Multi-dimensional safety monitoring method and system for infant gastric lavage machine
Through the deep learning model, real-time monitoring of facial images of infants and young children is solved, and the problem of difficulty in detecting discomfort during gastric lavage in infants and young children is achieved. The accurate recognition and timely alarm of facial emotions of infants and young children is achieved, and the safety of gastric lavage is improved. .
Patent Information
- Application Number
- CN202411160719.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-12-17
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
During the gastric lavage process of infants and young children, it is difficult to detect physical discomfort in time, which affects the effect of gastric lavage and threatens life safety.
Using multi-dimensional security monitoring method, the facial images of infants and young children are collected, preprocessed and deep learning model training is performed, local and global features are extracted using convolutional neural networks and self-attention networks, and recognition capabilities are enhanced through feature fusion structures to monitor the face in real time. Notify medical staff in a timely manner.
Real-time and accurate monitoring and recognition of facial emotions of infants and young children, timely alarms are called, and the response speed of medical staff and the effectiveness of measures are improved.
Smart Images
Figure CN119152427A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of artificial foot detection equipment, and more specifically, relates to a multi-dimensional safety monitoring method and system for an infant gastric lavage machine. Background Art
[0002] Gastric lavage refers to the process of pouring a certain amount of liquid into the stomach cavity, mixing it with the stomach contents, and then extracting it. This is repeated many times. The purpose is to remove unabsorbed poisons in the stomach or clean the stomach cavity. It is used clinically for gastric surgery and pre-examination preparation. For acute poisoning, such as swallowing organic phosphorus, inorganic phosphorus, alkaloids, barbiturates, etc. in a short period of time, gastric lavage is an important rescue measure. Gastric lavage includes emetic gastric lavage and gastric tube gastric lavage. Gastric tube gastric lavage is to insert a gastric tube from the nasal cavity or oral cavity, through the esophagus to the stomach, first suck out the poison, then inject gastric lavage fluid, and discharge the stomach contents to achieve the purpose of eliminating the poison. In certain circumstances, infants and young children need to undergo gastric lavage. However, infants and young children are not physically mature, and it is difficult for them to express their emotions and thoughts during the gastric lavage process. There will be problems such as not being able to detect physical discomfort in infants and young children in time and taking corresponding measures, which will affect the effect of gastric lavage and even damage the physical and life safety of infants and young children. Summary of the invention
[0003] The purpose of the present invention is to provide a multi-dimensional safety monitoring method and system for infant gastric lavage machines, aiming to solve the technical problems in the prior art that it is difficult to express one's own emotions and thoughts during the gastric lavage process, and that it is not possible to promptly discover and take corresponding measures when the infant is physically unwell.
[0004] To achieve the above object, the technical solution adopted by the present invention is: to provide a multi-dimensional safety monitoring method for an infant gastric lavage machine, comprising the following steps:
[0005] S1: Collect facial images of infants and young children in different scenarios;
[0006] S2: preprocessing the facial image to obtain a preprocessed facial image;
[0007] S3: Input the preprocessed facial image as a sample into a deep learning model for training to obtain a facial emotion recognition model; wherein the facial emotion recognition model includes: extracting local features using a convolutional neural network; extracting global features using a self-attention network structure; and fusing local features and global features in a spatial domain and a channel domain using a feature fusion structure to enhance the expressiveness of features;
[0008] S4: inputting the target facial image into the facial emotion recognition model to determine the emotional state of the target face;
[0009] S5: When the target facial emotion is identified as abnormal, relevant medical staff are notified in a timely manner.
[0010] Preferably, the facial emotion recognition model further comprises: estimating the probability distribution of the corresponding infant expression category using a joint loss function to determine the final expression category of the infant image;
[0011] The formula of the joint loss function is:
[0012]
[0013] in, Represents element-wise multiplication; a ij represents the weight of the i-th deep feature along dimension j calculated by the deep attention network in the embedding space; j∈{1,2,...,d}.
[0014] Preferably, S2 comprises the following steps:
[0015] S2.1, adjust the size of the image;
[0016] S2.2. Use data enhancement methods to enhance the image data.
[0017] S2.3. Use image segmentation algorithm to process the enhanced image to obtain an image containing only the facial area of the infant.
[0018] Preferably, the feature fusion structure is composed of a parallel spatial attention module and a channel attention module.
[0019] Preferably, the formula of the channel attention module is:
[0020]
[0021]
[0022] a c +b c =1
[0023]
[0024] V=[V1,V2,......,V c ]
[0025] Among them, a c express The corresponding attention vector; b c express The corresponding attention vector; A c represents the cth row of A; B c represents the c-th row of B; [.] represents the concatenation of matrices.
[0026] Preferably, the formula of the spatial attention module is:
[0027] I=S(f 5×5 [Avgpool(X in );Maxpool(X in )]);
[0028]
[0029] Among them, S represents the activation function; f 5×5 represents a 5*5 convolution; [.] represents matrix concatenation; AvgPool represents a global average pooling operation; MaxPool represents a global maximum pooling operation; X in ∈R H×W×C represents the features obtained by adding the convolutional neural network and the self-attention network; I represents the fusion feature; f 1×1 Indicates 1*1 convolution; * indicates matrix multiplication; Represents the activation function.
[0030] A multi-dimensional safety monitoring system for an infant gastric lavage machine, characterized in that it is used to implement the steps in a multi-dimensional safety monitoring method for an infant gastric lavage machine as described in any one of the above items; comprising:
[0031] A collection module, used to collect facial images of infants and young children in different scenes;
[0032] A preprocessing module is used to preprocess the facial image to obtain a preprocessed facial image;
[0033] An emotion recognition module, used for inputting the preprocessed facial image as a sample into a deep learning model for training to obtain a facial emotion recognition model; wherein the facial emotion recognition model includes: extracting local features using a convolutional neural network; extracting global features using a self-attention network structure; and fusing local features and global features in a spatial domain and a channel domain using a feature fusion structure to enhance the expressiveness of features;
[0034] An input module, used for inputting a target facial image into the facial emotion recognition model to determine the target facial emotional state;
[0035] The result output module is used to promptly notify relevant medical staff when the target facial emotion is identified as abnormal.
[0036] An electronic device comprises a bus, a transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory and the processor are connected via the bus, and wherein the computer program, when executed by the processor, implements the steps of a multi-dimensional safety monitoring method for an infant gastric lavage machine as described in any one of the above.
[0037] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a multi-dimensional safety monitoring method for an infant gastric lavage machine as described in any one of the above are implemented.
[0038] The beneficial effect of the multi-dimensional safety monitoring method and system for infant gastric lavage machines provided by the present invention is that: compared with the prior art, the multi-dimensional safety monitoring method and system for infant gastric lavage machines provided by the present invention can monitor and identify the facial emotions of infants in real time and accurately during the infant gastric lavage process. When the infant is uncomfortable and the collected facial image is identified as requiring an alarm, the monitoring method can promptly notify the medical staff so that the medical staff can provide relevant measures in a timely manner. The safety monitoring method also has the characteristics of high accuracy and high efficiency in identifying facial emotions. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative work.
[0040] Figure 1 A flowchart of a multi-dimensional safety monitoring method for an infant gastric lavage machine provided by an embodiment of the present invention;
[0041] Figure 2 A structural block diagram of a multi-dimensional safety monitoring system for an infant gastric lavage machine provided in an embodiment of the present invention.
[0042] In the figure: 1. Acquisition module; 2. Preprocessing module; 3. Emotion recognition module; 4. Input module; 5. Result output module. DETAILED DESCRIPTION
[0043] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0044] Please also read Figure 1 to Figure 2 Now, a multi-dimensional safety monitoring method for an infant gastric lavage machine provided by the present invention is described. The multi-dimensional safety monitoring method for an infant gastric lavage machine comprises the following steps:
[0045] Step S1: collecting facial images of infants and young children in different scenes;
[0046] The facial image contains the facial expressions of babies. The facial expressions of babies are divided into two categories: alarm-required and normal emotion-required. The normal emotion-required category includes calm and happy emotions. Sadness, anger, surprise, disgust and fear are alarm-required.
[0047] Step S2: preprocessing the facial image to obtain a preprocessed facial image;
[0048] Preferably, step S2 uses data enhancement means and image segmentation algorithms to pre-process the facial image. Specifically, it includes the following steps:
[0049] Step 2.1, adjust the size of the image;
[0050] Step 2.2: Use data enhancement methods to enhance the image data.
[0051] Step 2.4: Use an image segmentation algorithm to process the enhanced image to obtain an image containing only the infant's facial area.
[0052] Step S3: input the preprocessed facial image as a sample into the deep learning model for training to obtain a facial emotion recognition model; wherein the facial emotion recognition model includes: using a convolutional neural network to extract local features; using a self-attention network structure to extract global features; using a feature fusion structure to fuse local features and global features in the spatial domain and the channel domain to enhance the expressiveness of the features.
[0053] In this step, the backbone network of the convolutional neural network uses ResNet18. The backbone network of the self-attention network uses Swin Transformer.
[0054] Convolutional neural networks can extract features that highlight local details. Self-attention networks can extract features that establish global dependencies between pixels. The feature fusion structure fuses the two in the spatial domain and channel domain. The use of feature fusion structures can improve the accuracy of facial expression recognition. The feature fusion structure obtains the spatial weight information of the input features by fusing the feature maps generated by dimensionality reduction to guide the allocation of feature weights.
[0055] In this step, the feature fusion structure consists of a parallel spatial attention module and a channel attention module.
[0056] The formula used by the spatial attention module is:
[0057] I=S(f 5×5 [Avgpool(X in );Maxpool(X in )]);
[0058] Among them, S represents the activation function; f 5×5 represents a 5*5 convolution; [.] represents matrix concatenation; AvgPool represents a global average pooling operation; MaxPool represents a global maximum pooling operation. X in ∈R H×W×C It represents the features obtained by adding the convolutional neural network and the self-attention network. I represents the fusion feature.
[0059] The specific meaning of this formula is: the features obtained by adding the convolutional neural network and the self-attention network are represented by the maximum element value in each channel vector of the feature through the global maximum pooling operation. The features obtained by adding the convolutional neural network and the self-attention network are represented by the average element value in each channel vector of the feature through the global average pooling operation. The above features are concatenated. Then the concatenated features are enhanced by 5*5 convolution. Then, the fused features are activated by the activation function.
[0060]
[0061] Among them, f 1×1 Indicates 1*1 convolution; * indicates matrix multiplication; Represents the activation function.
[0062] The specific meaning of this formula is: the features obtained by adding the convolutional neural network and the self-attention network are transformed through 1*1 convolution. The dimensionality reduction processing of the baby's facial expression features is converted into a one-dimensional vector. The obtained dimensionality reduction features are transposed and then multiplied by the dimensionality reduction features to fuse the low-dimensional information of the expression features, and then the normalized feature J is obtained through the activation function.
[0063] Feature J is transposed and then multiplied by feature I to promote the interaction of information between the two and restore the obtained features to the dimensions of the original input features of the baby's facial expressions. The information of the features in the spatial dimension is enhanced.
[0064] The use of the channel attention module can enhance the model's attention to important channel features. The expression formula of the feature map in the channel attention module is:
[0065]
[0066]
[0067] a c +b c =1
[0068]
[0069] V=[V1,V2,......,V c ]
[0070] Among them, a c express The corresponding attention vector; b c express The corresponding attention vector; A c represents the cth row of A; B c represents the c-th row of B; [.] represents the concatenation of matrices.
[0071] The specific meaning of this formula is: the features obtained by adding the convolutional neural network and the self-attention network are simultaneously input into the hole convolution and group convolution to obtain the features respectively. as well as Then, the multi-branch information is integrated through the gating unit to obtain The integrated feature U is subjected to global average pooling to obtain the feature map s∈R 1×1×C ; Then through f c (s) function generates vector z∈R d×1 ; and get a c 、b c According to a c 、b c Get the calculation formula for obtaining the feature map V.
[0072] In this step, the facial emotion recognition model also includes estimating the probability distribution of the corresponding infant expression category using a joint loss function to determine the final expression category of the infant image.
[0073] The formula of the joint loss function is:
[0074]
[0075] in, Represents element-wise multiplication; a ij represents the weight of the i-th deep feature along dimension j calculated by the deep attention network in the embedding space; j∈{1,2,...,d}.
[0076] Step S4: inputting the target facial image into the facial emotion recognition model to determine the emotional state of the target face;
[0077] Step S5: When the target face is identified as abnormal, the corresponding emotion is extracted and the situation is promptly notified to the relevant medical staff.
[0078] The present invention provides a multi-dimensional safety monitoring method for infant gastric lavage machines. Compared with the prior art, the monitoring method can monitor and identify infant facial emotions in real time and accurately during infant gastric lavage. When the infant is uncomfortable and the collected facial image is identified as requiring an alarm, the monitoring method can promptly notify medical staff so that the medical staff can provide relevant measures in a timely manner. The safety monitoring method also has the characteristics of high accuracy and high efficiency in identifying facial emotions.
[0079] The present invention also provides a multi-dimensional safety monitoring system for infant gastric lavage machine, please refer to Figure 1 to Figure 2 , used to implement the steps in a multi-dimensional safety monitoring method for an infant gastric lavage machine as described in any one of the above, including: an acquisition module 1, a preprocessing module 2, an emotion recognition module 3, an input module 4 and a result output module 5, wherein the acquisition module 1 is used to acquire facial images of infants in different scenes; the preprocessing module 2 is used to preprocess the facial images to obtain preprocessed facial images; the emotion recognition module 3 is used to input the preprocessed facial images as samples into a deep learning model for training to obtain a facial emotion recognition model; wherein the facial emotion recognition model includes: using a convolutional neural network to extract local features; using a self-attention network structure to extract global features; using a feature fusion structure to fuse local features and global features in the spatial domain and the channel domain to enhance the expression ability of the features; the input module 4 is used to input the target facial image into the facial emotion recognition model to determine the emotional state of the target face; the result output module 5 is used to promptly notify relevant medical staff when the target facial emotion is identified as abnormal.
[0080] Compared with the prior art, the beneficial effects of the multi-dimensional safety monitoring system for infant gastric lavage machines provided by the present invention are the same as the beneficial effects of the multi-dimensional safety monitoring method for infant gastric lavage machines described in the above technical solution, which will not be elaborated here.
[0081] The present invention also provides an electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and wherein the computer program, when executed by the processor, implements the steps of the above-mentioned multi-dimensional safety monitoring method for an infant gastric lavage machine.
[0082] Compared with the prior art, the beneficial effects of the electronic device provided by the present invention are the same as the beneficial effects of the multi-dimensional safety monitoring method for an infant gastric lavage machine described in the above technical solution, and will not be elaborated here.
[0083] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps in the above-mentioned multi-dimensional safety monitoring method for an infant gastric lavage machine.
[0084] Compared with the prior art, the beneficial effects of a computer-readable storage medium provided by the present invention are the same as the beneficial effects of a multi-dimensional safety monitoring method for an infant gastric lavage machine described in the above technical solution, and will not be elaborated here.
[0085] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technical solution that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.
Claims
1. A multi-dimensional safety monitoring method for infant gastric lavage machine, It is characterized in that The following steps are involved: S1: Collect facial images of infants and young children in different scenarios; S2: preprocessing the facial image to obtain a preprocessed facial image; S3: Input the preprocessed facial image as a sample into a deep learning model for training to obtain a facial emotion recognition model; wherein the facial emotion recognition model includes: extracting local features using a convolutional neural network; extracting global features using a self-attention network structure; and fusing local features and global features in a spatial domain and a channel domain using a feature fusion structure to enhance the expressiveness of features; S4: inputting the target facial image into the facial emotion recognition model to determine the emotional state of the target face; S5: When the target facial emotion is identified as abnormal, relevant medical staff are notified in a timely manner.
2. A multi-dimensional safety monitoring method for an infant gastric lavage machine as claimed in claim 1, It is characterized in that The facial emotion recognition model further comprises: estimating the probability distribution of the corresponding infant expression category using a joint loss function to determine the final expression category of the infant image; The formula of the joint loss function is: in, Represents element-wise multiplication; a ij represents the weight of the i-th deep feature along dimension j calculated by the deep attention network in the embedding space; j∈{1,2,...,d}.
3. A multi-dimensional safety monitoring method for an infant gastric lavage machine as claimed in claim 2, It is characterized in that The S2 comprises the following steps: S2.1, adjust the size of the image; S2.
2. Use data enhancement methods to enhance the image data. S2.
3. Use an image segmentation algorithm to process the enhanced image to obtain an image containing only the infant's facial area.
4. A multi-dimensional safety monitoring method and system for an infant gastric lavage machine as claimed in claim 2, It is characterized in that The feature fusion structure is composed of a parallel spatial attention module and a channel attention module.
5. A multi-dimensional safety monitoring method for an infant gastric lavage machine as claimed in claim 4, It is characterized in that The formula of the channel attention module is: a c +b c =1 V=[V 1 ,V 2 ,......,V c ] Among them, a c express The corresponding attention vector; b c express The corresponding attention vector; A c represents the cth row of A; B c represents the c-th row of B; [.] represents the concatenation of matrices.
6. A multi-dimensional safety monitoring method and system for an infant gastric lavage machine according to any one of claims 4 or 5, It is characterized in that The formula of the spatial attention module is: I=S(f 5×5 [Avgpool(X in );Maxpool(X in )]); Among them, S represents the activation function; f 5×5 represents a 5*5 convolution; [.] represents matrix concatenation; AvgPool represents a global average pooling operation; MaxPool represents a global maximum pooling operation; X in ∈R H×W×C represents the features obtained by adding the convolutional neural network and the self-attention network; I represents the fusion feature; f 1×1 Indicates 1*1 convolution; * indicates matrix multiplication; Represents the activation function.
7. A multi-dimensional safety monitoring system for infant gastric lavage machines, It is characterized in that Used to implement the steps in a multi-dimensional safety monitoring method for an infant gastric lavage machine according to any one of claims 1 to 6; comprising: A collection module, used to collect facial images of infants and young children in different scenes; A preprocessing module, used for preprocessing the facial image to obtain a preprocessed facial image; An emotion recognition module, used for inputting the preprocessed facial image as a sample into a deep learning model for training to obtain a facial emotion recognition model; wherein the facial emotion recognition model includes: extracting local features using a convolutional neural network; extracting global features using a self-attention network structure; and fusing local features and global features in a spatial domain and a channel domain using a feature fusion structure to enhance the expressiveness of features; An input module, used for inputting a target facial image into the facial emotion recognition model to determine the target facial emotional state; The result output module is used to promptly notify relevant medical staff when the target facial emotion is identified as abnormal.
8. An electronic device comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus. It is characterized in that When the computer program is executed by the processor, the steps of a multi-dimensional safety monitoring method for an infant gastric lavage machine as described in any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of a multi-dimensional safety monitoring method for an infant gastric lavage machine as described in any one of claims 1 to 6 are implemented.