligence technology, large models have shown powerful learning capabilities. Therefore, the present invention proposes a difficult airway assessment method based on a large model of audio and video features, which fully utilizes the learning capabilities of the large model and the multidimensionality of audio and video data. Summary of the invention
[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.
[0005] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a difficult airway assessment method based on a large model of audio and video features to solve the problems raised in the background technology.
[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a difficult airway assessment method based on a large model of audio and video features, comprising:
[0007] The difficult airway status characteristics are collected in real time through the airway audio collection system and the airway video collection system, and the collected difficult airway status characteristics are transmitted and processed using a high-speed, low-latency data network communication method;
[0008] Constructing a difficult airway big model through a big model server, receiving difficult airway state characteristics transmitted by a high-speed and low-latency data network communication method, performing current big model training on the difficult airway state characteristics based on the difficult airway big model, and obtaining a prediction evaluation of the difficult airway big model;
[0009] Based on the prediction and evaluation, a large model of difficult airway is established to assist decision making, and the large model of difficult airway is iteratively optimized through a real-time back propagation mechanism.
[0010] As a preferred solution of the difficult airway assessment method based on the large audio and video feature model described in the present invention, the difficult airway state features are collected in real time by an airway audio acquisition system and an airway video acquisition system, and the collected difficult airway state features are transmitted and processed by a high-speed and low-latency data network communication method, including:
[0011] The processing method is to perform normalization processing and out-of-threshold data elimination processing on the data, and to perform large model training data sequence processing of a fusion of a spatial mixed sequence of difficult airway audio and video feature data and a temporal mixed sequence of difficult airway audio and video feature data;
[0012] The difficult airway state characteristics at least include an audio characteristic data sequence of the difficult airway and a video characteristic data sequence of the difficult airway.
[0013] As a preferred solution of the difficult airway assessment method based on the large audio and video feature model of the present invention, the large model training data sequence processing of the fusion of the spatial mixed sequence of the difficult airway audio and video feature data and the temporal mixed sequence of the difficult airway audio and video feature data includes:
[0014] Calculate spatially mixed sequences of spatially mixed sequences of difficult airway audio and video feature data;
[0015] Calculate the fusion values of the time mixed sequence of difficult airway audio and video feature data, the space mixed sequence of difficult airway audio and video feature data, and the time mixed sequence of difficult airway audio and video feature data.
[0016] As a preferred embodiment of the difficult airway assessment method based on the large audio and video feature model of the present invention, it also includes:
[0017] Calculate the fusion value adjustment coefficient of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data;
[0018] Calculate the fused large model training data sequence of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data.
[0019] As a preferred solution of the difficult airway assessment method based on the large model of audio and video features described in the present invention, wherein: a large model of difficult airway is constructed by a large model server, and difficult airway state features transmitted by a high-speed and low-latency data network communication method are received, including:
[0020] The large model server constructs the difficult airway large model based on the difficult airway state features collected in the past and the difficult airway state features collected currently;
[0021] Among them, the difficult airway large model adopts the open source large model framework for private and localized deployment.
[0022] As a preferred embodiment of the difficult airway assessment method based on the audio and video feature large model of the present invention, wherein: based on the difficult airway large model, the difficult airway state feature is trained with a large model at the current moment to obtain a prediction assessment of the difficult airway large model, including:
[0023] The difficult airway large model iterates the training accuracy of the difficult airway large model by comparing the actually collected difficult airway state characteristics with the difficult airway large model prediction evaluation results.
[0024] As a preferred embodiment of the difficult airway assessment method based on the large audio and video feature model of the present invention, it also includes:
[0025] The difficult airway large model uses a localized large model neural network to calculate the changing trend of the difficult airway data sequence characteristics, and evaluates the intubation feasibility and intubation safety of the actual difficult airway according to preset rules.
[0026] As a preferred solution of the difficult airway assessment method based on the large audio and video feature model of the present invention, wherein: according to the predicted assessment, a large difficult airway model is established to assist decision making, including:
[0027] Adaptive particle swarm optimization decision making as a decision aid for large models of difficult airways;
[0028] The adaptive particle swarm optimization decision-making obtains the operational safety score of the difficult airway according to the intubability and intubation safety of the actual difficult airway and the prediction evaluation results of the difficult airway large model.
[0029] Compared with the prior art, the invention has the following beneficial effects:
[0030] 1. The present invention introduces a fusion value adjustment coefficient of a spatial mixed sequence of difficult airway audio and video feature data and a temporal mixed sequence of difficult airway audio and video feature data, and calculates the fusion value adjustment coefficient of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data through the fusion value of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data. By adjusting the data sequence, abnormal fluctuations in the fusion value of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data can be achieved, thereby reducing the impact on the fusion large model training data sequence of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data;
[0031] 2. The present invention comprehensively considers the audio feature data sequence of the difficult airway and the video feature data sequence of the difficult airway, and can realize multi-parameter evaluation of the difficult airway; at the same time, it calculates the spatial mixed sequence of the audio and video feature data of the difficult airway at the current sampling moment, and calculates the temporal mixed sequence of the audio and video feature data of the difficult airway at the current sampling moment, thereby realizing the calculation of the fusion value of the spatial mixed sequence of the audio and video feature data of the difficult airway and the temporal mixed sequence of the audio and video feature data of the difficult airway in the actual dimension of the multi-parameter space, so that the fusion large model of the spatial mixed sequence of the audio and video feature data of the difficult airway and the temporal mixed sequence of the audio and video feature data of the difficult airway can fully consider the time and space factor scales during the training process. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:
[0033] Figure 1 This is an overall flow chart of a difficult airway assessment method based on a large model of audio and video features according to an embodiment of the present invention;
[0034] Figure 2 A flowchart of a fusion big model training data sequence processing of a difficult airway audio and video feature data spatial mixed sequence and a difficult airway audio and video feature data temporal mixed sequence of a difficult airway assessment method based on a big audio and video feature model according to an embodiment of the present invention;
[0035] Figure 3 This is a comparison diagram before and after the fusion of audio and video data of a difficult airway assessment method based on a large model of audio and video features described in an embodiment of the present invention. DETAILED DESCRIPTION
[0036] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0037] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0038] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0039] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.
[0040] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0041] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0042] Example 1
[0043] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a difficult airway assessment method based on a large model of audio and video features, comprising:
[0044] S1. Real-time collection of difficult airway status characteristics through the airway audio collection system and the airway video collection system, and transmission and processing of the collected difficult airway status characteristics using a high-speed, low-latency data network communication method;
[0045] Further, the processing method is to perform normalization processing and out-of-threshold data elimination processing on the data, and to perform large model training data sequence processing of the fusion of the spatial mixed sequence of the difficult airway audio and video feature data and the temporal mixed sequence of the difficult airway audio and video feature data;
[0046] Furthermore, the difficult airway state feature includes at least an audio feature data sequence of the difficult airway and a video feature data sequence of the difficult airway;
[0047] Specifically, the steps for processing the training data sequence of the fusion model of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data are as follows:
[0048] S101, calculating a spatial mixed sequence of a spatial mixed sequence of audio and video feature data of a difficult airway;
[0049] S102, calculating the fusion value of the time mixed sequence of the difficult airway audio and video feature data, the space mixed sequence of the difficult airway audio and video feature data, and the time mixed sequence of the difficult airway audio and video feature data;
[0050] S103, calculating a fusion value adjustment coefficient of a spatial mixed sequence of difficult airway audio and video feature data and a temporal mixed sequence of difficult airway audio and video feature data;
[0051] S104, calculating a fused large model training data sequence of a spatial mixed sequence of difficult airway audio and video feature data and a temporal mixed sequence of difficult airway audio and video feature data;
[0052] Specifically, the spatial mixed sequence formula for calculating the spatial mixed sequence of difficult airway audio and video feature data is expressed as:
[0053]
[0054] Among them, d t is the spatial mixed sequence of audio and video feature data of difficult airway at the current sampling moment; s is the audio feature data sequence of difficult airway; v is the video feature data sequence of difficult airway; n is the length of feature data sequence; a1 is the balance coefficient of audio feature data sequence of difficult airway; a2 is the balance coefficient of video feature data sequence of difficult airway; β1 is the harmonic strength of audio feature data sequence of difficult airway; β2 is the harmonic strength of video feature data sequence of difficult airway;
[0055] Furthermore, the constraint condition of the spatial mixed sequence formula for calculating the spatial mixed sequence of the difficult airway audio and video feature data is expressed as:
[0056] a1+a2=1
[0057] β1+β2<a1s+a2v
[0058] Specifically, the formula for calculating the time mixed sequence of difficult airway audio and video feature data is expressed as:
[0059]
[0060] Among them, g t is the time-mixed sequence of audio and video feature data of the difficult airway at the current sampling moment; s(t-nΔt) is the audio feature data sequence of the difficult airway with a translation of Δt; v(t-nΔt) is the video feature data sequence of the difficult airway with a translation of Δt; ω n is the weight value of the n-th Δt difficult airway characteristic data sequence;
[0061] Specifically, the formula for calculating the fusion value of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data is expressed as:
[0062]
[0063] Among them, X t is a fusion value of a spatial mixed sequence of difficult airway audio and video feature data and a temporal mixed sequence of difficult airway audio and video feature data;
[0064] Specifically, the formula for calculating the fusion value adjustment coefficient of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data is expressed as:
[0065]
[0066] Among them, γ s is the variance of the audio feature data sequence of difficult airway; γ v is the variance of the video feature data sequence of difficult airway; s t The fusion value adjustment coefficient of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data; w t is the fusion large model training data sequence of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data; k is the fusion value adjustment coefficient scaling factor of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data, and its value range is 0-1; η is the fusion large model training data sequence scaling factor of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data, and its value range is 0-1;
[0067] It should be noted that s t With X t For related relationships, As a fusion large model training data sequence of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data, The value of the variable s t In the range of (0,0.25), it increases and decreases, and X t Considering γ s The variance of the audio feature data sequence of difficult airway, X t Considering γ v The variance of the video feature data sequence of difficult airway, when their values increase, Can be decreasing;
[0068] S2. construct a difficult airway big model through a big model server, receive difficult airway state characteristics transmitted by a high-speed, low-latency data network communication method, perform big model training on the difficult airway state characteristics at the current moment based on the difficult airway big model, and obtain a prediction evaluation of the difficult airway big model;
[0069] Further, the large model server constructs the difficult airway large model based on the difficult airway state features collected in the past and the difficult airway state features collected currently;
[0070] Among them, the difficult airway large model uses an open source large model framework for private local deployment;
[0071] It should be noted that private local deployment can ensure that data is processed locally, avoid transmitting sensitive medical data to external servers, reduce the risk of data leakage, and protect patient privacy;
[0072] Furthermore, the difficult airway model iterates the training accuracy of the difficult airway model by comparing the actually collected difficult airway status features with the predicted evaluation results of the difficult airway model.
[0073] It should be noted that the iterative training process reduces the need for human intervention, reduces the workload of manual adjustments and possible human errors, improves the automation and work efficiency of the overall system, and can quickly respond to clinical needs, especially in emergency situations, ensuring that the model can be improved and optimized in a very short time;
[0074] Furthermore, the difficult airway large model uses a localized large model neural network to calculate the changing trend of the difficult airway data sequence characteristics, and evaluates the intubability and safety of the actual difficult airway according to preset rules;
[0075] Specifically, the evaluation of preset rules can realize personalized tracheal intubation for each patient according to the patient's own difficult airway status, further improving the success rate and safety of tracheal intubation operations;
[0076] S3. Based on the prediction and evaluation, a large model of difficult airway is established to assist decision making, and the large model of difficult airway is iteratively optimized through a real-time back propagation mechanism;
[0077] Furthermore, adaptive particle swarm optimization was used as decision-making aid for the difficult airway large model;
[0078] Furthermore, the adaptive particle swarm optimization decision-making obtains the operational safety score of the difficult airway based on the actual difficult airway intubation capability and intubation safety and the predicted evaluation results of the difficult airway large model;
[0079] It should be noted that adaptive particle swarm optimization decision-making can reduce the complications caused by tracheal intubation on the basis of the success rate and safety of tracheal intubation operations; and by obtaining the operational safety score of difficult airway, it can help doctors of different qualifications better understand and judge the specific situation of difficult airway.
[0080] Example 2
[0081] Reference Figure 3 , which is the second embodiment of the present invention, and this embodiment provides a difficult airway assessment method based on a large audio and video feature model, including: In this embodiment, through a simulation experiment, a difficult airway assessment method based on a large audio and video feature model is used to assess the intubability and safety of actual patients;
[0082] The equipment used in the simulation experiment includes an airway audio acquisition system, an airway video acquisition system, and a transmission processing device for high-speed, low-latency data network communication. The large model server uses an open source large model framework for private local deployment, and uses a localized large model neural network to calculate the changing trend of difficult airway data sequence characteristics.
[0083] Ten patients with difficult airways were selected for data collection. Before endotracheal intubation, each patient's airway status feature data, including audio feature data sequence and video feature data sequence, were collected in real time using the airway audio collection system and airway video collection system. First, the collected audio and video data were normalized to eliminate individual differences between different patients. Then, the data were subjected to out-of-threshold data removal to remove noise and outliers. The processed data were used to calculate spatial mixed sequences and temporal mixed sequences, respectively, and fused into a large model training data sequence. Secondly, the large model service was used to train the large model training data sequence. The server trains the large model. The server uses a neural network to train the data based on historical collected data and current real-time data. During the training process, the model parameters are optimized through a real-time back propagation mechanism to continuously improve the prediction accuracy of the model. Finally, after the model training is completed, the prediction and evaluation results of the difficult airway large model are obtained. Based on the prediction and evaluation results, the difficult airway large model evaluates the intubability and safety of the actual difficult airway through adaptive particle swarm optimization (APSO) decision-making. APSO is used to calculate the operation safety score of each patient to provide auxiliary decision support for clinicians.
[0084] The method of the present invention was compared with the traditional subjective experience evaluation method, the single audio analysis evaluation method and the single video analysis evaluation method, and the performance of each method in terms of the success rate and safety of endotracheal intubation was evaluated. The simulated experimental results are shown in Tables 1 and 2.
[0085] Table 1 The method of the present invention
[0086]
[0087] Table 2 Traditional methods
[0088]
[0089] From Table 1 and Table 2, it can be seen that the operational safety scores of the method of the present invention are 95, 90, 92, 97, 91, 86, 89, 93, 90 and 87, respectively; compared with the traditional method (85, 80, 82, 87, 81, 76, 79, 83, 80 and 77, respectively), the single audio analysis evaluation method (88, 83, 85, 90, 84, 79, 82, 86, 83 and 80, respectively) and the single video analysis evaluation method (90, 85, 87, 92, 86, 81, 84, 88, 85 and 82, respectively), the method of the present invention shows obvious advantages in the scores of each patient;
[0090] And through Figure 3It can be seen that the reduction in the variance of the fusion value reflects the degree of improvement in the stability of the data sequence after adjustment. The smaller the variance, the smaller the data fluctuation and the better the stability of the system. This shows that after the data sequence is adjusted by the method of the present invention, the abnormal fluctuation is significantly reduced, and the stability and reliability of the model training data are improved.
[0091] By comparing the simulated experimental data, it can be found that the difficult airway assessment method based on the large model of audio and video features is significantly higher than the traditional method and the single-dimensional analysis method in terms of operation safety score; the traditional method relies on the subjective experience of doctors and single-dimensional data, which is easily affected by personal experience and data noise, and the evaluation results are highly subjective and uncertain; the method of the present invention combines audio and video feature data, and through comprehensive analysis of multi-dimensional data and adaptive particle swarm optimization decision-making, it can provide more comprehensive, scientific and accurate evaluation results.
[0092] Those skilled in the art will appreciate that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program codes. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.
[0093] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0094] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1A function specified in one or more boxes.
[0095] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0096] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.
[0097] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.
Claims
1. A difficult airway assessment method based on a large model of audio and video features, characterized in that: include: The difficult airway status characteristics are collected in real time through the airway audio collection system and the airway video collection system, and the collected difficult airway status characteristics are transmitted and processed using a high-speed, low-latency data network communication method; Constructing a difficult airway big model through a big model server, receiving difficult airway state characteristics transmitted by a high-speed and low-latency data network communication method, performing current big model training on the difficult airway state characteristics based on the difficult airway big model, and obtaining a prediction evaluation of the difficult airway big model; Based on the prediction and evaluation, a large model of difficult airway is established to assist decision making, and the large model of difficult airway is iteratively optimized through a real-time back propagation mechanism.
2. The difficult airway assessment method based on a large audio and video feature model as claimed in claim 1, characterized in that: The difficult airway status characteristics are collected in real time through the airway audio collection system and the airway video collection system, and the collected difficult airway status characteristics are transmitted and processed using a high-speed, low-latency data network communication method, including: The processing method is to perform normalization processing and out-of-threshold data elimination processing on the data, and to perform large model training data sequence processing of a fusion of a spatial mixed sequence of difficult airway audio and video feature data and a temporal mixed sequence of difficult airway audio and video feature data; The difficult airway state characteristics at least include an audio characteristic data sequence of the difficult airway and a video characteristic data sequence of the difficult airway.
3. The difficult airway assessment method based on a large audio and video feature model as claimed in claim 2, characterized in that: The processing of the fused large model training data sequence of the difficult airway audio and video feature data spatial mixed sequence and the difficult airway audio and video feature data temporal mixed sequence includes: Calculate spatially mixed sequences of spatially mixed sequences of difficult airway audio and video feature data; Calculate the fusion values of the time mixed sequence of difficult airway audio and video feature data, the space mixed sequence of difficult airway audio and video feature data, and the time mixed sequence of difficult airway audio and video feature data.
4. The difficult airway assessment method based on a large audio and video feature model as claimed in claim 3, characterized in that: Also includes: Calculate the fusion value adjustment coefficient of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data; Calculate the fused large model training data sequence of the spatial mixed sequence of difficult airway audio and video feature data and the temporal mixed sequence of difficult airway audio and video feature data.
5. The difficult airway assessment method based on a large audio and video feature model as claimed in claim 3 or 4, characterized in that: A large model of difficult airway is constructed through a large model server, and the difficult airway status characteristics transmitted by a high-speed and low-latency data network communication method are received, including: The large model server constructs the difficult airway large model based on the difficult airway state features collected in the past and the difficult airway state features collected currently; Among them, the difficult airway large model adopts the open source large model framework for private and localized deployment.
6. The difficult airway assessment method based on a large audio and video feature model as claimed in claim 5, characterized in that: Based on the difficult airway big model, the difficult airway state characteristics are trained on the big model at the current moment to obtain a prediction evaluation of the difficult airway big model, including: The difficult airway large model iterates the training accuracy of the difficult airway large model by comparing the actually collected difficult airway state characteristics with the difficult airway large model prediction evaluation results.
7. The difficult airway assessment method based on a large audio and video feature model as claimed in claim 6, characterized in that: Also includes: The difficult airway large model uses a localized large model neural network to calculate the changing trend of the difficult airway data sequence characteristics, and evaluates the intubation feasibility and intubation safety of the actual difficult airway according to preset rules.
8. The difficult airway assessment method based on a large audio and video feature model according to claim 6 or 7, characterized in that: Based on the above prediction and evaluation, a large model of difficult airway is established to assist decision making, including: Adaptive particle swarm optimization decision making as a decision aid for difficult airway large models; The adaptive particle swarm optimization decision-making obtains the operational safety score of the difficult airway according to the intubability and intubation safety of the actual difficult airway and the prediction evaluation results of the difficult airway large model.
Citation Information
Patent Citations
Neural network training method and device and electronic equipment
CN111930992A
Image classification method based on multilevel adaptive feature fusion class incremental learning
CN114612721A
MD patient emotion fluctuation monitoring and affective disorder state evaluation method and system
CN115517681A
Method and system for evaluating difficult airway
CN115602320A
Training method and system of audio and video speech recognition model, and electronic equipment
CN116013270A
Cited By
Identification method and system for Markov grading and storage medium
CN120511065A
