Gas medicine inhalation control method and system for special patient

Abstract

The invention relates to the field of intelligent control of medical equipment, in particular to a gas medicine inhalation control method and system for a special patient. A gas drug inhalation control system for a special patient comprises a state recognition module, a physiological signal processing module, a gas supply control suggestion module and a scan trigger suggestion module. According to the invention, an intelligent decision-making process integrating a short time sequence prediction model and fuzzy logic reasoning is constructed, so that full-automatic closed loop from respiratory signal acquisition, phase recognition to gas drug supply control is realized; the system can predict the inspiration starting point in advance based on the real-time breathing waveform and dynamically adjust the gas contrast agent flow and the valve time sequence, a traditional operation mode depending on manual voice instructions and gesture feedback is replaced, the scanning failure rate caused by the matching degree difference of patients is remarkably reduced, and the scanning efficiency is improved. The medicine is especially suitable for children, old people, unconscious special patient groups and the like.

Description

Technical Field

[0001] This invention relates to the field of intelligent control of medical devices, and specifically to a method and system for controlling the inhalation of gaseous drugs for special patients. Background Technology

[0002] In medical imaging examinations such as CT and MRI, contrast-enhanced scanning is an important means to improve image contrast and diagnostic accuracy. In recent years, gas contrast enhancement technology (such as lung gas contrast CT / MRI) has been gradually promoted and applied, which observes changes in imaging signals of the lungs and related areas by having the subject inhale a gaseous contrast agent.

[0003] However, existing gas contrast agent inhalation control technology has significant drawbacks: First, its operation process heavily relies on manual intervention. The scanner needs to give instructions such as "inhale gas" and "hold your breath" to the subject via voice, and the subject then provides feedback through gestures. The operator then manually triggers the scan. This method is extremely unfriendly to special patients such as children, the elderly, or those with impaired consciousness, as it is difficult for them to respond to instructions accurately, resulting in a high failure rate. Second, the entire process lacks standardization, and the timing of the scan depends entirely on the operator's experience. The examination procedures vary greatly between different personnel and institutions, which seriously affects the comparability and repeatability of imaging results.

[0004] Therefore, this invention proposes a gaseous drug inhalation control method and system for special patients, which replaces manual operation and realizes fully automated control of the process, so as to solve the problems of large individual cooperation differences and non-standardized process in the prior art. Summary of the Invention

[0005] This invention constructs an intelligent decision-making process that integrates short-time prediction models and fuzzy logic reasoning, achieving a fully automated closed loop from respiratory signal acquisition and phase recognition to gas drug supply control. The system can predict the inspiratory start point in advance based on real-time respiratory waveforms and dynamically adjust the gas contrast agent flow rate and valve timing, replacing the traditional operation mode that relies on manual voice commands and gesture feedback. This significantly reduces the scan failure rate caused by differences in patient cooperation and is especially suitable for special patient groups such as children, the elderly, and those with impaired consciousness.

[0006] A method for controlling the inhalation of a gaseous drug for a specific patient includes: Step S1: Receive the control parameters set by the operator, including the scan trigger mode, breath-hold duration requirement, and delayed scan time; Collect the patient's respiratory waveform in a quiet state, extract the variance features of the mean and peak interval of the respiratory cycle, identify the patient's state type, and output a matching preset parameter template loading scheme accordingly. Step S2: Continuously and in parallel acquire the patient's respiratory airflow signal, electrocardiogram signal, and blood oxygen saturation signal; The ECG and blood oxygen saturation signals are analyzed in real time. When a preset abnormal physiological event characteristic is identified, a process interruption suggestion and alarm prompt are output. If no abnormal physiological event characteristics are identified, the respiratory airflow signal is filtered and input into the short-time prediction model to output the predicted time of the patient's next inspiratory start. Simultaneously, real-time respiratory rate, airflow pressure fluctuations, and heart rate variability parameters are input into the fuzzy logic reasoning process, and the identification result of the current respiratory phase and the corresponding confidence score are output. The respiratory phase includes the inspiratory phase, breath-holding phase, and expiratory phase. Step S3: Based on the inhalation start point predicted in step S2, generate a suggestion on the opening timing of the gas drug supply valve; Based on the recognition results output by the fuzzy logic reasoning process, a dynamic adjustment scheme for the gas drug supply flow rate is generated. Based on the identified breath-hold start signal, an operational suggestion to close the gas drug supply valve is generated; Based on the identified expiratory phase initiation signal, a switching suggestion for the gas emission pathway is generated; Step S4: Generate a suggested sequence of scan triggering times based on the preset scan triggering mode and the respiratory phase identified in step S2.

[0007] Preferably, in step S1, the variance feature of the interval between the mean and peak values ​​of the respiratory cycle is extracted, and the specific operation is as follows: Collect respiratory waveform data of the patient in a resting state to obtain a continuous respiratory airflow signal sequence; The respiratory airflow signal is preprocessed, including noise reduction and filtering. The following feature parameters were extracted from the preprocessed respiratory airflow signal: Mean respiratory cycle: Calculates the average duration of multiple consecutive respiratory cycles; Peak interval variance: Calculates the variance of the time interval between adjacent peaks in the respiratory waveform.

[0008] Preferably, the patient's condition type is identified based on the variance characteristics of the interval between the mean and peak values ​​of the respiratory cycle, as follows: Based on the extracted respiratory cycle mean and peak interval variance, the state type is identified according to the following rules: When the mean respiratory cycle is within the preset normal respiratory range and the variance of the peak interval is lower than the unconscious type threshold, it is identified as normal. When the mean respiratory cycle is less than the preset pediatric respiratory threshold and the variance of the peak interval is less than the unconscious type threshold, it is identified as pediatric type; When the mean respiratory cycle is greater than the preset respiratory threshold for the elderly and the variance of the peak interval is lower than the unconscious type threshold, it is identified as elderly. When the peak interval variance exceeds the preset unconsciousness threshold, it is identified as unconsciousness.

[0009] Preferably, the preset parameter template is constructed using the following method: Step T1: Establish a clinical data analysis set, which collects historical monitoring records from different patient groups. The records include respiratory waveform data, corresponding gas drug delivery flow rate pattern data, and successful case records of scan triggering timing. Step T2: Perform data cleaning and standardization preprocessing on the historical monitoring records to extract respiratory characteristic parameters for each independent case. The respiratory characteristic parameters include the mean respiratory cycle, the variance of the peak interval, and the proportion of inspiratory phase duration. Step T3: Based on the respiratory feature parameters extracted in step T2, the historical cases are divided into multiple categories using the K-means clustering algorithm, specifically including pediatric type, normal type, elderly type, and unconscious type; Step T4: Perform statistical analysis on success cases within each data category: For gaseous drug delivery modes, the flow rate variation characteristics in the inspiratory phase of such cases were analyzed, and the recommended flow rate parameter range was determined based on the 25th-75th percentile range. Regarding the timing of the scan trigger, the time characteristics from the start of breath-holding to successful scan triggering in this type of case are statistically analyzed, and the median is used to determine the baseline trigger delay parameter; Step T5: Combine the parameter ranges and baseline values ​​obtained from Step T4 to construct a preset parameter template corresponding to each patient status type.

[0010] Preferably, the short-time-series prediction model in step S2 is constructed and applied in the following manner: Step M1: Collect a historical dataset containing time series of respiratory airflow signals and corresponding inspiratory start markers. Use a sliding window to process the time series data to construct training samples. Each sample contains signal data from 2-3 complete respiratory cycles before the current moment. Step M2: A gated recurrent unit network is used as the core computational structure of the short-term prediction model. The network includes an input layer, two GRU hidden layers and an output layer. The output layer uses the Sigmoid activation function and the output value is a prediction confidence score between 0 and 1. Step M3: Using the mean squared error as the loss function, the Adam optimizer is used to optimize the network parameters. The training objective is to minimize the time error between the predicted inhalation start and the actual inhalation start. Step M4: Input the filtered respiratory airflow signal collected in real time into the trained prediction model. When the output confidence exceeds the preset threshold, generate the prediction result of the next inhalation start point.

[0011] Preferably, the construction and reasoning of the fuzzy logic reasoning process in step S2 includes: Step F1: Define the input variables and their membership functions, wherein the input variables include: The respiratory rate has its universe of discourse divided into three fuzzy sets: {slow, normal, fast}. The universe of discourse for airflow pressure fluctuations is divided into three fuzzy sets: {low, medium, and high}. Heart rate variability is divided into three fuzzy sets: {stable, moderate fluctuation, and violent fluctuation}. Step F2: Define the output variables and their membership functions, wherein the output variables include: The confidence level for respiratory phase recognition is divided into three fuzzy sets: {low, medium, and high}. The gas flow rate adjustment factor has its universe of discourse divided into three fuzzy sets: {decrease, maintain, increase}. Step F3: Establish a fuzzy rule base, constructed in the following way: Initial rules are extracted from the experience of clinical experts to form fuzzy rules in the form of "if-then"; Based on respiratory signals and successful control records from historical data, an adaptive neurofuzzy inference method is used to optimize the rules; The final rule base contains no fewer than 27 rules, covering various combinations of breathing states; Step F4: Blur the real-time collected respiratory rate, airflow pressure fluctuations, and heart rate variability parameters; Inference is performed using a fuzzy rule base and the Mamdani inference method. The centroid method is used to deblur the data, and the specific breathing phase recognition confidence score and gas flow rate adjustment factor value are output. Step F5: Compare the confidence score of the deblurred respiratory phase recognition with a preset threshold. When the confidence score is lower than 0.7, generate a re-recognition suggestion. Convert the gas flow rate adjustment factor value into an adjustment suggestion for the gas drug supply flow rate, which is used to generate the dynamic adjustment scheme in step S3.

[0012] Preferably, the specific implementation of generating the scan trigger timing suggestion sequence in step S4 includes: Step P1: Based on the respiratory phase recognition results and confidence score output in step S2, and combined with the preset scan trigger mode, generate scan trigger suggestions in the following manner: For single-scan mode, based on the time characteristics of the first breath-hold period reaching the preset breath-hold duration, corresponding trigger time point suggestions are generated; For the multi-scan mode, based on the identified start time characteristics of each breath-hold period, corresponding trigger time sequence suggestions are generated; For delayed scanning mode, based on the matching analysis of current patient respiratory characteristic parameters and historical data, a recommended delay time value and corresponding trigger time point suggestion are generated; Step P2: For the delayed scan mode, generate a trigger suggestion scheme through the following steps: Extract the current patient's mean respiratory cycle and breath-hold stability index as input features; Perform similarity matching analysis between the input features and a database of historical success cases; Based on the delay time distribution characteristics of similar cases, the recommended delay time interval is calculated; Output multiple alternative trigger time point suggestions based on the median of the recommended delay time interval; Step P3: Perform a quality assessment on each generated scan trigger suggestion: Taking into account respiratory phase recognition confidence, historical data matching degree, and signal stability index; When the overall confidence level is below 0.7, the trigger suggestion is marked as low reliability. When a sudden change in breathing pattern is detected, an instruction to recalculate the trigger suggestion is output.

[0013] A gaseous drug inhalation control system for special patients, comprising: The status recognition module includes: The data acquisition unit is used to acquire respiratory waveform data of the patient in a resting state; A state classification unit is used to extract the variance features of the mean and peak interval of the respiratory cycle from respiratory waveform data; based on the features, the patient state type is identified, and a matching preset parameter template loading scheme is output. The physiological signal processing module includes: The signal acquisition unit is used to continuously and in parallel acquire the patient's respiratory airflow signal, electrocardiogram signal, and blood oxygen saturation signal; The anomaly detection unit is used to perform real-time analysis of electrocardiogram signals and blood oxygen saturation signals. When a preset abnormal physiological event characteristic is identified, it outputs a process interruption suggestion and an alarm prompt. The respiratory prediction unit is used to filter the respiratory airflow signal and output the predicted time of the patient's next inspiratory start through a short-time prediction model. The fuzzy inference unit is used to perform fuzzy logic inference on real-time respiratory rate, airflow pressure fluctuation and heart rate variability parameters, and output the identification result of the current respiratory phase and the corresponding confidence score. The gas supply control suggestion module is used to generate suggestions on the opening timing of the gas drug supply valve based on the predicted inspiratory start point; generate a dynamic adjustment scheme for the gas drug supply flow rate based on the recognition results output by fuzzy logic reasoning; generate operation suggestions for closing the gas drug supply valve based on the identified breath-hold start signal; and generate suggestions for switching the gas emission pathway based on the identified expiration start signal. The scan trigger suggestion module is used to generate a suggested sequence of scan trigger times based on the preset scan trigger mode and the identified respiratory phase.

[0014] The present invention has the following beneficial effects: 1. This invention constructs an intelligent decision-making process that integrates short-time prediction models and fuzzy logic reasoning, realizing a fully automated closed loop from respiratory signal acquisition and phase recognition to gas drug supply control. The system can predict the inspiratory start point in advance based on real-time respiratory waveforms and dynamically adjust the gas contrast agent flow rate and valve timing, replacing the traditional operation mode that relies on manual voice commands and gesture feedback. This significantly reduces the scan failure rate caused by differences in patient cooperation and is especially suitable for special patient groups such as children, the elderly, and those with impaired consciousness.

[0015] 2. This invention achieves standardization of control procedures and personalization of decision-making by constructing a pre-set parameter template library based on clinical big data and an automatic patient status recognition mechanism, combined with real-time safety monitoring of multimodal physiological signals. The system can automatically classify patients according to their respiratory characteristics and load the optimal control parameters that match them. At the same time, by continuously analyzing ECG and blood oxygenation signals, it actively outputs interruption suggestions when abnormalities are detected. Thus, without the need for manual intervention, it provides a consistent and highly reliable inhalation control solution for special patients in different physiological states, effectively improving the standardization level and overall safety of the examination process. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the structure of a gaseous drug inhalation control system for special patients used in an embodiment of the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0018] Example 1 details a specific implementation of a gaseous drug inhalation control method for special patients. It should be noted that, in the context of medical imaging examinations, the "gaseous drug" specifically refers to a "gaseous contrast agent" used to enhance image contrast. The specific implementation process of this method, combined with an application example in a pulmonary pneumatic MRI examination scenario, is as follows: Step S1: Receive the control parameters set by the operator, including the scan trigger mode, breath-hold duration requirement and delayed scan time; collect the respiratory waveform of the patient in a quiet state, extract the variance features of the mean and peak interval of the respiratory cycle, identify the patient's state type, and output a matching preset parameter template loading scheme accordingly. The operator sets the control parameters for this examination through the user interface; for example, the scan trigger mode can be selected as "single scan", "multiple scans" or "delayed scan"; the breath-holding duration requirement can be set to 8 seconds to accommodate a complete lung scan; the delayed scan time can be set to 30 seconds to observe the distribution of gaseous contrast agent in the blood circulation; After identifying the patient's condition type, the system will automatically call the matching template from the preset parameter template library; for example, the template loaded for "children" will suggest using a smaller gas flow rate (such as 1.0-2.0 L / min) and a shorter breath-hold trigger window to accommodate their smaller vital capacity and faster respiratory metabolism. Step S2: Continuously and in parallel acquire the patient's respiratory airflow signal, electrocardiogram signal, and blood oxygen saturation signal; The ECG and blood oxygen saturation signals are analyzed in real time. When a preset abnormal physiological event characteristic is identified, a process interruption suggestion and alarm prompt are output. If no abnormal physiological event characteristics are identified, the respiratory airflow signal is filtered and input into the short-time prediction model to output the predicted time of the patient's next inspiratory start. Simultaneously, real-time respiratory rate, airflow pressure fluctuations, and heart rate variability parameters are input into the fuzzy logic reasoning process, and the identification result of the current respiratory phase and the corresponding confidence score are output. The respiratory phase includes the inspiratory phase, breath-holding phase, and expiratory phase. The fuzzy logic reasoning process is used to handle uncertainties and individual differences in respiratory signals. It reasons through a rule base containing dozens of "if-then" rules, considering fuzzy and imprecise input parameters such as "rapid respiratory rate," "moderate airflow pressure fluctuation," and "high heart rate variability." For example, a rule might be: "IF respiratory rate is fast AND airflow pressure fluctuation is high THEN respiratory phase recognition confidence is medium." Ultimately, it outputs a quantified "confidence score" (e.g., 0.85), which reflects the system's degree of certainty about the current respiratory phase (e.g., "breath-holding period"). When the score is low (e.g., below 0.7), the system considers the current judgment unreliable and thus relies on other signals or waits for a more definitive signal, enhancing the system's robustness. Step S3: Based on the inhalation start point predicted in step S2, generate a suggestion on the opening timing of the gas drug supply valve; Based on the recognition results output by the fuzzy logic reasoning process, a dynamic adjustment scheme for the gas drug supply flow rate is generated. Based on the identified breath-hold start signal, an operational suggestion is generated to close the gas drug supply valve to prevent drug waste and over-inhalation; Based on the identified expiratory phase initiation signal, a switching suggestion for the gas emission pathway is generated; Step S4: Generate a suggested sequence of scan triggering times based on the preset scan triggering mode and the respiratory phase identified in step S2; Specifically, the identification of preset abnormal physiological event characteristics includes: Step A1: ECG signal abnormality detection. ECG signals are analyzed in real time. An abnormal ECG event is determined when any of the following conditions are met: The detected ventricular premature beat frequency exceeded 10 times per minute; The occurrence of three or more consecutive premature ventricular contractions; A heart rate below 40 beats per minute or above 140 beats per minute is detected for more than 30 seconds. Step A2: Detection of abnormal blood oxygen saturation. Blood oxygen saturation values ​​are monitored in real time. An abnormal blood oxygenation event is determined when any of the following conditions are met: Blood oxygen saturation remains below 90% for more than 10 seconds; Blood oxygen saturation drops by more than 5% within 30 seconds; Step A3: Abnormal Event Response Handling. When the above-mentioned abnormal physiological events are detected, the following processing flow is executed: Immediately interrupt the gaseous drug supply and close all gas supply valves; Activate the audible and visual alarm device to alert medical staff to intervene; Display the specific exception type and the time of occurrence in the user interface; Record physiological signal data for 30 seconds before and after the abnormal event for subsequent analysis; Step A4: Recovery mechanism. After the abnormal event is handled, the system must be confirmed by medical staff and manually reset before the control process can be restarted.

[0019] The specific steps for extracting the mean and peak interval variance features of the respiratory cycle in step S1 to identify the patient's condition type are as follows: The respiratory waveform data of the patient in a quiet state is collected to obtain a continuous respiratory airflow signal sequence. In actual implementation, the patient needs to maintain quiet breathing for about 30 seconds, and the raw respiratory signal is collected by a medical-grade airflow sensor at a sampling rate of 100Hz. The respiratory airflow signal is preprocessed, including noise reduction and filtering. The following feature parameters were extracted from the preprocessed respiratory airflow signal: Mean respiratory cycle: Calculates the average duration of multiple consecutive respiratory cycles. This parameter directly reflects the respiratory rate characteristics. Peak interval variance: Calculates the variance of the time interval between adjacent peaks in the respiratory waveform. This parameter is used to quantify the stability of the respiratory rhythm. Based on the extracted respiratory cycle mean and peak interval variance, the state type is identified according to the following rules: When the mean respiratory cycle is within the preset normal breathing range (e.g., 3.0-5.0 seconds) and the variance of the peak interval is lower than the unconscious type threshold, it is identified as normal. When the mean respiratory cycle is less than the preset pediatric respiratory threshold (e.g., 3.0 seconds) and the variance of the peak interval is less than the unconscious type threshold, it is identified as pediatric type; When the mean respiratory cycle is greater than the preset respiratory threshold for the elderly (e.g., 5.0 seconds) and the variance of the peak interval is lower than the unconscious type threshold, it is identified as elderly. When the peak interval variance exceeds the preset unconsciousness threshold, it is identified as unconsciousness; The classification thresholds and ranges used in this method, including the normal breathing range, the pediatric / elderly breathing threshold, and the unconscious breathing threshold, are all set based on the clinical experience of professional medical personnel and refer to the consensus standards of respiratory physiology. For example, the normal adult resting respiratory rate range (12-20 breaths / min) corresponds to a respiratory cycle of 3.0-5.0 seconds; the setting of the pediatric breathing threshold takes into account the respiratory rate characteristics of children of different age groups; the differences in the stability thresholds reflect the clinical observation summary of the variability of breathing patterns in different populations.

[0020] The preset parameter template is constructed using the following method: Step T1: Establish a clinical data analysis set, which collects historical monitoring records from different patient groups. These data usually come from the radiology departments of multiple tertiary hospitals, covering patients of different ages, genders and pathological conditions. The records include respiratory waveform data, corresponding gas and drug delivery flow rate mode data, and successful case records of scan triggering timing. The criteria for a successful case is obtaining images of acceptable diagnostic quality and no adverse reactions in the patient. Step T2: Perform data cleaning and standardization preprocessing on the historical monitoring records, specifically including removing records with severely missing or interfered signals, and unifying the sampling rate and amplitude scale of all respiratory waveform data; extract respiratory characteristic parameters for each independent case, including respiratory cycle mean, peak interval variance, and inspiratory phase duration percentage, where the inspiratory phase duration percentage refers to the proportion of inspiratory time in a single respiratory cycle, which can effectively reflect the patient's respiratory pattern characteristics; Step T3: Based on the respiratory feature parameters extracted in step T2, the historical cases are divided into multiple categories using the K-means clustering algorithm. In the specific implementation, k=4 is selected for clustering. This algorithm is based on Euclidean distance metric and automatically groups patients with similar respiratory features through iterative optimization. The four categories formed are highly correlated with the pediatric, normal, elderly and unconscious types. Step T4: Perform statistical analysis on success cases within each data category: For gaseous drug delivery modes, the flow rate variation characteristics in the inspiratory phase of such cases were analyzed, and the recommended flow rate parameter range was determined based on the 25th-75th percentile range. Regarding the timing of the scan trigger, the time characteristics from the start of breath-holding to successful scan triggering in this type of case are statistically analyzed, and the median is used to determine the baseline trigger delay parameter; Step T5: Combine the parameter ranges and baseline values ​​obtained from Step T4. For example, the template constructed for "children" includes parameters such as flow range of 1.5-2.0 L / min and baseline trigger delay of 1.2 seconds. Construct preset parameter templates corresponding to each patient status type. These templates can be fine-tuned according to specific circumstances in actual applications, but they provide scientific data support for the initial settings.

[0021] The short-time-series prediction model in step S2 is constructed and applied in the following ways: Step M1: Collect a historical dataset containing time series of respiratory airflow signals and corresponding inspiratory start markers. Use a sliding window to construct training samples from the time series data. Each sample contains signal data from 2-3 complete respiratory cycles before the current moment. For example, for a patient with a respiratory cycle of 4 seconds, each training sample covers a continuous signal of 8-12 seconds. Step M2: A gated recurrent unit network (GRU) is used as the core computational structure of the prediction model. GRU is chosen because it can effectively capture the temporal dependencies in respiratory signals and has higher computational efficiency than traditional LSTM. The network consists of an input layer, two GRU hidden layers, and an output layer. The output layer uses the sigmoid activation function, and the output value is a prediction confidence score between 0 and 1. This score directly reflects the model's confidence in the upcoming inspiratory initiation. Step M3: Use mean squared error as the loss function, which can effectively measure the deviation between the predicted time point and the actual time point; use the Adam optimizer to optimize the network parameters, set the initial learning rate to 0.001, and use a learning rate decay strategy; the training objective is to minimize the time error between the predicted inhalation start point and the actual inhalation start point. Step M4: Input the filtered respiratory airflow signal collected in real time into the trained prediction model. When the output confidence exceeds the preset threshold, the threshold can be set to 0.8 to minimize false negatives while ensuring accuracy, and generate the prediction result of the next inhalation start point.

[0022] The construction and reasoning of the fuzzy logic reasoning process in step S2 includes: Step F1: Define the input variables and their membership functions, wherein the input variables include: The respiratory rate is divided into three fuzzy sets: {slow, normal, fast}. Specifically, a respiratory rate of <12 breaths per minute is defined as "slow", 12-20 breaths per minute is "normal", and >20 breaths per minute is "fast". A triangular membership function is used to achieve a smooth transition. The airflow pressure fluctuation is divided into three fuzzy sets: {low, medium, and high}. This parameter is obtained by calculating the standard deviation of the respiratory signal within a 0.5-second window and can be set to correspond to three intervals: <0.5 kPa, 0.5-1.0 kPa, and >1.0 kPa, respectively. Heart rate variability is divided into three fuzzy sets: {stable, moderate fluctuation, and violent fluctuation}. The division is based on the standard deviation of heart rate (SDNN) and can be set to correspond to SDNN < 50ms, 50-100ms, and > 100ms, respectively. Step F2: Define the output variables and their membership functions, wherein the output variables include: The confidence level for respiratory phase recognition is divided into three fuzzy sets: {low, medium, and high}. The output range is 0-1, which can be set to correspond to the three intervals [0, 0.6], (0.6, 0.8], and (0.8, 1.0], respectively. The gas flow rate adjustment factor has its universe of discourse divided into three fuzzy sets: {decrease, maintain, increase}, which can be set to correspond to flow rate adjustment from -30% to 0%, maintaining the original state, and increasing from 0% to +30%, respectively. Step F3: Establish a fuzzy rule base, constructed in the following way: Initial rules are extracted from the experience of clinical experts to form fuzzy rules in the form of "if-then", such as: "If the respiratory rate is fast and the airflow pressure fluctuation is high, then the confidence level of respiratory phase identification is medium". Based on respiratory signals and successful control records from historical data, an adaptive neurofuzzy inference method is used to optimize the rules. This method automatically adjusts the membership function parameters through the backpropagation algorithm, making the rules more consistent with the actual breathing pattern. The final rule base contains no fewer than 27 rules, covering various combinations of breathing states, ensuring that reasonable output suggestions can be generated for all possible combinations of inputs; Step F4: Fuzzyize the real-time collected respiratory rate, airflow pressure fluctuation and heart rate variability parameters, and convert the precise input values ​​into the membership degrees of the corresponding fuzzy sets; The reasoning is performed using a fuzzy rule base and the Mamdani reasoning method is adopted, which synthesizes the output of each rule through min-max operations. By using the centroid method for defuzzification, specific respiratory phase recognition confidence scores and gas flow rate adjustment factor values ​​are output, for example, a confidence score of 0.75 and a flow rate adjustment factor of +0.15. Step F5: Compare the confidence score of the deblurred respiratory phase recognition with a preset threshold. When the confidence score is lower than 0.7, generate a re-recognition suggestion. At this time, the system will prompt the operator to confirm or wait for a clearer respiratory signal. Convert the gas flow rate adjustment factor value into an adjustment suggestion for the gas drug supply flow rate. For example, if the current basic flow rate is 2.0 L / min, the adjustment factor +0.15 corresponds to the suggestion to increase the flow rate to 2.3 L / min, which is used to generate the dynamic adjustment plan in step S3.

[0023] The specific implementation of generating the scan trigger timing suggestion sequence in step S4 includes: Step P1: Based on the respiratory phase recognition results and confidence score output in step S2, and combined with the preset scan trigger mode, generate scan trigger suggestions in the following manner: For single-scan mode, based on the time characteristics of the first breath-hold period reaching the preset breath-hold duration, corresponding trigger time point suggestions are generated; For the multi-scan mode, based on the identified start time characteristics of each breath-hold period, corresponding trigger time sequence suggestions are generated; For delayed scanning mode, based on the matching analysis of the current patient's respiratory characteristic parameters and historical data, the current patient's respiratory cycle, tidal volume and other parameters are compared with successful cases in the database to generate a recommended delay time value and corresponding trigger time point suggestions. Step P2: For the delayed scan mode, generate a trigger suggestion scheme through the following steps: The mean respiratory cycle and breath-hold stability index of the current patient are extracted as input features, where breath-hold stability is quantified by the coefficient of variation of continuous breath-hold duration; The input features are matched with a database of historical success cases for similarity analysis. The Euclidean distance metric is used to select the top 10% of cases with the highest similarity as the reference set. Based on the distribution characteristics of delay time in similar cases, a recommended delay time interval is calculated. For example, for a study on pulmonary gas exchange, the recommended interval may be 30-60 seconds. The output suggests multiple alternative trigger times based on the median of the recommended delay time interval, with each time point spaced 5-10 seconds apart. Step P3: Perform a quality assessment on each generated scan trigger suggestion: Taking into account respiratory phase recognition confidence, historical data matching degree, and signal stability index; When the overall confidence level is below 0.7, the trigger suggestion is marked as low reliability. When a sudden change in breathing pattern is detected, an instruction to recalculate the trigger suggestion is output.

[0024] Example 2, a gaseous drug inhalation control system for special patients, such as Figure 1 As shown, it includes: The status recognition module includes: The data acquisition unit is used to acquire respiratory waveform data of the patient in a resting state; A state classification unit is used to extract the variance features of the mean and peak interval of the respiratory cycle from respiratory waveform data; based on the features, the patient state type is identified, and a matching preset parameter template loading scheme is output. The physiological signal processing module includes: The signal acquisition unit is used to continuously and in parallel acquire the patient's respiratory airflow signal, electrocardiogram signal, and blood oxygen saturation signal; The anomaly detection unit is used to perform real-time analysis of electrocardiogram signals and blood oxygen saturation signals. When a preset abnormal physiological event characteristic is identified, it outputs a process interruption suggestion and an alarm prompt. The respiratory prediction unit is used to filter the respiratory airflow signal and output the predicted time of the patient's next inspiratory start through a short-time prediction model. The fuzzy inference unit is used to perform fuzzy logic inference on real-time respiratory rate, airflow pressure fluctuation and heart rate variability parameters, and output the identification result of the current respiratory phase and the corresponding confidence score. The gas supply control suggestion module is used to generate suggestions on the opening timing of the gas drug supply valve based on the predicted inspiratory start point; generate a dynamic adjustment scheme for the gas drug supply flow rate based on the recognition results output by fuzzy logic reasoning; generate operation suggestions for closing the gas drug supply valve based on the identified breath-hold start signal; and generate suggestions for switching the gas emission pathway based on the identified expiration start signal. The scan trigger suggestion module is used to generate a suggested sequence of scan trigger times based on the preset scan trigger mode and the identified respiratory phase.

[0025] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for controlling the inhalation of gaseous medications for special patients, characterized in that, include: Step S1: Receive the control parameters set by the operator, including the scan trigger mode, breath-hold duration requirement, and delayed scan time; Collect the patient's respiratory waveform in a quiet state, extract the variance features of the mean and peak interval of the respiratory cycle, identify the patient's state type, and output a matching preset parameter template loading scheme accordingly. Step S2: Continuously and in parallel acquire the patient's respiratory airflow signal, electrocardiogram signal, and blood oxygen saturation signal; The ECG and blood oxygen saturation signals are analyzed in real time. When a preset abnormal physiological event characteristic is identified, a process interruption suggestion and alarm prompt are output. If no abnormal physiological event characteristics are identified, the respiratory airflow signal is filtered and input into the short-time prediction model to output the predicted time of the patient's next inspiratory start. Simultaneously, real-time respiratory rate, airflow pressure fluctuations, and heart rate variability parameters are input into the fuzzy logic reasoning process, and the identification result of the current respiratory phase and the corresponding confidence score are output. The respiratory phase includes the inspiratory phase, breath-holding phase, and expiratory phase. Step S3: Based on the inhalation start point predicted in step S2, generate a suggestion on the opening timing of the gas drug supply valve; Based on the recognition results output by the fuzzy logic reasoning process, a dynamic adjustment scheme for the gas drug supply flow rate is generated. Based on the identified breath-hold start signal, an operational suggestion to close the gas drug supply valve is generated; Based on the identified expiratory phase initiation signal, a switching suggestion for the gas emission pathway is generated; Step S4: Generate a suggested sequence of scan triggering times based on the preset scan triggering mode and the respiratory phase identified in step S2.

2. The method for controlling the inhalation of gaseous drugs for special patients according to claim 1, characterized in that, In step S1, the variance feature of the interval between the mean and peak values ​​of the respiratory cycle is extracted. The specific operation is as follows: Collect respiratory waveform data of the patient in a resting state to obtain a continuous respiratory airflow signal sequence; The respiratory airflow signal is preprocessed, including noise reduction and filtering. The following feature parameters were extracted from the preprocessed respiratory airflow signal: Mean respiratory cycle: Calculates the average duration of multiple consecutive respiratory cycles; Peak interval variance: Calculates the variance of the time interval between adjacent peaks in the respiratory waveform.

3. The method for controlling the inhalation of gaseous drugs for special patients according to claim 2, characterized in that, The patient status type is identified based on the variance characteristics of the interval between the mean and peak values ​​of the respiratory cycle. The specific operation is as follows: Based on the extracted respiratory cycle mean and peak interval variance, the state type is identified according to the following rules: When the mean respiratory cycle is within the preset normal respiratory range and the variance of the peak interval is lower than the unconscious type threshold, it is identified as normal. When the mean respiratory cycle is less than the preset pediatric respiratory threshold and the variance of the peak interval is less than the unconscious type threshold, it is identified as pediatric type; When the mean respiratory cycle is greater than the preset respiratory threshold for the elderly and the variance of the peak interval is lower than the unconscious type threshold, it is identified as elderly. When the peak interval variance exceeds the preset unconsciousness threshold, it is identified as unconsciousness.

4. The method for controlling the inhalation of gaseous drugs for special patients according to claim 3, characterized in that, The preset parameter template is constructed using the following method: Step T1: Establish a clinical data analysis set, which collects historical monitoring records from different patient groups. The records include respiratory waveform data, corresponding gas drug delivery flow rate pattern data, and successful case records of scan triggering timing. Step T2: Perform data cleaning and standardization preprocessing on the historical monitoring records to extract respiratory characteristic parameters for each independent case. The respiratory characteristic parameters include the mean respiratory cycle, the variance of the peak interval, and the proportion of inspiratory phase duration. Step T3: Based on the respiratory feature parameters extracted in step T2, the historical cases are divided into multiple categories using the K-means clustering algorithm, specifically including pediatric type, normal type, elderly type, and unconscious type; Step T4: Perform statistical analysis on success cases within each data category: For gaseous drug delivery modes, the flow rate variation characteristics in the inspiratory phase of such cases were analyzed, and the recommended flow rate parameter range was determined based on the 25th-75th percentile range. Regarding the timing of the scan trigger, the time characteristics from the start of breath-holding to successful scan triggering in this type of case are statistically analyzed, and the median is used to determine the baseline trigger delay parameter; Step T5: Combine the parameter ranges and baseline values ​​obtained from Step T4 to construct a preset parameter template corresponding to each patient status type.

5. The method for controlling the inhalation of gaseous drugs for special patients according to claim 4, characterized in that, The short-time-series prediction model in step S2 is constructed and applied in the following ways: Step M1: Collect a historical dataset containing time series of respiratory airflow signals and corresponding inspiratory start markers. Use a sliding window to process the time series data to construct training samples. Each sample contains signal data from 2-3 complete respiratory cycles before the current moment. Step M2: A gated recurrent unit network is used as the core computational structure of the short-term prediction model. The network includes an input layer, two GRU hidden layers and an output layer. The output layer uses the Sigmoid activation function and the output value is a prediction confidence score between 0 and 1. Step M3: Using the mean squared error as the loss function, the Adam optimizer is used to optimize the network parameters. The training objective is to minimize the time error between the predicted inhalation start and the actual inhalation start. Step M4: Input the filtered respiratory airflow signal collected in real time into the trained prediction model. When the output confidence exceeds the preset threshold, generate the prediction result of the next inhalation start point.

6. A method for controlling the inhalation of gaseous drugs for special patients according to claim 5, characterized in that, The construction and reasoning of the fuzzy logic reasoning process in step S2 includes: Step F1: Define the input variables and their membership functions, wherein the input variables include: The respiratory rate has its universe of discourse divided into three fuzzy sets: {slow, normal, fast}. The universe of discourse for airflow pressure fluctuations is divided into three fuzzy sets: {low, medium, and high}. Heart rate variability is divided into three fuzzy sets: {stable, moderate fluctuation, and violent fluctuation}. Step F2: Define the output variables and their membership functions, wherein the output variables include: The confidence level for respiratory phase recognition is divided into three fuzzy sets: {low, medium, and high}. The gas flow rate adjustment factor has its universe of discourse divided into three fuzzy sets: {decrease, maintain, increase}. Step F3: Establish a fuzzy rule base, constructed in the following way: Initial rules are extracted from the experience of clinical experts to form fuzzy rules in the form of "if-then"; Based on respiratory signals and successful control records from historical data, an adaptive neurofuzzy inference method is used to optimize the rules; The final rule base contains no fewer than 27 rules, covering various combinations of breathing states; Step F4: Blur the real-time collected respiratory rate, airflow pressure fluctuations, and heart rate variability parameters; Inference is performed using a fuzzy rule base and the Mamdani inference method. The centroid method is used to deblur the data, and the specific breathing phase recognition confidence score and gas flow rate adjustment factor value are output. Step F5: Compare the confidence score of the deblurred respiratory phase recognition with a preset threshold. When the confidence score is lower than 0.7, generate a re-recognition suggestion. Convert the gas flow rate adjustment factor value into an adjustment suggestion for the gas drug supply flow rate, which is used to generate the dynamic adjustment scheme in step S3.

7. A method for controlling the inhalation of gaseous drugs for special patients according to claim 6, characterized in that, The specific implementation of generating the scan trigger timing suggestion sequence in step S4 includes: Step P1: Based on the respiratory phase recognition results and confidence score output in step S2, and combined with the preset scan trigger mode, generate scan trigger suggestions in the following manner: For single-scan mode, based on the time characteristics of the first breath-hold period reaching the preset breath-hold duration, corresponding trigger time point suggestions are generated; For the multi-scan mode, based on the identified start time characteristics of each breath-hold period, corresponding trigger time sequence suggestions are generated; For delayed scanning mode, based on the matching analysis of current patient respiratory characteristic parameters and historical data, a recommended delay time value and corresponding trigger time point suggestion are generated; Step P2: For the delayed scan mode, generate a trigger suggestion scheme through the following steps: Extract the current patient's mean respiratory cycle and breath-hold stability index as input features; Perform similarity matching analysis between the input features and a database of historical success cases; Based on the delay time distribution characteristics of similar cases, the recommended delay time interval is calculated; Output multiple alternative trigger time point suggestions based on the median of the recommended delay time interval; Step P3: Perform a quality assessment on each generated scan trigger suggestion: Taking into account respiratory phase recognition confidence, historical data matching degree, and signal stability index; When the overall confidence level is below 0.7, the trigger suggestion is marked as low reliability. When a sudden change in breathing pattern is detected, an instruction to recalculate the trigger suggestion is output.

8. A gaseous drug inhalation control system for special patients, characterized in that, The system is applied to a gaseous drug inhalation control method for special patients as described in any one of claims 1-7, comprising: The status recognition module includes: The data acquisition unit is used to acquire respiratory waveform data of the patient in a resting state; A state classification unit is used to extract the variance features of the mean and peak interval of the respiratory cycle from respiratory waveform data; based on the features, the patient state type is identified, and a matching preset parameter template loading scheme is output. The physiological signal processing module includes: The signal acquisition unit is used to continuously and in parallel acquire the patient's respiratory airflow signal, electrocardiogram signal, and blood oxygen saturation signal; The anomaly detection unit is used to perform real-time analysis of electrocardiogram signals and blood oxygen saturation signals. When a preset abnormal physiological event characteristic is identified, it outputs a process interruption suggestion and an alarm prompt. The respiratory prediction unit is used to filter the respiratory airflow signal and output the predicted time of the patient's next inspiratory start through a short-time prediction model. The fuzzy inference unit is used to perform fuzzy logic inference on real-time respiratory rate, airflow pressure fluctuation and heart rate variability parameters, and output the identification result of the current respiratory phase and the corresponding confidence score. The gas supply control suggestion module is used to generate suggestions on the opening timing of the gas drug supply valve based on the predicted inspiratory start point; generate a dynamic adjustment scheme for the gas drug supply flow rate based on the recognition results output by fuzzy logic reasoning; generate operation suggestions for closing the gas drug supply valve based on the identified breath-hold start signal; and generate suggestions for switching the gas emission pathway based on the identified expiration start signal. The scan trigger suggestion module is used to generate a suggested sequence of scan trigger times based on the preset scan trigger mode and the identified respiratory phase.

Images