Medical staff emergency capability assessment method and system based on deep neural network
By employing a deep neural network-based assessment method, combined with multi-layer neural networks and natural language processing technology, the accuracy and resource consumption issues of existing technologies for assessing the emergency response capabilities of medical personnel have been resolved, enabling the provision of highly accurate and timely emergency response capability assessments and training programs.
Patent Information
- Application Number
- CN202311152192.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-07
- Publication Date
- 2026-04-14
AI Technical Summary
Existing methods for assessing the emergency response capabilities of medical and nursing personnel cannot guarantee the accuracy of the assessment results, consume a lot of resources, and cannot identify weaknesses in emergency response capabilities in a timely manner. They also cannot be conducted routinely, which leads to a failure to improve emergency response capabilities in a timely manner.
An assessment method based on deep neural networks is adopted. By collecting personality information, basic information and emergency response knowledge description information of medical staff, multiple feature vectors are constructed using multi-layer neural networks and natural language processing technology. Combined with TextCNN and GRU attention mechanism text classification model, the final emergency response capability assessment result is output.
It achieves highly accurate and timely emergency response capability assessment, improves the accuracy of the algorithm by more than 12%, and can identify weaknesses in emergency response capabilities and provide targeted training programs.
Smart Images
Figure CN121862335A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of artificial intelligence technology, specifically relating to a method and system for assessing the emergency response capabilities of medical personnel based on deep neural networks. Background Technology
[0002] Given the continuous emergence of public health emergencies, the ability of hospitals at all levels to respond to and handle such emergencies is a crucial guarantee for social harmony and stability and national public safety. The emergency response capabilities of medical staff are a vital foundation for a hospital's ability to respond to and handle public health emergencies.
[0003] Emergency room environments are crowded and noisy, with high work intensity and a fast pace, placing extremely high demands on the psychological qualities and emergency response capabilities of emergency medical personnel. Therefore, it is of great significance to assess the emergency response capabilities of medical personnel in daily work and to provide training to address any deficiencies and improve their emergency response capabilities.
[0004] In the process of realizing this invention, the inventors discovered that the prior art has at least the following problems: traditional emergency response capability evaluation methods include filling out questionnaires for self-assessment of emergency response capabilities, assessment of emergency theoretical knowledge, implementation of emergency plan drills, etc. These methods either cannot guarantee the accuracy of the evaluation results, or cannot reflect the true emergency response capability, or require a lot of resources and time. Because they cannot be carried out on a regular basis, they cannot detect weaknesses in emergency response capabilities in a timely manner. Summary of the Invention
[0005] The present invention aims to at least partially solve the aforementioned technical problems. Therefore, the objective of this invention is to provide a method and system for assessing the emergency response capabilities of medical personnel based on deep neural networks.
[0006] The technical solution adopted in this invention is as follows:
[0007] A method for assessing the emergency response capabilities of healthcare workers based on deep neural networks, comprising the following steps:
[0008] Collect personality information, basic information, and descriptions of emergency response capabilities of medical personnel; basic information includes general basic information and basic information on emergency response.
[0009] The first neural network model is constructed using personality information as input. After passing through two layers of neural networks and nonlinear transformation, the first feature vector is output.
[0010] A second neural network model is constructed using general basic information as input. After passing through two layers of neural networks and nonlinear transformation, the second feature vector is output.
[0011] Using basic information about emergency response as input, a third neural network model is constructed. After passing through two layers of neural networks and nonlinear transformation, the third feature vector is output.
[0012] The second and third feature vectors are concatenated to form the fourth feature vector, and the fourth feature vector is concatenated to form the first feature vector to form the fifth feature vector;
[0013] The fifth feature vector is used as input to construct the fourth neural network model, obtain the parameters of the impact of different types of features on emergency response capabilities, and output the sixth feature vector.
[0014] The text classification model, based on the TextCNN attention mechanism and using word vectors to describe information relevant to emergency work, outputs a seventh feature vector.
[0015] The text classification model, based on the GRU attention mechanism and using word vectors to describe information related to emergency response, outputs the eighth feature vector.
[0016] The seventh and eighth eigenvectors are concatenated to form the ninth eigenvector;
[0017] The sixth and ninth eigenvectors are concatenated to form the tenth eigenvector;
[0018] The fifth neural network model is constructed using the tenth feature vector as input, and after nonlinear transformation, the eleventh feature vector is output.
[0019] The eleventh feature vector is classified using the softmax function to obtain the emergency response capability assessment result.
[0020] Preferably, the general basic information of the medical staff includes the medical staff's gender, marital status, professional title, education, length of service, age, and number of children; the basic information for emergency work includes nursing years, reasons for work, and willingness to be transferred.
[0021] Preferably, the emergency work awareness description information includes a description in natural language of the importance, intensity, characteristics, requirements, and capabilities of emergency work.
[0022] Preferably, the lengths of the first and third feature vectors are 4, the length of the second feature vector is 8, the length of the fourth feature vector is 12, the length of the fifth feature vector is 16, the length of the ninth feature vector is 1024, and the lengths of the tenth and eleventh feature vectors are 80.
[0023] Preferably, the emergency response capability assessment results include three categories: excellent, good, and poor.
[0024] A deep neural network-based emergency response assessment system for healthcare workers includes:
[0025] The data collection module is used to collect personality information, basic information, and descriptions of emergency response capabilities of medical personnel; the basic information includes general basic information and emergency response basic information.
[0026] The first neural network model uses personality information as input to construct and output the first feature vector;
[0027] The second neural network model constructs and outputs a second feature vector using general basic information as input.
[0028] The third neural network model uses basic information about emergency work as input to construct and output a third feature vector.
[0029] The fourth neural network model constructs and outputs a sixth feature vector using the fifth feature vector as input; the fifth feature vector is formed by concatenating the third and first feature vectors; the fourth feature vector is formed by concatenating the second and third feature vectors.
[0030] The TextCNN attention mechanism-based text classification model takes word vectors as input to describe emergency work information and outputs a seventh feature vector.
[0031] A text classification model based on the GRU attention mechanism takes word vectors as input to describe emergency work information and outputs an eighth feature vector.
[0032] The fifth neural network model constructs and outputs the eleventh feature vector using the tenth feature vector as input; the tenth feature vector is formed by concatenating the sixth and ninth feature vectors; the ninth feature vector is formed by concatenating the seventh and eighth feature vectors.
[0033] The processing module is used to classify the eleventh feature vector using the softmax function to obtain the emergency response capability assessment result.
[0034] This invention uses a deep neural network to train and model the personality information, basic information, and emergency work awareness description information of medical staff, and discovers the correlation between personality information, basic information, emergency work awareness description information and emergency capabilities, and obtains emergency capability assessment results. This can provide a scientific basis for identifying the weaknesses of medical staff's emergency capabilities and proposing targeted training programs.
[0035] The beneficial effects of this invention are as follows:
[0036] 1. This invention uses deep neural networks to conveniently and promptly assess the emergency response capabilities of medical personnel with high accuracy, overcoming the weaknesses of traditional methods such as poor accuracy, high resource consumption, untimely assessment, and lack of objectivity. It comprehensively analyzes multiple aspects, including the personality, basic information, and understanding of emergency work of medical personnel, thereby enabling a more accurate assessment of their emergency response capabilities.
[0037] 2. This invention first models the collected information separately, extracting key features that affect the emergency response capabilities of medical personnel from each type of information. Then, based on the relationships between the information, it progressively assembles the data and extracts a higher level of key features. This approach can fully extract the features that have the greatest impact on emergency response capabilities from each type of information, and also analyze the relationships between different types of features to extract common features. Practical experience shows that this method can improve the accuracy of the algorithm by more than 12%.
[0038] 3. This invention utilizes text classification models based on TextCNN attention mechanisms and GRU attention mechanisms, which fully leverage the advantages of TextCNN, GRU, and attention mechanisms. It can extract local key information, extract contextual semantic information, and focus attention on information that has a significant impact on the final result, thereby greatly improving the text classification effect. Attached Figure Description
[0039] Figure 1 This is a flowchart of part of the method for assessing the emergency response capabilities of medical personnel based on deep neural networks, which is part of the present invention.
[0040] Figure 2 This is another flowchart of the method for assessing the emergency response capabilities of medical personnel based on deep neural networks, which is another part of the present invention.
[0041] Figure 3 This is another flowchart of the method for assessing the emergency response capabilities of medical personnel based on deep neural networks, which is another part of the present invention. Detailed Implementation
[0042] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0043] It should be understood that, and also noted, in the embodiments, the functions / actions may appear in a different order than those shown in the figures. For example, depending on the functions / actions involved, they may actually be performed substantially concurrently, or sometimes the two figures shown consecutively may be performed in reverse order.
[0044] Emergency response capability of medical personnel refers to their ability to keenly observe changes in the vital signs of those being rescued, analyze and judge them, and calmly and decisively cooperate in rescue and nursing care using skilled techniques during emergency response.
[0045] The emergency response capability evaluation includes two aspects: patient observation and emergency cooperation. The patient observation capability assessment indicators are broken down into four aspects: mastery of key observation points, correct observation methods, accurate patient assessment, and timely reporting and handling of problems. The emergency cooperation capability assessment indicators are broken down into four aspects: rapid and standardized operation and proceduralized rescue, proactive and swift cooperation, adequate preparation of first-aid supplies, and emotional stability during the rescue process. Each indicator is evaluated on a three-level scale: Excellent, Good, and Poor.
[0046] The personality traits referred to in this invention are one of the important factors influencing an individual's perception of stress. Under the same environment and when exposed to the same stressor, people with different personalities will perceive and cope differently, which will also affect their emergency response capabilities. This invention studies the relationship between personality and emergency response capability assessment indicators for medical personnel, thereby identifying which emergency response indicators a person with a given personality is most likely to have weaknesses in, and thus enabling targeted training.
[0047] This invention uses the Eysenck Personality Questionnaire to classify the personality of medical staff. The classification results include introversion / extroversion (E), neuroticism (N), psychoticism (P), and covertness (L), which serve as the personality information of the medical staff.
[0048] like Figures 1 to 3 As shown in this embodiment, a method for assessing the emergency response capabilities of medical personnel based on deep neural networks includes the following steps:
[0049] S1. Collect personality information, basic information, and descriptions of emergency response capabilities of medical staff.
[0050] Specifically, basic information was collected using a cross-sectional questionnaire survey method. The questionnaire included information on gender, marital status, professional title, education level, length of service, age, number of children, years of nursing experience, reasons for working in the emergency department, and willingness to transfer. Reasons for working in the emergency department were the reasons for coming to the emergency department, and willingness to transfer was the willingness to transfer out of the emergency department.
[0051] Among them, the gender, marital status, professional title, education, length of service, age and number of children of medical staff are used as general basic information, while nursing years, reasons for work and willingness to transfer are used as basic information for emergency work.
[0052] Emergency response awareness descriptions include descriptions of the importance, intensity, characteristics, requirements, and capabilities of emergency work in natural language.
[0053] S2. The personality, general basic information, and emergency work basic information of medical staff all affect their emergency response capabilities, but the aspects and degrees of influence differ. Therefore, this invention first establishes neural network models for each factor separately, and then constructs a further deep neural network model by concatenating their outputs.
[0054] Specifically, the first neural network model is constructed using personality information (input_1) as input. After two layers of neural network and nonlinear transformation, the first feature vector (dense_2) with a length of 4 is output. After training, the first feature vector will reflect the main feature information of the influence of medical staff's personality on their emergency response capabilities.
[0055] The second neural network model is constructed using general basic information (input_2) as input. After two layers of neural network and nonlinear transformation, the output is a second feature vector (dense_4) with a length of 8.
[0056] Using basic emergency work information (input_3) as input, a third neural network model is constructed. After two layers of neural networks and nonlinear transformation, the output is a third feature vector (dense_6) with a length of 4.
[0057] After training, the second and third feature vectors will respectively reflect the main feature information of the impact of general basic information and emergency work basic information on the emergency response capabilities of medical staff.
[0058] The second and third feature vectors are concatenated to form a fourth feature vector (concatenate_1) of length 12. The fourth feature vector contains all the feature information that affects the emergency response capabilities of medical staff, including general basic information and emergency work basic information.
[0059] The fourth and first feature vectors are concatenated to form a fifth feature vector of length 16 (concatenate_2). The fifth feature vector contains all the feature information about the impact of medical staff's personality and basic information on their emergency response capabilities.
[0060] The fourth neural network model was constructed using the fifth feature vector as input. The feature vectors of the impact of these three types of information (personality information, general basic information, and emergency work basic information) on the emergency response capabilities of medical staff were trained to obtain the parameters of the impact of different types of features on emergency response capabilities, and a sixth feature vector (dense_8) of length 16 was output. The feature vectors of the impact of the three types of information on emergency response capabilities were further optimized after training.
[0061] S3. The description of emergency response awareness largely reflects medical staff's attitude towards emergency work and the extent of their knowledge of emergency response, thus also reflecting their emergency response capabilities to some extent. This invention uses natural language processing technology to establish a deep neural network model to analyze medical staff's descriptions of emergency response awareness and to explore their emergency response capabilities.
[0062] The text classification model, based on the TextCNN attention mechanism, outputs a seventh feature vector (including four vectors of length 128) by inputting word vectors describing emergency response information.
[0063] The text classification model based on the GRU attention mechanism outputs the eighth feature vector (including four vectors of length 128) by inputting word vectors describing information about emergency work.
[0064] The seventh and eighth feature vectors are concatenated to form a ninth feature vector (concatenate_4) with a length of 1024. The ninth feature vector can be regarded as two sets of feature information reflecting the impact of emergency response work understanding description on emergency response capabilities.
[0065] By setting convolutional kernels of various sizes for TextCNN, local key information can be effectively extracted. GRU can handle long texts more effectively and extract contextual semantic information better. Their respective attention mechanisms, by setting different weights, can focus attention on words that have a greater impact on the final result. Therefore, combining the characteristics of TextCNN and GRU, along with the attention mechanism, can fully utilize the most valuable information in the text and improve text classification performance.
[0066] S4. Concatenate the sixth and ninth feature vectors to form a tenth feature vector of length 80 (concatenate_5); the tenth feature vector contains the feature information that has the most influence on emergency response capability, which is obtained through training multiple models separately and concatenating the training.
[0067] The fifth neural network model is constructed using the tenth feature vector as input. After nonlinear transformation, the feature information is further abstracted and enhanced, and the output is the eleventh feature vector (dense_9) with a length of 80. The eleventh feature vector is the feature that has the greatest influence on emergency response capability, which is automatically proposed by the algorithm after training the deep neural network.
[0068] The eleventh feature vector was classified using the softmax function to obtain the emergency response capability assessment results, which were categorized into three levels: excellent, good, and poor. These results provide a scientific basis for identifying deficiencies in the emergency response capabilities of medical personnel and for developing targeted training programs.
[0069] This embodiment also provides a medical staff emergency response capability assessment system based on deep neural networks. The emergency response capability assessment method described above includes a data acquisition module, a first neural network model, a second neural network model, a third neural network model, a fourth neural network model, a text classification model based on the TextCNN attention mechanism, a text classification model based on the GRU attention mechanism, a fifth neural network model, and a processing module.
[0070] The data collection module is used to collect personality information, basic information, and emergency response awareness descriptions of medical staff; the basic information includes general basic information and emergency response basic information.
[0071] The first neural network model uses personality information as input to construct and output a first feature vector; the second neural network model uses general basic information as input to construct and output a second feature vector; the third neural network model uses basic emergency work information as input to construct and output a third feature vector; and the fourth neural network model uses a fifth feature vector as input to construct and output a sixth feature vector.
[0072] The TextCNN attention-based text classification model takes word vectors describing emergency work information as input and outputs the seventh feature vector; the GRU attention-based text classification model takes word vectors describing emergency work information as input and outputs the eighth feature vector.
[0073] The fifth neural network model constructs an eleventh feature vector from the tenth feature vector as input. The processing module uses the softmax function to classify the eleventh feature vector to obtain the emergency response capability assessment result.
[0074] The fifth eigenvector is formed by concatenating the third and first eigenvectors; the fourth eigenvector is formed by concatenating the second and third eigenvectors; the tenth eigenvector is formed by concatenating the sixth and ninth eigenvectors; and the ninth eigenvector is formed by concatenating the seventh and eighth eigenvectors.
[0075] This invention is not limited to the above-described optional embodiments. Anyone can derive other various forms of products under the guidance of this invention. However, regardless of any changes made in their shape or structure, any technical solution that falls within the scope of the claims of this invention shall be protected by this invention.
Claims
1. A method for assessing the emergency response capabilities of medical personnel based on deep neural networks, characterized in that, Including the following steps: Collect personality information, basic information, and descriptions of emergency response capabilities of medical personnel; basic information includes general basic information and basic information on emergency response. The first neural network model is constructed using personality information as input. After passing through two layers of neural networks and nonlinear transformation, the first feature vector is output. A second neural network model is constructed using general basic information as input. After passing through two layers of neural networks and nonlinear transformation, the second feature vector is output. Using basic information about emergency response as input, a third neural network model is constructed. After passing through two layers of neural networks and nonlinear transformation, the third feature vector is output. The second and third feature vectors are concatenated to form the fourth feature vector, and the fourth feature vector is concatenated to form the first feature vector to form the fifth feature vector; The fifth feature vector is used as input to construct the fourth neural network model, obtain the parameters of the impact of different types of features on emergency response capabilities, and output the sixth feature vector. The text classification model, based on the TextCNN attention mechanism and using word vectors to describe information relevant to emergency work, outputs a seventh feature vector. The text classification model, based on the GRU attention mechanism and using word vectors to describe information related to emergency response, outputs the eighth feature vector. The seventh and eighth eigenvectors are concatenated to form the ninth eigenvector; The sixth and ninth eigenvectors are concatenated to form the tenth eigenvector; The fifth neural network model is constructed using the tenth feature vector as input, and after nonlinear transformation, the eleventh feature vector is output. The eleventh feature vector is classified using the softmax function to obtain the emergency response capability assessment result.
2. The method for assessing the emergency response capabilities of medical personnel according to claim 1, characterized in that: The general basic information of the medical staff includes their gender, marital status, professional title, education, length of service, age, and number of children; the basic information for emergency work includes years of nursing experience, reasons for working, and willingness to be transferred.
3. The method for assessing the emergency response capabilities of medical personnel according to claim 1, characterized in that: The description of emergency work awareness includes a description in natural language of the importance, intensity, characteristics, requirements, and capabilities of emergency work.
4. The method for assessing the emergency response capabilities of medical personnel according to claim 1, characterized in that: The lengths of the first and third feature vectors are 4, the length of the second feature vector is 8, the length of the fourth feature vector is 12, the length of the fifth feature vector is 16, the length of the ninth feature vector is 1024, and the lengths of the tenth and eleventh feature vectors are 80.
5. The method for assessing the emergency response capabilities of medical personnel according to claim 1, characterized in that: The emergency response capability assessment results are categorized into three levels: excellent, good, and poor.
6. A system for assessing the emergency response capabilities of medical personnel based on deep neural networks, characterized in that, include: The data collection module is used to collect personality information, basic information, and descriptions of emergency response capabilities of medical staff. Basic information includes general basic information and emergency response basic information; The first neural network model uses personality information as input to construct and output the first feature vector; The second neural network model constructs and outputs a second feature vector using general basic information as input. The third neural network model uses basic information about emergency work as input to construct and output a third feature vector. The fourth neural network model uses the fifth feature vector as input to construct and output the sixth feature vector; The fifth eigenvector is formed by concatenating the third eigenvector and the first eigenvector; The fourth eigenvector is formed by concatenating the second and third eigenvectors; The TextCNN attention mechanism-based text classification model takes word vectors as input to describe emergency work information and outputs a seventh feature vector. A text classification model based on the GRU attention mechanism takes word vectors as input to describe emergency work information and outputs an eighth feature vector. The fifth neural network model uses the tenth feature vector as input to construct and output the eleventh feature vector; The tenth eigenvector is formed by concatenating the sixth and ninth eigenvectors; The ninth eigenvector is formed by concatenating the seventh and eighth eigenvectors; The processing module is used to classify the eleventh feature vector using the softmax function to obtain the emergency response capability assessment result.
7. The emergency response assessment system for medical personnel according to claim 6, characterized in that: The general basic information of the medical staff includes their gender, marital status, professional title, education, length of service, age, and number of children; the basic information for emergency work includes years of nursing experience, reasons for working, and willingness to be transferred.
8. The emergency response assessment system for medical personnel according to claim 6, characterized in that: The description of emergency work awareness includes a description in natural language of the importance, intensity, characteristics, requirements, and capabilities of emergency work.
9. The emergency response assessment system for medical personnel according to claim 6, characterized in that: The lengths of the first and third feature vectors are 4, the length of the second feature vector is 8, the length of the fourth feature vector is 12, the length of the fifth feature vector is 16, the length of the ninth feature vector is 1024, and the lengths of the tenth and eleventh feature vectors are 80.
10. The emergency response assessment system for medical personnel according to claim 6, characterized in that: The emergency response capability assessment results are categorized into three levels: excellent, good, and poor.