Intelligent early warning method for febrile convulsion of children and related equipment

Abstract

The invention discloses an intelligent early warning method and related equipment for febrile convulsion of children, and relates to the field of intelligent medical treatment, and the method comprises the steps: obtaining body temperature information, chest wall temperature information, muscle electric signal information and heart rate information; the body temperature information, the chest wall temperature information, the muscle electric signal information and the heart rate information are input into an intelligent early warning model to obtain an intelligent early warning result, and the intelligent early warning model is obtained based on LSTM neural network training; and sending the intelligent prediction result to a target terminal. According to the embodiment of the invention, by introducing multi-channel depth time sequence modeling, a personalized time window, adaptive precursor mode learning and a multi-modal fusion mechanism, the sensitivity, accuracy and universality of a febrile convulsion early warning system are greatly improved, and more intelligent, more timely and more reliable health risk early warning support is provided for children.

Description

Technical Field

[0001] This specification relates to the field of smart healthcare, and more specifically, this application relates to a smart early warning method and related equipment for febrile seizures in children. Background Technology

[0002] Febrile seizures in children (also known as febrile convulsions) are an acute neurological reaction commonly seen in children aged 6 months to 5 years. They typically occur during a rapid rise in body temperature (≥38°C) and are diagnosed after ruling out other causes such as central nervous system infections and metabolic disorders. Typical manifestations include loss of consciousness, limb convulsions, and upward rolling of the eyes; some cases are accompanied by respiratory arrest and cyanosis. Most seizures resolve spontaneously within 3–5 minutes, but a few can develop into status epilepticus requiring emergency treatment. The global prevalence is 2%–5%, while the prevalence in Chinese children is approximately 3%–4%, with a peak incidence between 6 months and 3 years of age.

[0003] The risks of febrile seizures include not only accidental injuries and acute brain damage during the seizure, but also, in complex cases, the increased long-term risk of neurological disorders such as epilepsy, placing a significant psychological burden on families. However, current monitoring and early warning methods are clearly inadequate, mainly in the following aspects:

[0004] 1. High dependence on operation: Traditional thermometers require manual measurement, while non-contact devices are greatly affected by environmental interference and have low levels of automation;

[0005] 2. Limited data dimensions: Only static body temperature readings are provided, lacking dynamic features such as the rate of temperature rise and continuous trends;

[0006] 3. Lack of multi-parameter integration: Existing systems cannot integrate other key physiological signals such as heart rate and electromyography, making it difficult to conduct comprehensive risk assessment;

[0007] 4. Poor early warning timeliness: The current alarm for elevated body temperature cannot identify the characteristics of pre-seizure signs, and cannot achieve early intervention;

[0008] 5. Weak interaction and responsiveness: Limited user feedback and personalized settings make it difficult to meet the real-time and individual differences in family care needs.

[0009] Therefore, there is an urgent need to develop an intelligent early warning method for febrile seizures in children based on non-contact sensing, real-time continuous data acquisition, and multi-parameter fusion analysis, so as to realize the transformation from single-point temperature measurement to comprehensive monitoring, upgrade from post-event response to forward-looking prediction, and provide safer, smarter, and more efficient technical support for children's health management. Summary of the Invention

[0010] The summary section introduces a series of simplified concepts, which will be further explained in detail in the detailed description section. This summary section is not intended to limit the key and essential technical features of the claimed technical solution, nor is it intended to determine the scope of protection of the claimed technical solution.

[0011] Firstly, this application proposes an intelligent early warning method for febrile seizures in children, including:

[0012] Acquire body temperature, chest wall temperature, electromyographic signals, and heart rate information;

[0013] The above-mentioned body temperature information, chest wall temperature information, electromyography signal information and heart rate information are input into the intelligent early warning model to obtain intelligent early warning results. The intelligent early warning model is trained based on LSTM neural network.

[0014] The above intelligent prediction results are sent to the target terminal.

[0015] In one feasible implementation, the above-mentioned intelligent early warning model includes:

[0016] A multi-channel LSTM timing coding module is used to extract dynamic features of body temperature, chest wall temperature, electromyography signal, and heart rate information, respectively. The dynamic features include real-time body temperature features, body temperature change rate features, electromyography signal features, and heart rate features.

[0017] The adaptive time window module is used to determine the optimal monitoring period length based on the historical physiological data of different children;

[0018] An abnormal prodrome memory module is used to generate a combination pattern of prodromal febrile seizures among the above dynamic features;

[0019] The cross-modal fusion module is used to fuse the above-mentioned multi-channel features and establish multi-modal temporal correlation information;

[0020] The risk output module is used to output the risk level of febrile seizures in children based on the above-mentioned multimodal temporal correlation information and the above-mentioned febrile seizure precursor combination pattern.

[0021] In one feasible implementation, the above-mentioned abnormal precursor memory module includes:

[0022] The feature sliding window caching submodule is used to cache multimodal input feature sequences of body temperature, chest wall temperature, electromyography signal, and heart rate within a preset time period.

[0023] The prodrome discriminator submodule is used to calculate the prodrome score based on the above multimodal input feature sequence, determine whether there is a high-risk prodrome pattern, and generate a prodrome confirmation result when the risk threshold is exceeded.

[0024] The abnormal memory bank submodule is used to store high-risk prodromal feature patterns identified by the prodromal discriminator module and to establish a prodromal pattern index to generate the above-mentioned febrile seizure prodromal combination patterns.

[0025] In one feasible implementation, the aforementioned body temperature information is obtained based on a non-contact infrared sensing module.

[0026] The chest wall temperature and heart rate information mentioned above were obtained based on the data collected by the adhesive temperature acquisition module.

[0027] The aforementioned adhesive acquisition module includes a temperature sensor, a heart rate monitoring unit, and a flexible circuit board. The adhesive acquisition module is fixed to the second intercostal space along the midline of the sternum.

[0028] The aforementioned electromyographic (EMG) signal information was acquired through EMG electrode patches, which were applied to the biceps brachii, frontalis, and occipital muscles, respectively.

[0029] In one feasible implementation, sending the intelligent prediction result to the target terminal includes:

[0030] Obtain the risk level of the above intelligent prediction results;

[0031] Select the corresponding target terminal based on the aforementioned risk level;

[0032] The intelligent prediction results are sent to the target terminal so that the target terminal displays the intelligent prediction results and generates a prompt message corresponding to the intelligent prediction results.

[0033] In one feasible implementation, the above method further includes:

[0034] Obtain medical records and questionnaire information of the target children;

[0035] Based on the above intelligent prediction results, questionnaire information, and medical record information, suggested measures are generated.

[0036] In one feasible implementation, the specific steps for training the aforementioned intelligent early warning model include:

[0037] We acquired a historical time-series training dataset containing body temperature, chest wall temperature, electromyography, and heart rate information, and labeled the multimodal feature sequences within the corresponding time window with aura tags based on the occurrence time of the diagnosed seizure events.

[0038] Multimodal input samples are constructed based on a sliding window mechanism. These multimodal input samples include the time series of each feature and its rate of change features.

[0039] The above multimodal input samples are fed into a temporal modeling network constructed from multiple LSTM coding layers to extract feature evolution patterns;

[0040] A precursor discrimination layer is set at the feature output end, and key moments are weighted based on the attention mechanism, and the precursor score is output through a multilayer perceptron structure.

[0041] Backpropagation optimization is performed based on the loss between the aforementioned precursor scores and corresponding labels to complete the training of the intelligent early warning model.

[0042] Secondly, the present invention also proposes an intelligent early warning device for febrile seizures in children, comprising:

[0043] The first acquisition unit is used to acquire body temperature information, chest wall temperature information, electromyographic signal information, and heart rate information;

[0044] The second acquisition unit is used to input the above-mentioned body temperature information, chest wall temperature information, electromyography signal information and heart rate information into the intelligent early warning model to obtain intelligent early warning results, wherein the above-mentioned intelligent early warning model is based on training based on LSTM neural network;

[0045] The sending unit is used to send the above-mentioned intelligent prediction results to the target terminal.

[0046] Thirdly, the present invention also proposes an electronic device comprising: a memory and a processor, characterized in that the processor is configured to execute a computer program stored in the memory to implement the steps of the intelligent early warning method for febrile seizures in children as described in any of the first aspects.

[0047] Fourthly, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the intelligent early warning method for febrile seizures in children as described in any one of the first aspects.

[0048] In summary, the method proposed in this embodiment acquires body temperature, chest wall temperature, electromyography (EMG) signals, and heart rate information; inputs these information into an intelligent early warning model to obtain intelligent early warning results, wherein the intelligent early warning model is trained based on an LSTM neural network; and sends the intelligent prediction results to the target terminal. This implementation method, by introducing multi-channel deep temporal modeling, personalized time windows, adaptive precursor pattern learning, and multimodal fusion mechanisms, greatly improves the sensitivity, accuracy, and versatility of the febrile seizure early warning system, providing children with more intelligent, timely, and reliable health risk early warning support.

[0049] Other advantages, objectives and features of this application will be apparent in part from the description which follows, and in part from what those skilled in the art will understand through study and practice of this application. Attached Figure Description

[0050] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit this specification. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0051] Figure 1 This is a schematic flowchart of an intelligent early warning method for febrile seizures in children, provided in an embodiment of this application.

[0052] Figure 2 A structural schematic diagram of an intelligent early warning device for febrile seizures in children provided in this application embodiment;

[0053] Figure 3 This is a schematic diagram of an electronic device structure provided in an embodiment of this application. Detailed Implementation

[0054] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The technical solutions of the embodiments of this application will now be clearly and completely described in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0055] Please see Figure 1 This is a flowchart illustrating a method for intelligent early warning of febrile seizures in children, provided in an embodiment of this application. Specifically, it may include:

[0056] Firstly, this application proposes an intelligent early warning method for febrile seizures in children, including:

[0057] S110: Acquire body temperature information, chest wall temperature information, electromyographic signal information, and heart rate information;

[0058] S120. Input the above-mentioned body temperature information, chest wall temperature information, electromyography signal information and heart rate information into the intelligent early warning model to obtain intelligent early warning results. The intelligent early warning model is trained based on LSTM neural network.

[0059] S130. Send the above intelligent prediction results to the target terminal.

[0060] For example, in step S110, the system first collects various physiological signal data of the child in real time through a wearable device or a multi-source monitoring system, mainly including body temperature information, chest wall temperature information, electromyography (EMG) signal information, and heart rate information. Body temperature information is used to assess the current fever status, and chest wall temperature, as a supplement to the spatial distribution of body temperature changes, helps to more accurately determine the local and overall heat distribution; EMG signal reflects the activity level of the neuromuscular system and is an important indicator of the prodromal symptoms of seizures; heart rate information can reflect whether the child has potential symptoms such as abnormal sympathetic nerve excitation.

[0061] In step S120, the aforementioned multimodal physiological information is input into a trained intelligent early warning model. This model is constructed based on a Long Short-Term Memory (LSTM) neural network, which can effectively process time-series data. LSTM has the ability to remember changes in historical states, making it suitable for identifying subtle changes and potential patterns in children's physical signs during the development of febrile seizures. Through this model, key temporal features such as the rate of body temperature rise, electromyographic signal fluctuation characteristics, and abnormal heart rate rhythms can be extracted, and a comprehensive judgment can be made as to whether there are signs of febrile seizures.

[0062] In step S130, the system sends the early warning results output by the intelligent model to the target terminal in real time. The target terminal can be a parent's smartphone, a hospital nurse station terminal, or a remote monitoring platform. The early warning results can be presented in various forms such as graphics, text, or voice, including level prompts such as "Current status is safe," "Suspected signs of febrile seizures, please pay attention," or "High-risk warning, please intervene immediately," so that users can respond quickly and take appropriate medical or nursing measures.

[0063] In summary, this embodiment, by integrating multimodal physiological monitoring with an LSTM deep learning model, achieves early identification and proactive warning of the risk of febrile seizures in children. It has the advantages of strong real-time performance, high identification accuracy, and good adaptability, which helps to reduce irreversible damage to children's nervous system caused by seizures and improve the ability of families and medical institutions to respond to sudden health events.

[0064] In one feasible implementation, the above-mentioned intelligent early warning model includes:

[0065] A multi-channel LSTM timing coding module is used to extract dynamic features of body temperature, chest wall temperature, electromyography signal, and heart rate information, respectively. The dynamic features include real-time body temperature features, body temperature change rate features, electromyography signal features, and heart rate features.

[0066] The adaptive time window module is used to determine the optimal monitoring period length based on the historical physiological data of different children;

[0067] An abnormal prodrome memory module is used to generate a combination pattern of prodromal febrile seizures among the above dynamic features;

[0068] The cross-modal fusion module is used to fuse the above-mentioned multi-channel features and establish multi-modal temporal correlation information;

[0069] The risk output module is used to output the risk level of febrile seizures in children based on the above-mentioned multimodal temporal correlation information and the above-mentioned febrile seizure precursor combination pattern.

[0070] For example, this embodiment further structures the intelligent early warning model for febrile seizures in children, and constructs a deep neural network system with multi-level perception and fusion capabilities to more accurately and personally determine whether a child is at risk of seizures.

[0071] The multi-channel LSTM temporal coding module is the basic feature extraction unit of the model. It receives four types of physiological information: body temperature, chest wall temperature, electromyography (EMG), and heart rate, and configures an independent LSTM channel for each type of information.

[0072] Each channel can capture the time-series characteristics of its respective input data, thereby generating corresponding dynamic features: for example, the body temperature channel can extract the current real-time body temperature value and the rate of change of body temperature; the electromyography channel can identify potential abnormal muscle discharge patterns; and the heart rate channel can uncover heart rhythm fluctuations, beating rate, and rhythmic patterns. These dynamic features provide a fine-grained temporal signal basis for subsequent judgments.

[0073] The adaptive time window module takes into account the significant differences in each child's physiological baseline and response patterns. Based on each child's historical monitoring data, this module dynamically adjusts the length of the time window. Specifically, it utilizes statistical analysis or a sliding window self-optimization mechanism to determine the optimal time period for modeling analysis. This ensures sufficient information is covered to capture abrupt changes while avoiding noise interference caused by redundant data, thus improving the model's response speed and sensitivity.

[0074] The function of the abnormal prodromal memory module is to identify and store characteristic combination patterns that may represent prodromal symptoms of febrile seizures. Based on labeled seizure samples or automatically identified mutation signal sequences during continuous monitoring, this module constructs a prodromal memory bank, learning and solidifying characteristic combination patterns such as "rapid rise in body temperature + abnormal electromyography frequency + increased heart rate fluctuations" as representatives of typical high-risk patterns for subsequent early warning reference.

[0075] To achieve holistic judgment from multiple physiological dimensions, the cross-modal fusion module utilizes a deep fusion mechanism to uniformly model the features extracted from each channel. Through methods such as attention mechanisms, multilayer perceptrons, or Transformer structures, this module can learn the temporal interaction relationships between different modalities, such as whether changes in electromyography signals closely follow increases in body temperature, or whether heart rate fluctuations lag behind temperature peaks, thereby forming multimodal temporal features with stronger discriminative power.

[0076] The risk output module, as the model's output, is responsible for comparing the fused temporal features with the aura memory pattern to calculate the risk level of febrile seizures. The output can be categorized into multiple levels (e.g., low risk, medium risk, high risk) and can be quantitatively presented as a risk score or confidence level. For example, if the current features highly match a certain aura pattern and exhibit an abnormal trend, a "high risk" level is output, triggering a terminal alarm mechanism.

[0077] In summary, this implementation method, by introducing multi-channel deep temporal modeling, personalized time windows, adaptive precursor pattern learning, and multimodal fusion mechanisms, greatly improves the sensitivity, accuracy, and versatility of the febrile seizure early warning system, providing children with smarter, more timely, and more reliable health risk early warning support.

[0078] In one feasible implementation, the above-mentioned abnormal precursor memory module includes:

[0079] The feature sliding window caching submodule is used to cache multimodal input feature sequences of body temperature, chest wall temperature, electromyography signal, and heart rate within a preset time period.

[0080] The prodrome discriminator submodule is used to calculate the prodrome score based on the above multimodal input feature sequence, determine whether there is a high-risk prodrome pattern, and generate a prodrome confirmation result when the risk threshold is exceeded.

[0081] The abnormal memory bank submodule is used to store high-risk prodromal feature patterns identified by the prodromal discriminator module and to establish a prodromal pattern index to generate the above-mentioned febrile seizure prodromal combination patterns.

[0082] For example, the feature sliding window caching submodule serves as the input layer for aura analysis, responsible for caching multimodal physiological feature data within a certain time range in real time. The system presets a sliding time window (e.g., 30 seconds to 5 minutes). Whenever a new data stream arrives, the window slides forward and automatically updates its content, ensuring that the latest continuous feature sequence is always retained. This cached sequence covers four key data categories: body temperature, chest wall temperature, electromyography (EMG), and heart rate. This mechanism not only ensures the model possesses "short-term memory" capabilities but also provides input feature sequences with temporal continuity and semantic integrity for subsequent aura discrimination.

[0083] The prodromal warning discriminator submodule is the core computational unit for anomaly identification. Based on cached multimodal input feature sequences, it performs real-time assessment of the presence of prodromal febrile seizures. Specifically, this module uses a trained neural network or discriminant function to calculate the prodromal score of the current sequence. This score quantifies the similarity between the current physiological state and known prodromal febrile seizure patterns. If the score exceeds a system-preset risk threshold (e.g., 0.85), the sequence is considered to contain significant prodromal febrile seizure features, and the system immediately generates a confirmation result. This confirmation result is not only used for the current warning output but is also submitted to the next submodule for long-term learning and accumulation.

[0084] To enable continuous model evolution and individualized adaptation, the anomaly memory submodule stores high-risk feature patterns identified by the precursor discriminator. Each time a valid precursor is identified, this module extracts key multimodal feature fragments within the corresponding time window and saves them as an independent precursor pattern sample. Simultaneously, a precursor index is constructed using feature summarization algorithms (such as hash indexing and vector encoding), allowing subsequent input features to be quickly matched and compared with historical high-risk patterns.

[0085] As the system runs longer, the abnormal memory library will continue to expand, forming a dynamically updated library of combined patterns of febrile seizure precursors. This will enhance the model's risk identification capabilities and support individualized model training for different children.

[0086] This abnormal prodromal memory module possesses both short-term perception and long-term memory mechanisms. Through a three-step process—feature caching, risk assessment, and pattern memorization—it enables intelligent perception and rapid response to prodromal symptoms of febrile seizures in children. This module not only improves the model's timeliness and accuracy but also exhibits strong self-learning capabilities, continuously adapting to individual differences among children and possessing broad clinical early warning application value.

[0087] In one feasible implementation, the aforementioned body temperature information is obtained based on a non-contact infrared sensing module.

[0088] The chest wall temperature and heart rate information mentioned above were obtained based on the data collected by the adhesive temperature acquisition module.

[0089] The aforementioned adhesive acquisition module includes a temperature sensor, a heart rate monitoring unit, and a flexible circuit board. The adhesive acquisition module is fixed to the second intercostal space along the midline of the sternum.

[0090] The aforementioned electromyographic (EMG) signal information was acquired through EMG electrode patches, which were applied to the biceps brachii, frontalis, and occipital muscles, respectively.

[0091] For example, body temperature information is collected using a non-contact infrared sensing module. This module is typically placed above the child's forehead or ear and measures the skin surface temperature through infrared radiation to estimate the core body temperature. This method has the advantages of being non-invasive and having a rapid response, making it particularly suitable for nighttime or continuous monitoring scenarios. It does not disturb the child's rest and avoids the risk of cross-infection. The non-contact infrared sensing module for body temperature information can use the MLX90614 sensor, paired with a Fresnel lens to focus the infrared signal, measuring at a distance of 1-5 cm, collecting data 10 times per second, and taking the average value as the current body temperature.

[0092] Chest wall temperature and heart rate information are simultaneously collected by a single adhesive temperature acquisition module. This module is designed with a flexible patch structure and integrates the following:

[0093] Temperature sensor: used to collect local temperature of the chest wall in real time, which can reflect changes in the temperature of the deep parts of the body;

[0094] Heart rate monitoring unit: typically based on photoplethysmography (PPG) or electrocardiogram (ECG) to accurately extract heart rate and rhythm;

[0095] Flexible circuit board: Connects the above components and completes data acquisition and transmission, while maintaining the module's flexibility to adapt to the child's chest contour.

[0096] The module is attached to the second intercostal space along the midline of the sternum. This location is close to the heart and far from areas with excessive muscle interference, enabling stable acquisition of high-quality heart rate signals and chest wall temperature data, while also providing good wearing comfort and fixation.

[0097] Electromyography (EMG) signals are acquired using EMG electrode patches. The system design selects three key muscle groups as electrode attachment points:

[0098] Biceps brachii: Reflects the activity of upper limb muscles and can monitor changes in muscle tone that may occur in children under high fever.

[0099] The frontalis muscle is highly sensitive to facial neuromuscular responses and is an important observation point for early signs of seizures.

[0100] Occipital muscles: Located at the back of the neck, they help identify abnormal movement patterns in children while sleeping or lying on their backs.

[0101] These three muscle electrode sites are representative, covering the main areas of electrical activity in the trunk and face of children, and can effectively identify pre-seizure signals such as abnormal discharges or muscle tension.

[0102] The chest patch can use a medical-grade silicone base, with a built-in DS18B20 temperature sensor and PPG heart rate sensor (green LED + phototransistor), powered by a CR2032 button battery, with a battery life of ≥72 hours; the muscle electrode patch uses disposable conductive adhesive patches, with an electrode diameter of 10mm and a wire length of 30cm, supporting magnetic connection with terminal devices.

[0103] This implementation utilizes a multi-source sensor configuration to achieve comprehensive, continuous, and high-precision acquisition of physiological characteristics related to febrile seizures in children, such as changes in body temperature, heart rate fluctuations, and abnormal electromyographic signals. The sensor system combines non-contact measurement (body temperature) with adhesive-based precise measurement (chest wall temperature + heart rate + EMG), featuring a scientifically designed layout suitable for daily wear and nighttime monitoring. Furthermore, it fully considers children's physiological characteristics and behavioral habits, ensuring safety, comfort, and reliability during device use, providing a solid foundation of raw data for intelligent early warning models.

[0104] In one feasible implementation, sending the intelligent prediction result to the target terminal includes:

[0105] Obtain the risk level of the above intelligent prediction results;

[0106] Select the corresponding target terminal based on the aforementioned risk level;

[0107] The intelligent prediction results are sent to the target terminal so that the target terminal displays the intelligent prediction results and generates a prompt message corresponding to the intelligent prediction results.

[0108] For example, in one feasible implementation, to achieve timely response and personalized intervention for the risk of febrile seizures, this application introduces a risk grading and terminal adaptation strategy into the mechanism for sending and displaying intelligent prediction results. This process includes the following key steps:

[0109] S210. Obtaining the risk level of the intelligent prediction result: After the intelligent early warning model completes the analysis of multimodal physiological data, the system will output a prediction result. This result includes not only the probability value of seizures but also, based on a set risk threshold, classifying the probability into different risk levels, such as Level 1 alert and Level 2 alert. Specifically, this may include:

[0110] (1) Data acquisition stage:

[0111] An infrared sensor can measure chest wall temperature every minute, while a chest patch simultaneously collects core body temperature and heart rate signals; a muscle electrode patch continuously collects electromyographic signals and outputs an RMS characteristic value every 5 seconds.

[0112] (2) Parameter calculation stage:

[0113] Calculation of the heating rate: ΔT = current body temperature - body temperature 15 minutes ago, Δt = 15 minutes;

[0114] Calibration of the EMG baseline value: When the system is initialized, the EMG signals in the resting state are collected for 5 minutes, and the average RMS value of each electrode is calculated as the baseline.

[0115] (3) Risk assessment stage:

[0116] When the body temperature ≥ 38.5°C and ΔT / Δt ≥ 0.5°C / 15 min, a temperature-related warning is triggered;

[0117] Simultaneously detect the EMG RMS value and heart rate. If the RMS of any muscle group > 2 times the baseline value and the heart rate > the corresponding age threshold, a secondary convulsion risk warning is initiated.

[0118] (4) Intervention response stage:

[0119] When a primary alarm is triggered, the APP pushes antipyretic measures (such as physical cooling, reminder of antipyretics);

[0120] When a secondary alarm is triggered, the location information and warning notice are sent to the emergency contacts synchronously, and it is recommended to seek medical attention immediately.

[0121] This grading mechanism enables the system to treat physiological abnormalities of different severities differently, avoid frequent false alarms, and at the same time ensure the timely response to truly high-risk events.

[0122] S220. Select the target terminal according to the risk level. The system will automatically select the most suitable target terminal device to receive the current warning information according to the above risk level to improve the information transmission efficiency and the accuracy of intervention. Specifically, it can include: for a primary alarm, it can be displayed only on the locally worn device or the child care end (such as the parent's mobile phone); for a secondary alarm, in addition to notifying the parent terminal, it can also be synchronously uploaded to the health management platform or the child smart wearable cloud system, and sent to multiple terminals at the same time, including the terminals of parents, medical care personnel, and on-duty doctors, to implement a multi-point linkage response mechanism. This selection mechanism can be dynamically adjusted according to the child custody configuration, terminal online status, and user preferences.

[0123] S230. Send the intelligent prediction result to the target terminal and generate a prompt message. Once the target terminal is selected, the system immediately pushes the intelligent prediction result to the corresponding device. At the same time, the terminal will generate and display the corresponding prompt message according to the received risk level. The prompt content can include: when a primary alarm is triggered, the APP pushes antipyretic measures (such as physical cooling, reminder of antipyretics); when a secondary alarm is triggered, the location information and warning notice are sent to the emergency contacts synchronously, and it is recommended to seek medical attention immediately.

[0124] This implementation method achieves a precise distribution of risk information and a differentiated response mechanism by binding intelligent prediction results with risk levels, automatically matching target terminals, and generating graded prompt information based on the risk levels.

[0125] In one feasible implementation, the above method further includes:

[0126] Obtain medical records and questionnaire information of the target children;

[0127] Based on the above intelligent prediction results, questionnaire information, and medical record information, suggested measures are generated.

[0128] For example, in one feasible implementation, the intelligent early warning method for febrile seizures in children proposed in this application not only has the function of real-time monitoring of physiological data and risk assessment, but also further introduces individualized health records and behavioral information to improve the interpretability of early warning results and the pertinence of recommended measures.

[0129] Specifically, the system first obtains the target child's medical records and questionnaire information. Medical records include the child's history of seizures, chronic diseases, vaccinations, hospitalizations, and other relevant medical information. The questionnaire, completed by parents or medical personnel, covers individual behavioral and cognitive data such as genetic background, lifestyle habits, experience in managing fever, and family history of seizures. This information can be automatically obtained through hospital interfaces or updated regularly through platforms such as apps and follow-up systems to ensure data integrity and timeliness.

[0130] After the system generates intelligent prediction results (i.e., the risk level of febrile seizures) based on multimodal physiological data, it further combines the aforementioned medical records and questionnaire information for comprehensive analysis. The system uses constructed knowledge rules or intelligent decision-making models to cross-compare current physiological risks with historical health risk factors, thereby generating individualized intervention recommendations. For example, if a child is predicted to be at moderate risk and their questionnaire records a family history of seizures, the system will advise parents to increase monitoring frequency, prepare antipyretic medication in advance, and reduce stimulating behaviors. If the child is predicted to be at high risk and has a history of hospitalization for seizures, the system will directly recommend further observation at a medical institution, triggering a remote doctor notification mechanism if necessary.

[0131] The recommended measures will be sent to parents' devices or care platforms in real time, presented in various formats such as text, images, and voice, to help guardians understand the sources of risk and the steps to take. Furthermore, the system allows parents to provide feedback on the implementation of the recommendations after receiving them, thereby further refining the child's personalized health management model.

[0132] In summary, this implementation method, by incorporating medical history and behavioral data, elevates the risk prediction of febrile seizures in children from "single-point identification" to "personalized intervention." This not only improves the practicality of early warning but also enhances parents' confidence and accuracy in responding to sudden health events, achieving a more forward-looking strategy for protecting children's health.

[0133] In one feasible implementation, the specific steps for training the aforementioned intelligent early warning model include:

[0134] We acquired a historical time-series training dataset containing body temperature, chest wall temperature, electromyography, and heart rate information, and labeled the multimodal feature sequences within the corresponding time window with aura tags based on the occurrence time of the diagnosed seizure events.

[0135] Multimodal input samples are constructed based on a sliding window mechanism. These multimodal input samples include the time series of each feature and its rate of change features.

[0136] The above multimodal input samples are fed into a temporal modeling network constructed from multiple LSTM coding layers to extract feature evolution patterns;

[0137] A precursor discrimination layer is set at the feature output end, and key moments are weighted based on the attention mechanism, and the precursor score is output through a multilayer perceptron structure.

[0138] Backpropagation optimization is performed based on the loss between the aforementioned precursor scores and corresponding labels to complete the training of the intelligent early warning model.

[0139] For example, in one feasible implementation, the training process of the intelligent early warning model for febrile seizures in children proposed in this application fully combines the temporal characteristics of multimodal physiological data with deep learning modeling capabilities to construct an adaptive early warning system with precursor recognition capabilities. The specific training steps are as follows:

[0140] First, the system needs to acquire a historical time-series training dataset, which includes a large amount of physiological parameter information of children collected from real clinical or home monitoring, covering key variables such as body temperature, chest wall temperature, electromyography (EMG), and heart rate. For samples in the dataset diagnosed as seizures, the system extracts the corresponding multimodal feature sequences within a fixed time window (e.g., 5 or 10 minutes) preceding the precise occurrence time of the seizure event, and labels these sequences as "premonitory seizures"; other time periods are considered as non-premonitory samples. In this way, the training data is clearly divided into labeled positive and negative samples, ensuring that the model learns early feature patterns that are truly related to seizures.

[0141] Next, to fully capture the changing trends of multimodal signals over time, the system constructs input samples based on a sliding window mechanism. Specifically, various physiological data are slidably divided into multiple overlapping short sequence samples at fixed time steps. Each sample not only contains the original time series values ​​but also extracts its rate of change features (such as the rate of temperature rise, heart rate fluctuation, and electromyography amplitude changes), enhancing the model's ability to perceive dynamic trends.

[0142] Subsequently, a time-series modeling network consisting of multiple LSTM encoding layers was constructed, and the aforementioned multimodal input samples were fed into this network. LSTM (Long Short-Term Memory) networks can effectively capture long-term dependencies in time series, helping to discover potential evolutionary paths from stability to mutation, thereby learning the intrinsic laws governing the changes of different physiological characteristics over time.

[0143] At the model's output, a prodrome discrimination layer is set up for prodrome identification. This layer integrates an attention mechanism to automatically identify the most critical time segments in the time series for prodrome judgment and assign them higher weights. Subsequently, a multilayer perceptron (MLP) structure is used to perform a nonlinear mapping on the weighted time series features, ultimately outputting a prodrome score representing the "probability of prodromal seizures" in the current sequence.

[0144] Finally, the system compares the precursor score output by the model with the true label of the sample, measures the prediction error by calculating loss functions such as cross-entropy or mean squared error, and uses the backpropagation algorithm to optimize and update the parameters of the entire neural network. As the number of training epochs increases, the model can gradually converge and achieve high precursor recognition accuracy, thus enabling its application in real-world environments.

[0145] In summary, this training process, based on real seizure data, combines sliding window sample construction, LSTM temporal modeling, attention mechanism to select key moments, and multi-layer neural network to discriminate outputs, achieving efficient modeling and accurate identification of prodromal symptoms of febrile seizures in children. This lays a solid model foundation for building a reliable and intelligent early warning system.

[0146] Secondly, this invention also proposes an intelligent early warning device for febrile seizures in children, such as... Figure 2 As shown, it includes:

[0147] The first acquisition unit 21 is used to acquire body temperature information, chest wall temperature information, electromyographic signal information, and heart rate information;

[0148] The second acquisition unit 22 is used to input the above-mentioned body temperature information, chest wall temperature information, electromyography signal information and heart rate information into the intelligent early warning model to obtain intelligent early warning results, wherein the above-mentioned intelligent early warning model is trained based on LSTM neural network.

[0149] The sending unit 23 is used to send the above-mentioned intelligent prediction results to the target terminal.

[0150] The aforementioned intelligent early warning device for febrile seizures in children can also perform the following functions:

[0151] In one feasible implementation, the above-mentioned intelligent early warning model includes:

[0152] A multi-channel LSTM timing coding module is used to extract dynamic features of body temperature, chest wall temperature, electromyography signal, and heart rate information, respectively. The dynamic features include real-time body temperature features, body temperature change rate features, electromyography signal features, and heart rate features.

[0153] The adaptive time window module is used to determine the optimal monitoring period length based on the historical physiological data of different children;

[0154] An abnormal prodrome memory module is used to generate a combination pattern of prodromal febrile seizures among the above dynamic features;

[0155] The cross-modal fusion module is used to fuse the above-mentioned multi-channel features and establish multi-modal temporal correlation information;

[0156] The risk output module is used to output the risk level of febrile seizures in children based on the above-mentioned multimodal temporal correlation information and the above-mentioned febrile seizure precursor combination pattern.

[0157] In one feasible implementation, the above-mentioned abnormal precursor memory module includes:

[0158] The feature sliding window caching submodule is used to cache multimodal input feature sequences of body temperature, chest wall temperature, electromyography signal, and heart rate within a preset time period.

[0159] The prodrome discriminator submodule is used to calculate the prodrome score based on the above multimodal input feature sequence, determine whether there is a high-risk prodrome pattern, and generate a prodrome confirmation result when the risk threshold is exceeded.

[0160] The abnormal memory bank submodule is used to store high-risk prodromal feature patterns identified by the prodromal discriminator module and to establish a prodromal pattern index to generate the above-mentioned febrile seizure prodromal combination patterns.

[0161] In one feasible implementation, the aforementioned body temperature information is obtained based on a non-contact infrared sensing module.

[0162] The chest wall temperature and heart rate information mentioned above were obtained based on the data collected by the adhesive temperature acquisition module.

[0163] The aforementioned adhesive acquisition module includes a temperature sensor, a heart rate monitoring unit, and a flexible circuit board. The adhesive acquisition module is fixed to the second intercostal space along the midline of the sternum.

[0164] The aforementioned electromyographic (EMG) signal information was acquired through EMG electrode patches, which were applied to the biceps brachii, frontalis, and occipital muscles, respectively.

[0165] In one feasible implementation, sending the intelligent prediction result to the target terminal includes:

[0166] Obtain the risk level of the above intelligent prediction results;

[0167] Select the corresponding target terminal based on the aforementioned risk level;

[0168] The intelligent prediction results are sent to the target terminal so that the target terminal displays the intelligent prediction results and generates a prompt message corresponding to the intelligent prediction results.

[0169] In one feasible implementation, the above method further includes:

[0170] Obtain medical records and questionnaire information of the target children;

[0171] Based on the above intelligent prediction results, questionnaire information, and medical record information, suggested measures are generated.

[0172] In one feasible implementation, the specific steps for training the aforementioned intelligent early warning model include:

[0173] We acquired a historical time-series training dataset containing body temperature, chest wall temperature, electromyography, and heart rate information, and labeled the multimodal feature sequences within the corresponding time window with aura tags based on the occurrence time of the diagnosed seizure events.

[0174] Multimodal input samples are constructed based on a sliding window mechanism. These multimodal input samples include the time series of each feature and its rate of change features.

[0175] The above multimodal input samples are fed into a temporal modeling network constructed from multiple LSTM coding layers to extract feature evolution patterns;

[0176] A precursor discrimination layer is set at the feature output end, and key moments are weighted based on the attention mechanism, and the precursor score is output through a multilayer perceptron structure.

[0177] Backpropagation optimization is performed based on the loss between the aforementioned precursor scores and corresponding labels to complete the training of the intelligent early warning model.

[0178] Thirdly, the present invention also proposes an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, it implements the steps of the intelligent early warning method for febrile seizures in children as described in any of the first aspects.

[0179] Fourthly, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the intelligent early warning method for febrile seizures in children as described in any one of the first aspects.

[0180] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0181] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0182] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0183] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0184] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0185] This application also provides a computer program product, which includes computer software instructions. When the computer software instructions are executed on a processing device, the processing device performs the voice-based identity recognition process in the corresponding embodiment.

[0186] A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0187] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0188] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0189] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0190] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0191] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0192] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A child high fever convulsion intelligent early warning method, characterized in that, The method comprises the following steps: acquiring body temperature information, chest wall temperature information, muscle electrical signal information, and heart rate information; inputting the body temperature information, the chest wall temperature information, the muscle electrical signal information, and the heart rate information into an intelligent early warning model to obtain an intelligent early warning result, wherein the intelligent early warning model is trained based on an LSTM neural network; sending the intelligent prediction result to a target terminal.

2. The method of claim 1, wherein the method further comprises: The intelligent early warning model comprises: a multi-channel LSTM time sequence encoding module for extracting dynamic features of body temperature information, chest wall temperature information, muscle electrical signal information, and heart rate information, wherein the dynamic features comprise real-time body temperature features, body temperature change rate features, muscle electrical signal features, and heart rate features; an adaptive time window module for determining an optimal monitoring cycle length based on historical physiological data of different children; an abnormal precursor memory module for generating a high febrile convulsion precursor combination mode in the dynamic features; a cross-modal fusion module for fusing the multi-channel features and establishing multi-modal time sequence correlation information; a risk output module for outputting a child high febrile convulsion risk level result based on the multi-modal time sequence correlation information and the high febrile convulsion precursor combination mode.

3. The method of claim 2, wherein the method further comprises: The abnormal precursor memory module comprises: a feature sliding window cache submodule for caching multi-modal input feature sequences of body temperature information, chest wall temperature information, muscle electrical signal information, and heart rate information within a preset time length; a precursor discriminator submodule for calculating a convulsion precursor score based on the multi-modal input feature sequences, judging whether a high-risk precursor mode exists, and generating a precursor confirmation result when a preset risk threshold is exceeded; an abnormal memory bank submodule for storing high-risk precursor feature modes identified by the precursor discriminator module and establishing a precursor mode index to generate the high febrile convulsion precursor combination mode.

4. The method of claim 1, wherein the method further comprises: The body temperature information is acquired based on a non-contact infrared sensing module; The chest wall temperature information and the heart rate information are acquired based on a pasting type temperature acquisition module; The pasting type acquisition module comprises a temperature sensor, a heart rate monitoring unit, and a flexible circuit board, and is fixed to the second intercostal space of the sternum; The muscle electrical signal information is acquired by a muscle electrical signal electrode patch, which is attached to the biceps brachii, frontal muscle, and occipital muscle.

5. The method of claim 1, wherein the method further comprises: The method further comprises: acquiring medical record information and questionnaire information of a target child; generating suggestion measure information according to the intelligent prediction result, the questionnaire information, and the medical record information. The training steps of the intelligent early warning model comprise:

6. The method of claim 1, wherein the method further comprises: ​ ​ ​ 7. The method of claim 1, wherein the method further comprises: ​ A historical time series training data set containing body temperature information, chest wall temperature information, muscle electrical signal information and heart rate information is acquired, and a precursor label is labeled for a multi-modal feature sequence in a corresponding time window according to the occurrence time of a confirmed convulsion event; A multi-modal input sample is constructed based on a sliding window mechanism, and the multi-modal input sample includes a time series of each feature and a rate of change feature thereof; The multi-modal input sample is input into a time series modeling network constructed by a plurality of LSTM encoding layers to extract a feature evolution law; A precursor discrimination layer is arranged at a feature output end, a key moment is weighted based on an attention mechanism, and a precursor score is output through a multi-layer perception structure; Loss between the precursor score and a corresponding label is back-propagated and optimized to complete training of the intelligent early warning model.

8. A child high fever convulsion intelligent early warning device, characterized in that, Comprise: A first acquisition unit configured to acquire body temperature information, chest wall temperature information, muscle electrical signal information and heart rate information; A second acquisition unit configured to input the body temperature information, the chest wall temperature information, the muscle electrical signal information and the heart rate information into an intelligent early warning model to obtain an intelligent early warning result, wherein the intelligent early warning model is trained based on an LSTM neural network; A sending unit configured to send the intelligent prediction result to a target terminal.

9. An electronic device comprising: A memory and a processor, wherein the processor is configured to implement the steps of the child high fever convulsion intelligent early warning method according to any one of claims 1-7 when executing a computer program stored in the memory.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the child high fever convulsion intelligent early warning method according to any one of claims 1-7.

Images