Intelligent body temperature real-time monitoring system and method

Abstract

The invention relates to the field of body temperature monitoring, and particularly discloses an intelligent body temperature real-time monitoring system and method.The intelligent body temperature real-time monitoring method comprises the steps that firstly, a patient infrared thermal imaging image collected by a thermal infrared imager and patient physiological environment text parameters collected by a database are obtained, and then feature extraction and correlation analysis are conducted on the two images through the deep learning technology; and finally, obtaining a classification result through a classifier to judge whether a patient body temperature abnormity early warning is sent out or not, so that the accuracy of body temperature abnormity detection is improved, and comprehensive and real-time monitoring and early warning of the patient body temperature are realized.

Description

Technical Field

[0001] The present application relates to the field of body temperature monitoring, and more specifically, to an intelligent body temperature real-time monitoring system and method. Background Art

[0002] Body temperature refers to the temperature inside the human body, which reflects the body's metabolic activity and physiological state. Normal body temperature is generally between 36.5℃ and 37.5℃. Deviation from this range may mean that something is wrong with the body. Body temperature monitoring is crucial for timely detection of health problems such as fever and hypothermia. It helps to assess the severity of the disease, monitor the effectiveness of treatment, and guide nursing measures.

[0003] Existing body temperature monitoring methods include traditional methods such as laser plasma, infrared, capacitance measurement, thermal resistance measurement, and mercury thermometers. Although these methods can effectively provide body temperature data, they usually have limitations, such as only being able to measure at a certain moment, or only being able to display the highest body temperature value in a certain period of time. Existing monitoring methods cannot conduct a comprehensive and in-depth analysis of body temperature, thus missing a lot of important human health information contained in body temperature.

[0004] Therefore, an intelligent real-time body temperature monitoring system and method are desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides an intelligent real-time temperature monitoring system and method, which first obtains the patient's infrared thermal imaging image collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database, and then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally obtains the classification result through the classifier to determine whether to issue an abnormal temperature warning for the patient, thereby improving the accuracy of abnormal temperature detection, so as to achieve comprehensive and real-time monitoring and early warning of the patient's body temperature.

[0006] According to one aspect of the present application, there is provided an intelligent real-time body temperature monitoring system, which includes:

[0007] A patient temperature data acquisition module is used to acquire the patient's infrared thermal imaging image acquired by the infrared thermal imager and the patient's physiological environment text parameters acquired by the database;

[0008] A patient temperature data extraction module, used for extracting a patient infrared thermal imaging area tracking association feature vector and a patient physiological environment text semantic association feature vector from the patient infrared thermal imaging image acquired by the infrared thermal imager and the patient physiological environment text parameters acquired by the database;

[0009] The patient temperature abnormality warning judgment module is used to determine whether to issue a patient temperature abnormality warning based on the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector.

[0010] According to another aspect of the present application, a smart body temperature real-time monitoring method is provided, which includes:

[0011] Acquire the patient's infrared thermal imaging image acquired by the infrared thermal imager and the patient's physiological environment text parameters acquired by the database;

[0012] Extracting the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector from the patient's infrared thermal imaging image acquired by the infrared thermal imager and the patient's physiological environment text parameters acquired by the database;

[0013] Based on the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector, it is determined whether to issue a warning of abnormal patient temperature.

[0014] Compared with the prior art, the present application provides an intelligent real-time body temperature monitoring system and method, which first obtains the patient's infrared thermal imaging image collected by an infrared thermal imager and the patient's physiological environment text parameters collected by a database, and then uses deep learning technology to perform feature extraction and correlation analysis on the two. Finally, a classifier is used to obtain a classification result to determine whether to issue an abnormal body temperature warning for the patient, thereby improving the accuracy of abnormal body temperature detection and realizing comprehensive, real-time monitoring and early warning of the patient's body temperature. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a block diagram of an intelligent body temperature real-time monitoring system according to an embodiment of the present application.

[0017] Figure 2 This is a block diagram of a patient temperature data extraction module in an intelligent temperature real-time monitoring system according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of a patient infrared thermal imaging feature extraction unit in an intelligent body temperature real-time monitoring system according to an embodiment of the present application.

[0019] Figure 4This is a block diagram of the patient temperature abnormality warning judgment module in the intelligent body temperature real-time monitoring system according to an embodiment of the present application.

[0020] Figure 5 Flow chart of the intelligent body temperature real-time monitoring method according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0022] Figure 1 : is a block diagram of the intelligent real-time temperature monitoring system of the embodiment of the present application. Figure 1 As shown, according to the embodiment of the present application, the intelligent body temperature real-time monitoring system 100 includes: a patient body temperature data acquisition module 110, which is used to obtain the patient's infrared thermal imaging image collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database; a patient body temperature data extraction module 120, which is used to extract the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector from the patient's infrared thermal imaging image collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database; a patient body temperature abnormal warning judgment module 130, which is used to judge whether to issue a patient body temperature abnormal warning based on the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector.

[0023] In the above-mentioned intelligent real-time temperature monitoring system 100, the patient temperature data acquisition module 110 is used to obtain the patient's infrared thermal imaging image collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database. It should be understood that the infrared thermal imager obtains the temperature distribution image by detecting the infrared radiation on the patient's body surface and displays the temperature on the patient's body surface. Through a computer or mobile device connected to the infrared thermal imager, these thermal imaging images are obtained and saved in a common image format (such as JPEG or TIFF) so that the image can be further processed and analyzed. At the same time, the patient's physiological environment data (such as heart rate, blood pressure, respiratory rate, body temperature, etc.) are usually collected and stored in real time by medical equipment or monitoring systems. These data monitor the patient's physiological state through sensors and are usually recorded in a medical database in text or numerical form. Combining these two types of data, namely infrared thermal imaging images and physiological environment text parameters, can provide a more comprehensive patient health assessment and determine whether to issue an abnormal patient temperature warning. Specifically, body temperature is the temperature inside the human body, which reflects the body's metabolic process and overall health status. Monitoring body temperature is crucial for timely detection of health problems such as fever and hypothermia. It can help assess the severity of the disease, observe the effect of treatment, and provide guidance for nursing work. At present, common body temperature monitoring methods include laser plasma, infrared, capacitance measurement, thermal resistance measurement, and traditional mercury thermometers. Although these technologies can provide body temperature data, they usually have some limitations. For example, they can only measure at a specific time point or can only display the highest body temperature at a certain moment. These existing monitoring methods cannot conduct a comprehensive and in-depth analysis of body temperature and may miss many key health information contained in body temperature. Therefore, in the technical solution of the present application, by obtaining the infrared thermal imaging image of the patient collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database, and combining deep learning technology, it is determined whether the patient's body temperature is abnormal, thereby improving the accuracy of abnormal body temperature detection, and then realizing all-round, real-time monitoring and early warning of the patient's body temperature.

[0024] In the above-mentioned intelligent real-time temperature monitoring system 100, the patient temperature data extraction module 120 is used to extract the patient infrared thermal imaging area tracking association feature vector and the patient physiological environment text semantic association feature vector from the patient infrared thermal imaging image collected by the infrared thermal imager and the patient physiological environment text parameter collected by the database. It should be understood that the thermal imaging image provided by the infrared thermal imager can reflect the temperature change of the patient's body surface, while the physiological environment text parameter records the patient's heart rate, body temperature, blood pressure and other key physiological data, which can reflect their overall health status. By extracting hot zone features (such as temperature distribution, regional morphology) from the thermal imaging image, and extracting semantic information of physiological state changes (such as body temperature fluctuations, abnormal heart rate) from the physiological text data, the patient infrared thermal imaging area tracking association feature vector and the patient physiological environment text semantic association feature vector can be obtained, which represent the patient's local physiological state (infrared image) and the whole body health state (physiological text), respectively. These two types of information are integrated through deep learning or machine learning models to form a complete health record, which provides data support for early diagnosis, dynamic monitoring and personalized treatment of diseases.

[0025] Figure 2 FIG. 1 is a block diagram of a patient temperature data extraction module in an intelligent temperature real-time monitoring system according to an embodiment of the present application. Figure 2 As shown, in a specific embodiment of the present application, the patient body temperature data extraction module 120 includes: a patient infrared thermal imaging feature extraction unit 121, which is used to perform feature extraction on the patient's infrared thermal imaging image collected by the infrared thermal imager to obtain the patient's infrared thermal imaging area tracking associated feature vector; a patient physiological environment text semantic encoding unit 122, which is used to perform semantic encoding on the patient's physiological environment text parameters collected by the database to obtain the patient's physiological environment text semantic associated feature vector.

[0026] It should be understood that infrared thermal imaging images can intuitively present the temperature differences of various areas on the body surface. By extracting features from infrared thermal imaging images, useful temperature and spatial features can be extracted from a large amount of pixel data, thereby generating feature vectors that are convenient for subsequent analysis. Specifically, in a specific embodiment of the present application, first, image preprocessing techniques such as denoising, image enhancement and resizing are used to ensure image quality and make it suitable for further analysis. Secondly, image segmentation technology is applied to divide the image into different areas, such as high temperature areas, low temperature areas and background areas. Then, feature extraction methods such as texture analysis, morphological features (such as the shape, size, and edge of the area) or features based on thermal changes (such as temperature gradients, local temperature fluctuations, etc.) are used to extract the temperature features of each area. Common feature extraction methods include gray level co-occurrence matrix (GLCM), local binary pattern (LBP), Hough transform, etc., which can capture the structural and pattern features in the image. Finally, these extracted regional features are integrated into feature vectors, that is, tracking association feature vectors for each thermal imaging area. These feature vectors not only reflect the temperature information of each area, but also capture the dynamic change characteristics of the area.

[0027] Further, considering that the physiological environment text parameters usually include the patient's vital signs, medical records, medical history information, etc., these data are usually unstructured or semi-structured text data, which are difficult to directly perform numerical processing. Therefore, the process of semantic encoding is a key step in processing and converting these text data so that they can be effectively analyzed by machine learning or deep learning models. Specifically, in a specific embodiment of the present application, first, text preprocessing is the first step in semantic encoding, which usually includes removing irrelevant information (such as stop words), punctuation marks, and unifying data formats. Afterwards, word segmentation technology is used to convert physiological environment text data into structured vocabulary units. These vocabulary units include important information describing the patient's current health status, such as body temperature, blood pressure, heart rate, respiratory rate, etc. Next, each vocabulary is converted into a high-dimensional vector through word embedding technology, such as Word2Vec, GloVe or a more advanced BERT model. Word embedding technology can capture the semantic relationship between words, making similar or related words closer in the vector space. For example, words such as body temperature, fever, and hypothermia can be mapped to similar vector space positions, reflecting their semantic relevance. In order to extract the patient's physiological environment characteristics, context modeling methods are usually used. For example, the BERT model can capture the relationship and semantics between different physiological states (such as cardiac arrest, hypoglycemia, hypertension, etc.) by modeling the context of the entire sentence. In this way, the health information involved in the text can be converted into a feature vector with a semantic hierarchy. Finally, the obtained physiological environment text semantic association feature vector can not only contain the semantic information of each word in the text, but also comprehensively consider the semantic relationship of each part of the entire text. This feature vector can comprehensively reflect the patient's health status, including changes in its physiological parameters, potential risks, and possible pathological states, and provide effective input for subsequent analysis (such as disease prediction and health assessment). The feature vector converted through semantic encoding can achieve more accurate patient health assessment and dynamic monitoring.

[0028] Figure 3 FIG. 1 is a block diagram of a patient infrared thermal imaging feature extraction unit in an intelligent real-time temperature monitoring system according to an embodiment of the present application. Figure 3As shown, in a specific embodiment of the present application, the patient infrared thermal imaging feature extraction unit 121 includes: a patient infrared thermal imaging image division subunit 1211, used to divide the patient infrared thermal imaging image collected by the infrared thermal imager into multiple patient infrared thermal imaging area tracking images; a patient infrared thermal imaging area tracking feature encoding subunit 1212, used to perform feature encoding on the multiple patient infrared thermal imaging area tracking images to obtain a patient infrared thermal imaging area tracking associated feature map; a global mean pooling subunit 1213, used to perform global mean pooling on the patient infrared thermal imaging area tracking associated feature map to obtain the patient infrared thermal imaging area tracking associated feature vector.

[0029] It should be understood that the purpose of dividing the infrared thermal imaging image of the patient collected by the infrared thermal imager into a plurality of infrared thermal imaging regional tracking images of the patient is to analyze the local temperature changes of the patient's body surface in more detail. Among them, the infrared thermal imaging image usually contains a lot of detailed information, so it is necessary to denoise the image, enhance the contrast or equalize the image to improve the image quality and make the regional division process more accurate. In a specific embodiment of the present application, an image segmentation algorithm is used to divide the entire image into multiple regions. Common image segmentation methods include threshold segmentation, region growing method, edge detection and K-means clustering. By setting a temperature threshold or temperature gradient, different regions in the image can be distinguished. Through these algorithms, the infrared image can be divided into multiple regions of interest, which have similar temperature characteristics and are called infrared thermal imaging regional tracking images.

[0030] Furthermore, feature encoding is performed on multiple infrared thermal imaging area tracking images of patients to obtain the patient infrared thermal imaging area tracking associated feature map. The main purpose is to digitize and structure the key information of each local area in the image, thereby realizing a comprehensive analysis and monitoring of the patient's body surface temperature changes. Specifically, after obtaining multiple infrared thermal imaging area tracking images, detailed feature extraction is required for each area. Among them, the temperature characteristics of each area can be described by calculating the statistical features such as the temperature mean, standard deviation, maximum and minimum temperature difference of each area to express the temperature distribution information in the thermal imaging image of each area. Morphological features such as area, boundary shape, location and extension of hot spots can also be extracted to reflect local temperature changes. In addition, deep learning methods can be used to further encode these features. A common method is to input the feature vector of each area into a convolutional neural network (CNN) for processing. CNN can automatically learn the local correlation features in space from the image, extract the local temperature change pattern through the convolution layer, and obtain a higher-level feature representation. For each area in the thermal imaging image, CNN can learn the relationship between the temperature change of the area and the surrounding area, as well as its position and importance in the overall image. Through these deep learning methods, the feature vector of each region can be converted into a patient infrared thermal imaging region tracking association feature map with high-dimensional semantic representation. This feature map uniformly encodes the temperature characteristics, morphological characteristics, time series characteristics and other information of each region, and maps them according to spatial position to form a comprehensive image representation that can be used for subsequent analysis.

[0031] Furthermore, the patient infrared thermal imaging region tracking associated feature map is subjected to global mean pooling to obtain the patient infrared thermal imaging region tracking associated feature vector, aiming to compress the complex, high-dimensional feature map information into a relatively concise, low-dimensional vector representation, so as to make the subsequent analysis, classification and prediction tasks more efficient. Global mean pooling can extract global and representative feature information by processing the feature map, and effectively reduce the complexity of calculation and storage. Specifically, in a neural network or other deep learning model, a feature map is the output after a series of convolution operations and pooling layers, and usually contains local feature information of an image or region. These feature maps are usually two-dimensional, and each element represents the feature value of a certain area of ​​the image, such as temperature, shape, texture, etc. In order to integrate the information of the entire feature map into a unified feature vector, a pooling operation is usually used, especially global mean pooling. Global average pooling (GAP) is a common pooling method, which generates a feature vector representing the entire image by calculating the average value of each channel of the entire feature map. The main advantage of global mean pooling is that it can retain the global information of the input feature map without losing important contextual features. Compared with the traditional fully connected layer, the GAP operation can not only significantly reduce the number of parameters and calculations, but also effectively avoid overfitting because it eliminates the dependency of spatial position, making the obtained feature vector more robust and more generalizable.

[0032] In a specific embodiment of the present application, the patient infrared thermal imaging area tracking feature encoding subunit 1212 includes: passing the multiple patient infrared thermal imaging area tracking images through a patient infrared thermal imaging area tracking convolutional neural network model with multi-feature extraction capabilities to obtain multiple patient infrared thermal imaging area tracking feature maps; passing the multiple patient infrared thermal imaging area tracking feature maps through a patient infrared thermal imaging area tracking target detection network to obtain multiple patient infrared thermal imaging tracking regions of interest; passing the multiple patient infrared thermal imaging area tracking regions of interest through a patient infrared thermal imaging area tracking significant target detector to obtain the patient infrared thermal imaging area tracking associated feature map.

[0033] It should be understood that processing multiple patient infrared thermal imaging area tracking images through a convolutional neural network (CNN) model with multi-feature extraction capabilities can effectively extract multi-level and multi-dimensional key information from complex thermal imaging data, thereby generating multiple patient infrared thermal imaging area tracking feature maps. These feature maps contain rich spatial, temperature change and structural information, which helps to make comprehensive temperature judgments of patients based on multi-region temperature changes. Specifically, convolutional neural networks can gradually extract different features in images through multiple convolutional layers. The primary convolutional layer extracts low-level local features, such as edges and corners in the image, while deeper convolutional layers can identify more complex patterns and structures, such as overall changes in temperature distribution, temperature differences between regions, dynamic changes, etc. In addition, the multi-channel feature extraction capability of the CNN model enables the network to learn different feature dimensions at the same time, thereby enhancing the richness and diversity of feature representation. Specifically, each layer of the patient infrared thermal imaging area tracking convolutional neural network model with multi-feature extraction capability is used to perform convolution processing, mean pooling processing based on the local feature matrix and nonlinear activation processing on the input data in the forward transfer of the layer, so that the last layer of the patient infrared thermal imaging area tracking convolutional neural network model with multi-feature extraction capability outputs the multiple patient infrared thermal imaging area tracking feature maps, wherein the input of the patient infrared thermal imaging area tracking convolutional neural network model with multi-feature extraction capability is the multiple patient infrared thermal imaging area tracking images.

[0034] Furthermore, the target detection network is usually composed of multiple convolutional layers and region extraction modules, which can further refine and locate the high-dimensional information in the infrared thermal imaging feature map through multiple convolution and pooling operations. Specifically, considering that the infrared thermal imaging region tracking feature map contains the temperature distribution and change information at different spatial positions, the target detection network will automatically identify the regions of interest in the image based on these features through specific detection algorithms (such as region proposal networks RPN, or YOLO and other frameworks). These regions often represent areas of abnormal temperature changes. The target detection network can not only identify the spatial positions of these regions. Specifically, the patient infrared thermal imaging region tracking target detection network is used to process the multiple patient infrared thermal imaging region tracking feature maps with the following target detection formula to obtain the multiple patient infrared thermal imaging tracking regions of interest;

[0035] Among them, the target detection formula is:

[0036] ROI=(cls(ψ det ,B),Regr(ψ det , B))

[0037] Among them, ψ detis the tracking feature map of multiple patients’ infrared thermal imaging regions, B is the anchor frame, ROI is the tracking region of interest of multiple patients’ infrared thermal imaging, cls(ψ det , B) indicates classification, Regr(ψ det , B) indicates regression.

[0038] Furthermore, the infrared thermal imaging tracking regions of interest of multiple patients are processed by the patient infrared thermal imaging region tracking salient target detector, aiming to further extract infrared thermal imaging region tracking associated feature maps with high recognition and diagnostic significance. Among them, the salient target detector can identify and highlight important areas in the image, especially when processing complex infrared thermal imaging data, it can effectively screen out key temperature change areas from a large range of thermal imaging images. Specifically, the salient target detector analyzes the regions of interest in the infrared thermal imaging images of multiple patients and uses a deep learning-based algorithm to automatically identify salient targets therein, such as areas with drastic temperature changes. Through salient target detection, the system can extract these salient areas, accurately locate and enhance them, to ensure that the subsequent diagnostic process can focus on these important areas. The detector usually performs feature learning on the input image through a multi-layer convolutional neural network (CNN) to identify areas that are important for disease diagnosis. In specific operations, the salient target detector first processes the infrared thermal imaging images of multiple patients to identify all potential salient target areas, and then generates associated feature maps through feature extraction and aggregation at different levels. These feature maps reflect the temperature distribution pattern and spatial characteristics of these salient areas through calibration and enhancement. Ultimately, the obtained infrared thermal imaging area tracking associated feature map has stronger distinguishability and can provide a more accurate reference for subsequent diagnosis, monitoring and treatment. Specifically, each layer of the patient infrared thermal imaging area tracking salient target detector performs the following on the input data in the forward pass of the layer: convolution processing is performed on the input data using a first convolution kernel to obtain a first convolution feature map; convolution processing is performed on the first convolution feature map using a second convolution kernel to obtain a second convolution feature map, wherein the size of the first convolution kernel is larger than the size of the second convolution kernel; and, pooling processing is performed on the second convolution feature map to obtain a pooled feature map; and, activation processing is performed on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the patient infrared thermal imaging area tracking salient target detector is the patient infrared thermal imaging area tracking associated feature map, and the input of the first layer of the patient infrared thermal imaging area tracking salient target detector is the multiple patient infrared thermal imaging tracking regions of interest.

[0039] In a specific embodiment of the present application, the patient physiological environment text semantic encoding unit 122 includes: passing the patient physiological environment text parameters collected from the database through a patient physiological environment text semantic encoder to obtain a plurality of patient physiological environment text semantic feature vectors; and cascading the plurality of patient physiological environment text semantic feature vectors into the patient physiological environment text semantic association feature vector.

[0040] It should be understood that the patient's physiological environment text parameters usually include physiological data such as the patient's body temperature, heart rate, respiratory rate, blood oxygen level, and related environmental information (such as indoor temperature, humidity, air quality, etc.). These text parameters are diverse and complex, and direct analysis may lead to information loss or misunderstanding. In order to effectively convert these text data into a format that can be understood by machine learning algorithms, semantic encoders (such as BERT, Transformer and other models based on deep learning) can be used to extract the potential semantic information in the text. In the specific operation process, the patient's physiological environment text semantic encoder first inputs the collected text data into the pre-trained language model. The model processes the words, sentence structures and grammatical relations in the text layer by layer through a deep neural network to obtain the semantic information in the text. The encoder analyzes the relationship between the various parts of the text through a self-attention mechanism to capture contextual information closely related to the patient's physiological and environmental status. Then, after multi-layer encoding and feature extraction, the model outputs a multi-dimensional semantic feature vector, which contains the core information of the patient's physiological environment text, such as temperature changes, physical sign fluctuations and other details, and can efficiently represent the key information and potential patterns in the text. Specifically, the patient physiological environment text parameters collected from the database are segmented to obtain a patient physiological environment word sequence; the embedding layer of the patient physiological environment text semantic encoder is used to map each patient physiological environment word in the patient physiological environment word sequence into a word embedding vector to obtain a sequence of patient physiological environment word embedding vectors; the converter-based Bert model of the patient physiological environment text semantic encoder is used to perform global contextual semantic encoding on the sequence of patient physiological environment word embedding vectors to obtain multiple patient physiological environment text semantic feature vectors.

[0041] Furthermore, multiple patient physiological environment text semantic feature vectors are cascaded into patient physiological environment text semantic association feature vectors, aiming to integrate the physiological environment information of each patient into a high-dimensional feature representation, so as to conduct a more comprehensive and accurate analysis of the patient's health status. It should be understood that the physiological environment text semantic feature vector of each patient is extracted by the text semantic encoder from physiological parameters (such as body temperature, heart rate, etc.) and environmental information (such as humidity, temperature, etc.), reflecting the health status and physiological environment characteristics of the patient at a certain point in time. However, these feature vectors themselves are local and only represent the health status of a single patient. If only the feature vector of a single patient is relied on for analysis, the commonalities or differences between patients may not be captured, especially in large-scale health data analysis, which lacks overall relevance. Therefore, cascading the physiological environment text semantic feature vectors of multiple patients is a necessary step. Specifically, the cascade operation usually splices or superimposes the text semantic feature vectors of multiple patients along the feature dimension, aligns the vectors of each patient by column, and forms a richer feature space, reflecting the overall health status and physiological environment of all patients. The concatenated semantically relevant feature vectors not only retain the unique information of each patient, but also extract the underlying patterns and trends among the patient groups.

[0042] In the above-mentioned intelligent real-time temperature monitoring system 100, the patient temperature abnormality warning judgment module 130 is used to determine whether to issue a patient temperature abnormality warning based on the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector. It should be understood that the accuracy and sensitivity of abnormality detection can be improved by comprehensively analyzing multi-source data. Among them, the infrared thermal imaging area tracking associated feature vector contains spatial information about the patient's body temperature distribution, which can intuitively reflect the changes in the patient's body surface temperature, especially the abnormal fluctuations in local temperature. The physiological environment text semantic associated feature vector helps understand the patient's overall health status and the physiological environment in which he is located by analyzing the patient's physiological parameters and environmental conditions, such as indoor temperature, humidity, heart rate, etc. These data provide important background information, which can assist in judging whether the temperature change is abnormal and whether there is a potential risk of disease.

[0043] Figure 4 FIG. 1 is a block diagram of a patient temperature abnormality warning and judgment module in the intelligent temperature real-time monitoring system according to an embodiment of the present application. Figure 4As shown, in a specific embodiment of the present application, the patient abnormal body temperature warning judgment module 130 includes: a patient body temperature feature fusion unit 131, which is used to perform target domain adaptation adjustment on the patient infrared thermal imaging area tracking association feature vector and the patient physiological environment text semantic association feature vector based on the fine-grained semantic information field to obtain a patient abnormal body temperature warning judgment feature vector; an abnormal body temperature warning classification unit 132, which is used to pass the patient abnormal body temperature warning judgment feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a patient abnormal body temperature warning.

[0044] It should be understandable that it is impossible to accurately monitor the patient's body temperature by relying solely on a certain type of data. Therefore, in the technical solution of the present application, by fusing the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector, multiple factors are comprehensively considered to provide a more accurate judgment of abnormal body temperature.

[0045] In particular, considering that infrared thermal imaging images mainly sense the body temperature distribution of patients through thermal radiation, their features reflect information related to thermal energy, such as local temperature anomalies and hot spots. Physiological environmental text parameters, on the other hand, are composed of various physiological indicators, environmental conditions and other information collected in the database, usually presented in the form of numerical values ​​or text, which is completely different from the distribution and dimension of thermal imaging images. Since these data come from different perception sources (vision, sensor, text), they have large differences in distribution mode, data type, noise characteristics, etc. in the feature space, resulting in inconsistent feature spaces between the source domain and the target domain. When the source domain model (e.g., a model trained based on a specific environment or device) is applied to the target domain (e.g., infrared images or physiological parameters of another patient), this difference in feature space may lead to a degradation in the performance of the model. Since the data distribution of the target domain is different from the training data distribution of the source domain, the model in the source domain may not be able to effectively identify new patterns or data features in the target domain, which limits the generalization ability of the model and manifests as performance degradation. In addition, traditional feature fusion methods often fuse through simple concatenation, weighting or averaging, but these methods usually ignore the more complex semantic associations between different data modalities. For example, there may be a potential relationship between the temperature information of infrared thermal imaging data and physiological parameters such as heart rate and blood pressure in physiological environment data. This semantic association cannot be effectively expressed by a simple combination of feature values. Traditional methods have failed to deeply explore the deep semantic relationship between the various modalities, such as how the heat map and physiological parameters jointly reflect the abnormal changes in the patient's body temperature. Traditional fusion methods often cannot make full use of the association information between different modalities, resulting in inaccurate feature representation after fusion, affecting the performance of subsequent classifiers. Therefore, in the technical solution of the present application, the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector are subjected to target domain adaptation adjustment based on a fine-grained semantic information field to obtain the patient's abnormal body temperature warning judgment feature vector.

[0046] Specifically, the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector are subjected to target domain adaptation adjustment based on a fine-grained semantic information field to obtain a patient's abnormal body temperature warning judgment feature vector, including: mapping the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector to an intrinsic decomposition space to obtain a patient's infrared thermal imaging area tracking associated intrinsic feature vector and a patient's physiological environment text semantic associated intrinsic feature vector; mapping the patient's infrared thermal imaging area tracking associated intrinsic feature vector and the patient's physiological environment text semantic associated intrinsic feature vector to an intrinsic decomposition space .... The method comprises the following steps: performing inter-domain feature calibration to obtain an intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and an intrinsic feature vector after the patient's physiological environment text semantic association modulation; constructing a multimodal fine-grained semantic information field of patient information between the intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient's physiological environment text semantic association modulation; and fusing the intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient's physiological environment text semantic association modulation based on the multimodal fine-grained semantic information field of patient information to obtain the feature vector for early warning judgment of abnormal body temperature of the patient.

[0047] More specifically, a multimodal fine-grained semantic information field of patient information is constructed between the intrinsic feature vector after modulation of the patient's infrared thermal imaging area tracking association and the intrinsic feature vector after modulation of the patient's physiological environment text semantic association, including: multiplying the transposed vectors of the intrinsic feature vector after modulation of the patient's infrared thermal imaging area tracking association and the intrinsic feature vector after modulation of the patient's physiological environment text semantic association, and then dividing by the scale of the intrinsic feature vector after modulation of the patient's physiological environment text semantic association to obtain a multimodal fine-grained interaction matrix of patient information; and convolutionally encoding the multimodal fine-grained semantic information field of patient information on the multimodal fine-grained matrix of patient information.

[0048] More specifically, the intrinsic feature vector after modulation of the patient's infrared thermal imaging area tracking association and the intrinsic feature vector after modulation of the patient's physiological environment text semantic association are fused based on the multimodal fine-grained semantic information field of the patient information to obtain the patient's abnormal body temperature warning judgment feature vector, including: multiplying the multimodal fine-grained semantic information field of the patient information by the intrinsic feature vector after modulation of the patient's infrared thermal imaging area tracking association to obtain the patient's infrared thermal imaging area tracking association feature vector after information field mapping; multiplying the multimodal fine-grained semantic information field of the patient information by the intrinsic feature vector after modulation of the patient's physiological environment text semantic association to obtain the patient's physiological environment text semantic association feature vector after information field mapping; weighted fusion of the patient's infrared thermal imaging area tracking association feature vector after information field mapping and the patient's physiological environment text semantic association feature vector after information field mapping to obtain the patient's abnormal body temperature warning judgment feature vector.

[0049] In the embodiment of the present application, specifically, the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector are subjected to target domain adaptation adjustment based on a fine-grained semantic information field to obtain a patient's abnormal body temperature warning judgment feature vector, including: processing the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector according to the following formula to obtain a patient's abnormal body temperature warning judgment feature vector; wherein the formula is:

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057] v c =αv kt +βv st

[0058] Where V1 represents the patient infrared thermal imaging region tracking associated feature vector, V2 represents the patient physiological environment text semantic associated feature vector, PCA(·) represents principal component analysis, U1 represents the sequence of patient infrared thermal imaging region tracking associated eigendecomposition vectors, Λ1 represents the patient infrared thermal imaging region tracking associated diagonal matrix, λ11 and λ 1m denote the eigenvalues ​​of the first and mth positions of the patient infrared thermal imaging region tracking correlation diagonal matrix, T denotes the transpose of the vector, concat{·;·} denotes vector concatenation, and v 11 、v 12 、v 1m They represent the first, second and mth eigenvectors of the sequence of the associated eigendecomposition vectors of the patient's infrared thermal imaging region tracking, V k represents the patient infrared thermal imaging area tracking associated eigenvector, U2 represents the sequence of the patient physiological environment text semantic association eigendecomposition vector, Λ2 represents the patient infrared thermal imaging area tracking associated diagonal matrix, λ 21 and λ 2m They represent the eigenvalues ​​of the first and mth positions of the semantic association diagonal matrix of the patient's physiological environment text, respectively. 21 、v 22 、v 2m The first, second and mth eigenvectors of the sequence of semantic association eigendecomposition vectors of the patient's physiological environment text are represented respectively, V s Represents the semantic association eigenvector of the patient's physiological environment text, v i represents the ith eigenvalue of the patient’s infrared thermal imaging region tracking associated eigenvector, v j The jth eigenvalue of the semantically associated eigenvector of the patient's physiological environment text, represents the square of the vector norm, exp represents the natural exponential function, V′ k represents the intrinsic feature vector after the patient's infrared thermal imaging area tracking correlation modulation, V′ s Represents the intrinsic feature vector after modulation of the semantic association of the patient's physiological environment text, represents matrix multiplication, S represents the scale of the vector, M x Represents the multimodal fine-grained interaction matrix of patient information, Conv 3×3 (M x ) indicates that M x A convolutional encoding operation based on a convolution kernel of 3×3 is performed, Ω represents the multimodal fine-grained semantic information field of patient information, and v kt represents the tracking correlation feature vector of the patient's infrared thermal imaging area after information field mapping, v st represents the semantic association feature vector of the patient's physiological environment text after information field mapping, α and β represent weighted hyperparameters, and v c Represents the patient's abnormal body temperature warning judgment feature vector.

[0059] That is, in the technical solution of the present application, the patient infrared thermal imaging area tracking associated feature vector and the patient physiological environment text semantic associated feature vector are subjected to target domain adaptation adjustment based on the fine-grained semantic information field. The process first maps the patient infrared thermal imaging area tracking associated feature vector and the patient physiological environment text semantic associated feature vector to the intrinsic decomposition space to obtain the patient infrared thermal imaging area tracking associated intrinsic feature vector and the patient physiological environment text semantic associated intrinsic feature vector. Those of ordinary skill in the art should know that the process of mapping to the intrinsic decomposition space is a combination of dimensionality reduction and feature extraction. In a specific example, the eigenvalue decomposition technology of principal component analysis (PCA) is used to find the intrinsic decomposition space, and the intrinsic decomposition space mapping is performed. The patient infrared thermal imaging area tracking associated intrinsic feature vector and the patient physiological environment text semantic associated intrinsic feature vector after information field mapping not only represent the projection of the original feature vector in the new coordinate system, but also reflect the main direction of change of the data, which helps to avoid the "dimensionality disaster", that is, the problem that the sample density drops sharply as the number of features increases.

[0060] Next, after mapping the feature vector to the intrinsic decomposition space, it is necessary to perform inter-domain feature calibration on the patient infrared thermal imaging region tracking associated intrinsic feature vector and the patient physiological environment text semantic associated intrinsic feature vector to adjust the source domain feature vector to better match the data distribution of the target domain. This operation aims to capture and compensate for the systematic deviation between the two domains to reduce the hidden disturbance of the class expression of the local feature distribution of the patient infrared thermal imaging region tracking associated intrinsic feature vector and the patient physiological environment text semantic associated intrinsic feature vector to the overall class expression. After modulation, the patient infrared thermal imaging region tracking associated modulated intrinsic feature vector and the patient physiological environment text semantic associated modulated intrinsic feature vector can retain the source domain features while being more in line with the actual needs of the target domain, thereby enhancing the model's understanding and application of the target domain data.

[0061] After completing the scale modulation of the feature vector, a multimodal fine-grained semantic information field of patient information is constructed between the intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient's physiological environment text semantic association modulation. The multimodal fine-grained semantic information field of patient information is used to describe a more detailed relationship between feature vectors. The introduction of the multimodal fine-grained semantic information field of patient information enables the model to go beyond simple point-to-point comparison and consider more complex patterns and contextual information, thereby enhancing the model's sensitivity to subtle differences and improving the quality of feature representation. Finally, based on the multimodal fine-grained semantic information field of patient information, the intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient's physiological environment text semantic association modulation are fused to obtain the patient's abnormal body temperature warning judgment feature vector.

[0062] Furthermore, the patient's abnormal body temperature warning judgment feature vector integrates a variety of information from infrared thermal imaging and physiological environment data, including the temperature distribution characteristics of the patient's body surface and the environmental and physiological factors that affect body temperature. After these feature vectors are processed by the classifier, they can be converted into clear classification results to determine whether it is necessary to issue an abnormal body temperature warning. Specifically, the patient's abnormal body temperature warning judgment feature vector will be used as the input of the classifier, and the classifier will determine whether the patient's body temperature status is "normal" or "abnormal" based on the feature vector. If the output of the classifier is "abnormal", the system will further determine whether an alarm needs to be issued based on the degree of abnormality (such as the temperature increase amplitude, duration, etc.).

[0063] In summary, the embodiment of the present application first obtains the infrared thermal imaging image of the patient collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database, and then uses deep learning technology to perform feature extraction and correlation analysis on the two. Finally, a classifier is used to obtain a classification result to determine whether to issue an abnormal temperature warning for the patient, thereby improving the accuracy of abnormal temperature detection and realizing comprehensive and real-time monitoring and early warning of the patient's body temperature.

[0064] As described above, the intelligent real-time body temperature monitoring system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the intelligent real-time body temperature monitoring system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent real-time body temperature monitoring system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the intelligent real-time body temperature monitoring system 100 can also be one of the many hardware modules of the terminal device.

[0065] Alternatively, in another example, the intelligent body temperature real-time monitoring system 100 and the terminal device may also be separate devices, and the intelligent body temperature real-time monitoring system 100 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0066] Figure 5 Flow chart of the intelligent real-time temperature monitoring method according to the embodiment of the present application. Figure 5As shown, according to the intelligent body temperature real-time monitoring method of the embodiment of the present application, it includes: S110, acquiring the patient's infrared thermal imaging image collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database; S120, extracting the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector from the patient's infrared thermal imaging image collected by the infrared thermal imager and the patient's physiological environment text parameters collected by the database; S130, judging whether to issue a patient's abnormal body temperature warning based on the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector.

[0067] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned intelligent body temperature real-time monitoring method have been referred to above. Figures 1 to 4 The description of the intelligent body temperature real-time monitoring system has been introduced in detail, and therefore, its repeated description will be omitted.

Claims

1. An intelligent real-time body temperature monitoring system, characterized in that: include: A patient temperature data acquisition module is used to acquire the patient's infrared thermal imaging image acquired by the infrared thermal imager and the patient's physiological environment text parameters acquired by the database; A patient temperature data extraction module, used for extracting a patient infrared thermal imaging area tracking association feature vector and a patient physiological environment text semantic association feature vector from the patient infrared thermal imaging image acquired by the infrared thermal imager and the patient physiological environment text parameters acquired by the database; The patient temperature abnormality warning judgment module is used to determine whether to issue a patient temperature abnormality warning based on the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector.

2. The intelligent body temperature real-time monitoring system according to claim 1 is characterized in that: The patient temperature data extraction module comprises: A patient infrared thermal imaging feature extraction unit, used for extracting features from the patient infrared thermal imaging image acquired by the infrared thermal imager to obtain a tracking associated feature vector for the patient infrared thermal imaging area; The patient physiological environment text semantic encoding unit is used to semantically encode the patient physiological environment text parameters collected from the database to obtain the patient physiological environment text semantic association feature vector.

3. The intelligent body temperature real-time monitoring system according to claim 2 is characterized in that: The patient infrared thermal imaging feature extraction unit comprises: A patient infrared thermal imaging image division subunit, used for dividing the patient infrared thermal imaging image acquired by the infrared thermal imager into a plurality of patient infrared thermal imaging area tracking images; A patient infrared thermal imaging area tracking feature encoding subunit, used for performing feature encoding on the plurality of patient infrared thermal imaging area tracking images to obtain a patient infrared thermal imaging area tracking associated feature map; The global mean pooling subunit is used to perform global mean pooling on the patient's infrared thermal imaging area tracking associated feature map to obtain the patient's infrared thermal imaging area tracking associated feature vector.

4. The intelligent real-time body temperature monitoring system according to claim 3 is characterized in that: The patient infrared thermal imaging area tracking feature encoding subunit comprises: The plurality of patient infrared thermal imaging area tracking images are passed through a patient infrared thermal imaging area tracking convolutional neural network model with multi-feature extraction capability to obtain a plurality of patient infrared thermal imaging area tracking feature maps; Passing the plurality of patient infrared thermal imaging region tracking feature maps through a patient infrared thermal imaging region tracking target detection network to obtain a plurality of patient infrared thermal imaging tracking regions of interest; The multiple patient infrared thermal imaging tracking regions of interest are passed through a patient infrared thermal imaging region tracking significant target detector to obtain the patient infrared thermal imaging region tracking associated feature map.

5. The intelligent real-time body temperature monitoring system according to claim 4 is characterized in that: The patient physiological environment text semantic encoding unit includes: Passing the patient physiological environment text parameters collected from the database through a patient physiological environment text semantic encoder to obtain a plurality of patient physiological environment text semantic feature vectors; The multiple patient physiological environment text semantic feature vectors are cascaded into the patient physiological environment text semantic association feature vector.

6. The intelligent real-time body temperature monitoring system according to claim 5, characterized in that: The patient's abnormal body temperature early warning and judgment module includes: A patient temperature feature fusion unit is used to perform target domain adaptation adjustment on the patient infrared thermal imaging area tracking association feature vector and the patient physiological environment text semantic association feature vector based on a fine-grained semantic information field to obtain a patient temperature abnormality warning judgment feature vector; The abnormal body temperature warning classification unit is used to pass the patient's abnormal body temperature warning judgment feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a patient's abnormal body temperature warning.

7. The intelligent real-time body temperature monitoring system according to claim 6, characterized in that: The patient temperature feature fusion unit comprises: Mapping the patient infrared thermal imaging region tracking associated feature vector and the patient physiological environment text semantic associated feature vector to an intrinsic decomposition space to obtain a patient infrared thermal imaging region tracking associated intrinsic feature vector and a patient physiological environment text semantic associated intrinsic feature vector; Performing inter-domain feature calibration on the patient infrared thermal imaging region tracking associated intrinsic feature vector and the patient physiological environment text semantic associated intrinsic feature vector to obtain the patient infrared thermal imaging region tracking associated modulated intrinsic feature vector and the patient physiological environment text semantic associated modulated intrinsic feature vector; Constructing a multimodal fine-grained semantic information field of patient information between the intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient's physiological environment text semantic association modulation; Based on the multimodal fine-grained semantic information field of the patient information, the intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient's physiological environment text semantic association modulation are fused to obtain the patient's abnormal body temperature warning judgment feature vector.

8. The intelligent real-time body temperature monitoring system according to claim 7, characterized in that: Constructing a patient information multimodal fine-grained semantic information field between the intrinsic feature vector after the patient's infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient's physiological environment text semantic association modulation, including: The patient's infrared thermal imaging area tracking modulated intrinsic feature vector and the patient's physiological environment text semantic association modulated intrinsic feature vector are multiplied, and then divided by the scale of the patient's physiological environment text semantic association modulated intrinsic feature vector to obtain a multimodal fine-grained interaction matrix of patient information; The multimodal fine-grained matrix of patient information is convolutionally encoded to obtain the multimodal fine-grained semantic information field of patient information.

9. The intelligent real-time body temperature monitoring system according to claim 8, characterized in that: Based on the multimodal fine-grained semantic information field of the patient information, the intrinsic feature vector after the patient infrared thermal imaging area tracking association modulation and the intrinsic feature vector after the patient physiological environment text semantic association modulation are fused to obtain the patient's abnormal body temperature warning judgment feature vector, including: Multiplying the patient information multimodal fine-grained semantic information field by the patient infrared thermal imaging region tracking associated modulated intrinsic feature vector to obtain the patient infrared thermal imaging region tracking associated feature vector after information field mapping; Multiplying the patient information multimodal fine-grained semantic information field by the intrinsic feature vector after the patient physiological environment text semantic association modulation to obtain the patient physiological environment text semantic association feature vector after information field mapping; The patient's infrared thermal imaging area tracking associated feature vector after the information field mapping and the patient's physiological environment text semantic associated feature vector after the information field mapping are weightedly fused to obtain the patient's abnormal body temperature warning judgment feature vector.

10. An intelligent real-time body temperature monitoring method, characterized in that: include: Acquire the patient's infrared thermal imaging image acquired by the infrared thermal imager and the patient's physiological environment text parameters acquired by the database; Extracting the patient's infrared thermal imaging area tracking association feature vector and the patient's physiological environment text semantic association feature vector from the patient's infrared thermal imaging image acquired by the infrared thermal imager and the patient's physiological environment text parameters acquired by the database; Based on the patient's infrared thermal imaging area tracking associated feature vector and the patient's physiological environment text semantic associated feature vector, it is determined whether to issue a warning of abnormal patient temperature.

Images