Artificial intelligence data set construction method and apparatus related to cardiac contrast agent injection

Abstract

An artificial intelligence data set construction method and apparatus related to cardiac contrast agent injection. The method comprises: measuring an expiration pressure of a patient to be tested (600); measuring electrocardiogram changes of said patient; determining, by a control unit (500), an injection timing of a contrast agent injection unit (300) and sending a recording instruction for a hemodynamic condition in a heart; when a time point at which the expiration pressure reaches a first predetermined threshold, instantly monitoring, by the control unit, a sudden change amplitude of an electrocardiogram change rate; when the sudden change amplitude exceeds a change amplitude from a time point at which expiration starts to the time point at which the expiration pressure reaches the first predetermined threshold, determining a time point at which the sudden change amplitude occurs as a change inflection point of the electrocardiogram change rate; after determining the change inflection point of the electrocardiogram change rate, sending, by the control unit, an injection start signal to the contrast agent injection unit; and marking, by an artificial intelligence semantic training model (540) in the control unit, the electrocardiogram change rate of said patient sent by an electrocardiogram monitoring unit (200), thereby forming a training set of the electrocardiogram change rate.

Description

An artificial intelligence dataset construction method and device related to cardiac contrast agent injection TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent medical treatment, in particular to an artificial intelligence dataset construction method and device related to cardiac contrast agent injection, and also relates to a patent foramen ovale auxiliary diagnosis device based on Valsalva action. BACKGROUND

[0002] The oval foramen is a physiological passageway of the atrial septum of the heart during the embryonic period. After birth, due to the increase in left atrial pressure and the decrease in pulmonary artery resistance, the two parts of the atrial septum gradually fuse, and most people close spontaneously within one year. If not closed, a patent foramen ovale (PFO) is formed in the atrial septum. In order to diagnose PFO, a provocation test is usually used to increase the diagnostic sensitivity. Among them, the Valsalva action is the most commonly used provocation test, which increases the chest pressure, reduces the venous return to the heart, and increases the right atrial pressure, thereby more obviously improving the diagnostic sensitivity of PFO. However, in critical patients, intubated patients and patients undergoing sedation TEE (transesophageal echocardiography), it is difficult to achieve a full Valsalva action.

[0003] During the diagnosis process, cardiac ultrasound contrast (ASCE) is an important tool. The standard for ASCE to diagnose PFO is to observe the contrast agent (ASC) passing through the oval foramen by cardiac ultrasound. Specifically, after the right atrium is filled with ASC, if ASC appears in the left atrium within 3-6 cardiac cycles, it can also be used as an alternative diagnostic indicator for PFO. This standard is applicable to cases where the shunt is not clearly displayed or the shunt is located in a non-current section, ensuring the accuracy and reliability of the diagnosis.

[0004] During the diagnosis of patent foramen ovale (PFO) using ASCE (cardiac ultrasound contrast), the accuracy of contrast agent injection directly affects the accuracy of diagnosis, the sensitivity of detection, the safety and comfort of patients, the utilization of medical resources and work efficiency. If the contrast agent fails to fully enter the target area, it may result in a small PFO being missed. Non-standard injection may cause uneven distribution of contrast agent in the heart cavity, producing artifacts or false signals, leading to misdiagnosis. Excessive contrast agent may increase the risk of adverse reactions such as allergic reactions, kidney function impairment, etc. In order to ensure the efficiency and reliability of the diagnosis, it is necessary to strictly control each link of contrast agent injection to ensure its standardization and accuracy.

[0005] The accuracy of contrast agent injection specifically includes five aspects. First, the injection time must be accurate. The contrast agent must be injected at a specific phase of the heart (such as the atrial contraction period) to ensure the best diagnostic results. Second, the injection dose must be accurate. The dose of contrast agent must be accurately calculated according to the specific situation of the patient (such as body weight, cardiac function), avoiding excess or deficiency. Third, the injection speed must be accurate. The injection speed must be moderate to ensure that the contrast agent can be fully mixed and evenly distributed. Fourth, the injection site must be accurate. The contrast agent should be injected into the appropriate blood vessel site (such as the right atrium), avoiding injection site deviation leading to inaccurate diagnostic results. Fifth, real-time monitoring and feedback must be accurate. The distribution of the contrast agent needs to be monitored in real time during the injection process and adjusted as needed.

[0006] The accuracy of contrast agent injection has a significant impact on the diagnostic results. First, accurate injection of contrast agent can ensure its uniform distribution in the heart chamber, generating high-quality, clear ultrasound images, which helps doctors accurately observe and identify PFO, reducing the possibility of misdiagnosis and missed diagnosis. Second, accurate injection of contrast agent can make the tiny PFO more clearly displayed through the development effect of the contrast agent, providing the best contrast effect and improving the diagnostic performance of ultrasound imaging.

[0007] In addition, precise injection of contrast agent can obtain high-quality diagnostic images at one time, reducing repeated examinations due to poor image quality, reducing the burden and discomfort of patients, and at the same time, by accurately controlling the dose and injection speed of contrast agent, it can reduce the side effects and potential safety risks caused by excessive use of contrast agent. Accurate injection avoids waste of contrast agent, improves resource utilization efficiency, saves medical costs, and through one-time accurate diagnostic results, avoids repeated operations, improves the work efficiency of the medical team and the smoothness of the diagnostic process.

[0008] Precise injection of contrast agent ensures the reliability of diagnostic results, providing a solid basis for doctors to develop appropriate treatment plans. Through accurate diagnostic results, subsequent interventional treatment or surgical operations can be better guided, improving treatment effectiveness and safety. Accurate injection of contrast agent helps to obtain standardized image data, improving data consistency and comparability, which plays an important role in subsequent patient management and research analysis. The accurate injection process reduces unnecessary repeated examinations and operations, reduces the pain and anxiety of patients, and improves patient satisfaction.

[0009] US2016213833(A1) discloses a method of determining an optimal volume of contrast agent to administer to a patient, the steps comprising: (a) measuring one or more patient-specific physiological parameters, the one or more patient-specific physiological parameters comprising at least a heart rate of the patient; and (b) determining the optimal volume of contrast agent based on the patient-specific physiological parameters determined under (a). This technical solution focuses on determining the optimal volume of contrast agent based on patient-specific physiological parameters such as heart rate. While the determination of volume is important, the ASCE's requirements for injection accuracy go beyond just volume, involving multiple aspects such as injection time, speed, and location. Simply determining the volume is not enough to ensure uniform distribution of contrast agent in the heart and optimal imaging results. ASCE requires real-time monitoring of contrast agent distribution during injection and adjustments as needed to ensure high-quality images and accurate diagnosis. However, the technical solution does not mention the mechanism of real-time monitoring and feedback, which is insufficient for the delicate operation of diagnosing patent foramen ovale (PFO) in ASCE.

[0010] US2014206991(A1) discloses a method of predicting a predicted contrast agent variation (CEP) for a contrast agent assisted examination of an examination object, as shown in FIG. 10, having the following steps: acquiring a patient-specific blood flow signal in a prediction measurement under the condition of using a defined examination protocol, and determining an individual impulse response function from the blood flow signal; acquiring a heart rate signal corresponding to the prediction measurement and obtaining heartbeat data; and predicting a contrast agent variation based on the individual impulse response function, the average CA heart rate, and the current examination protocol and the currently acquired heartbeat data. This technical solution is based on the blood flow signal, heart rate signal, and current examination protocol of a specific patient to predict the behavior of contrast agent. Among them, the CEP algorithm includes extracting the patient's function from the blood flow signal according to the AIR algorithm, then obtaining the corrected patient's function, and adjusting the corrected patient's function to the PRED algorithm based on the current examination protocol. Finally, the PRED algorithm is used to obtain the prediction of contrast agent variation. However, what ASCE needs is real-time monitoring and adjustment in actual operation to ensure that the contrast agent is accurately injected into the target area and to avoid artifacts and false signals. Relying solely on prediction may not be able to cope with complex situations and changes in actual operation. Precise control of injection time, speed, dose, and location is also needed during the ASCE diagnosis process to ensure that the contrast agent can fully enter the target area and be evenly distributed. This technical solution provides a prediction mechanism, but does not specifically address how to accurately control these parameters in actual operation.

[0011] Therefore, the prior art described above cannot be applied to the current ASCE. Therefore, for ASCE, the present application hopes to provide a method and device for building an artificial intelligence data set related to cardiac contrast agent injection, hoping to significantly improve the effect and efficiency of ASCE in diagnosing patent foramen ovale (PFO), thereby optimizing injection parameters, providing real-time monitoring and feedback, improving diagnostic accuracy and sensitivity, and reducing patient pain and discomfort. This data set also provides a powerful tool for doctors to make more accurate diagnoses and develop more effective treatment plans.

[0012] SUMMARY

[0013] Currently, transesophageal echocardiography combined with right heart acoustic contrast and sufficient provocation test is the gold standard imaging method for diagnosing patent foramen ovale, but there is no uniform standard for this operation procedure and diagnostic criteria in China. Medical staff instruct the patient to be tested to blow and hold their breath according to their work experience, and determine the duration of each action, which makes the duration of the Valsalva action of the patient to be tested not uniform. The timing of the medical staff injecting the contrast agent into the vein and the timing of taking the heart ultrasound image are all determined according to work experience, which causes some of the heart ultrasound images taken to be unable to accurately reflect the blood flow dynamics of the heart, thus forming a misdiagnosis of PFO.

[0014] The prior art has emerged technical solutions for detecting whether the patient's Valsalva respiration meets the standard by setting a Valsalva respiration detection mechanism and realizing the mixing of contrast agent by a contrast agent mixing injection mechanism to maintain the activity of the contrast agent. For example, CN114272470A discloses an oval foramen screening foaming injector, which comprises a Valsalva respiration detection mechanism for detecting whether the patient's Valsalva respiration meets the standard, a receiving module for receiving external injection instructions, a contrast agent mixing injection mechanism for mixing and injecting the contrast agent, and a control module for controlling the contrast agent mixing injection mechanism according to the Valsalva respiration detection mechanism and the receiving module. The control module controls the contrast agent mixing injection mechanism to mix the contrast agent, and controls the contrast agent mixing injection mechanism to inject after the Valsalva respiration detection mechanism detects that the patient's Valsalva respiration meets the standard and the receiving module detects that the external injection instructions are received. The blowing pressure sensor in the Valsalva respiration detection mechanism starts the corresponding timing device to record the duration of maintaining the detected expiration pressure when the detected expiration pressure reaches the preset pressure, and judges that the patient's Valsalva respiration action meets the standard when the duration reaches the preset time. However, the judgment of the patient's detection parameter index in this technical solution is limited to the duration of the state. According to this threshold setting method, the detection parameters recognized as qualified are the results after the continuous change of the patient's physical index for a period of time, and cannot reflect the continuous change process of the patient's physical state (such as cardiac blood flow dynamics). If only the data at the end of the physiological change characteristics is used to select the injection time of the contrast agent or the shooting time of the cardiac ultrasound image, the contrast agent cannot be matched with the most suitable cardiac blood flow dynamics, or the shooting time of the cardiac ultrasound image lags behind the actual shunt in the heart, thereby making the detection result inaccurate. That is, this technical solution only judges the execution standard of the patient's Valsalva respiration action from the form of the action, that is, whether the patient completes the specified action according to the existing evaluation standard, but cannot judge whether the same standard Valsalva respiration action can produce the same physiological change characteristics for different patients. For example, due to the differences in individual differences of patients, the same Valsalva respiration action usually produces different physiological change characteristics in different patients. The physiological change characteristics of the patient (such as cardiac blood flow dynamics) are directly related to the accurate selection of the injection time of the contrast agent. The appropriate injection time can ensure that the contrast agent reaches the highest concentration in a specific part of the heart, thereby ensuring that the contrast agent can realize more accurate image feedback under the driving of the corresponding physiological change, so as to obtain clear images and help doctors make accurate diagnoses.

[0015] The present application hopes to accurately judge and remind the heart shunt time and the application time of the contrast agent by providing a Valsalva action-based patent foramen ovale auxiliary diagnosis device, so as to generate accurate ultrasound images of the heart shunt and provide accurate diagnosis basis for patent foramen ovale.

[0016] In view of the deficiencies of the prior art, the present application provides an auxiliary diagnosis device for patent foramen ovale based on Valsalva action, which comprises a blowing unit, an electrocardio monitoring unit, a contrast agent injection unit and a control unit. The blowing unit is used to measure the blowing pressure of the patient to be tested; the electrocardio monitoring unit is used to measure the electrocardio change of the patient to be tested; the contrast agent injection unit is used to input the contrast agent into the patient to be tested; when the blowing pressure of the patient to be tested fed back by the blowing unit reaches the first predetermined threshold and remains for a predetermined time length, the control unit starts to request the electrocardio change rate of the patient to be tested output by the electrocardio monitoring unit. The present application uses the control unit and the pressure sensor to collect the data and time length of the blowing pressure, so as to provide accurate judgment opportunity for collecting the electrocardio change, and avoid the patient to be tested to repeatedly perform Valsalva action.

[0017] When it is determined that the electrocardio change rate of the patient to be tested appears the change inflection point related to Valsalva action, the control unit sends the prompt information of inputting the contrast agent to the contrast agent injection unit. The present application analyzes the electrocardio change rate and judges the opportunity of the appearance of the inflection point, so that the contrast agent injection unit can timely inject the contrast agent into the vein, avoid the injection opportunity of the contrast agent to be missed, and can make the injection opportunity of the contrast agent be unified standard.

[0018] In response to the received start information of the contrast agent vein input of the contrast agent injection unit, the control unit sends the blowing instruction to the patient to be tested through the prompt unit, and when it is determined that the blowing pressure of the patient to be tested measured by the blowing unit is lowered to the second predetermined threshold, the control unit sends the recording instruction of the blood flow force condition in the heart through the prompt unit.

[0019] Unlike the prior art, the auxiliary diagnostic device of the present application can measure the electrocardio change of the patient to be tested through the electrocardio monitoring unit to monitor the electrocardio change rate of the patient to be tested when the blowing pressure reaches the first predetermined threshold and remains for a predetermined time length, and when the electrocardio change rate of the patient to be tested appears a change inflection point related to the Valsalva action, the control unit sends a prompt information of inputting the contrast agent to the contrast agent injection unit, and the control unit can send a recording instruction of the blood flow dynamics in the heart through the prompt unit according to the start information of the intravenous input of the contrast agent. Based on the above technical features, the problems to be solved by the present application can include: how to improve the accuracy of the determination of the injection time of the contrast agent and how to improve the accuracy of the determination of the shooting time of the heart ultrasound image to generate and obtain an accurate ultrasound image of the shunt in the heart, and provide an accurate diagnosis basis for patent ductus arteriosus. The contrast agent injection unit of the present application is connected with the control unit, and the control unit prompts the injection of the contrast agent according to the monitoring results of the electrocardio change rate and the blowing pressure, so as to accurately control the injection time of the contrast agent according to the real-time data of the electrocardio change and the blowing pressure, and improve the accuracy of the diagnosis. In view of the defects of judging the shooting time according to the working experience and the defects that the blowing condition of the patient to be tested cannot be directly observed, the present application can determine the shooting time of the ultrasound image based on the change of the blowing condition, and timely capture the shunt condition in the heart to avoid missing the best shooting time.

[0020] Compared with the prior art which only detects the expiration pressure of the patient and ignores the correlation analysis between the actual Valsalva action and the electrocardio change rate, the present application can determine the blood flow dynamics in the heart suitable for the injection of the contrast agent through the correlation analysis between the electrocardio change rate and the Valsalva action in time, and the blood flow dynamics in the heart suitable for the injection of the contrast agent is characterized by the change inflection point related to the Valsalva action in time. Preferably, the determination of the change inflection point of the electrocardio change rate in time in association with the Valsalva action is performed as follows:

[0021] S01: Initial monitoring of the electrocardio change rate. At least from the start of blowing, the monitoring of the electrocardio change rate is started. When the patient to be tested starts to perform the Valsalva action, i.e. the blowing action, the electrocardio monitoring unit is immediately started to start real-time monitoring of the electrocardio change rate. The control unit receives the electrocardio signal through the electrocardio monitoring unit and performs preliminary analysis by using the artificial intelligence semantic training model.

[0022] S02: ECG change rate mutation monitoring after blowing pressure meets the standard. At the time point when the blowing pressure reaches the first predetermined threshold, the mutation amplitude of the ECG change rate is monitored. The blowing unit monitors that the blowing pressure reaches the first predetermined threshold (for example, 40 mmHg), and this information is immediately transmitted to the control unit. The control unit then switches to a more intensive monitoring mode and focuses on the mutation amplitude of the ECG change rate.

[0023] S03: Identify the turning point of the change in ECG change rate. When the mutation amplitude exceeds the change amplitude before the time point from the start of blowing to the time point when the blowing pressure reaches the first predetermined threshold, the time point of the mutation amplitude is determined as the turning point. The control unit analyzes the ECG change rate data during blowing and pressure meeting the standard to find the mutation amplitude. Once the monitored mutation amplitude exceeds the maximum change amplitude during the previous monitoring period, and the exceeding degree reaches or exceeds the preset percentage (for example, more than 30%, preferably more than 50%, and more preferably more than 70%), the time point of the mutation amplitude is determined as the turning point of the change in ECG change rate related to the Valsalva action.

[0024] S04: Preparation of contrast agent injection. After determining the turning point of the change in ECG change rate, the control unit immediately sends a prompt message to the contrast agent injection unit to prepare for injection. The medical staff prepares to inject the contrast agent into the vein of the patient to be tested according to the instructions of the control unit.

[0025] S05: Synchronize contrast agent injection with ECG change. The control unit sends an injection start signal to the contrast agent injection unit after confirming that the ECG change rate meets the preset conditions. While the medical staff performs the injection action, the control unit continues to monitor the ECG change rate to ensure that the injection timing is synchronized with the turning point of the change in ECG change rate.

[0026] S06: Record the dynamic of heart blood flow. After the injection is completed, the control unit sends an exhalation instruction to the patient to be tested through the prompt unit, and instructs the ultrasonic image imaging unit to start recording the dynamic of blood flow in the heart. The ultrasonic image imaging unit captures and records the change of blood flow in the heart, providing key information for diagnosis.

[0027] According to a preferred embodiment, when the blowing instruction is issued, the control unit forms a labeling prompt message through the prompt unit, and the artificial intelligence semantic training model in the control unit labels the ECG change rate of the patient to be tested sent by the ECG monitoring unit, thereby forming a training set of the ECG change rate. By collecting related data of ECG change and forming a training set, it is helpful for the artificial intelligence semantic training model to learn and identify the turning point related to the Valsalva action based on the training set.

[0028] According to a preferred embodiment, the control unit integrates an artificial intelligence semantic training model, which, after being trained, analyzes and determines the change inflection point related to the Valsalva action according to the heart rate change rate of the patient to be tested sent by the electrocardio monitoring unit. The present application analyzes the heart rate change rate and determines the change inflection point of the heart rate in time by training the artificial intelligence semantic training model, thereby achieving the good effect of accurately reminding the contrast agent input time.

[0029] According to a preferred embodiment, in the case that the control unit issues the blowing instruction through the prompting unit, the blowing unit monitors the blowing pressure of the pipeline and its change, and the control unit triggers the timing instruction that makes the timer connected thereto start timing based on the information condition that the blowing pressure reaches the first predetermined threshold; in the case that the time when the blowing pressure reaches the first predetermined threshold reaches a predetermined length of time, the control unit generates the instruction information that makes the patient to be tested hold breath and the timing prompt information of the holding breath length. The control unit of the present application can monitor the change condition and length of time of the blowing pressure through the timer and the pressure sensor, and can also remind the patient to be tested to cooperate to realize the blowing pressure standardization, i.e., to make the Valsalva action of the patient to be tested standardized, thereby reducing the case that the Valsalva action of the patient to be tested is not standardized and improving the efficiency of the diagnosis process of the patent for invention.

[0030] According to a preferred embodiment, in response to the end information of the contrast agent intravenous input of the contrast agent injection unit, the control unit issues the exhale instruction to the patient to be tested through the prompting unit, and determines whether the blowing pressure of the patient to be tested measured by the blowing unit is as low as the second predetermined threshold. Unlike the prior art, the control unit of the present application can issue the corresponding breathing instruction to the patient to be tested through the prompting unit according to the contrast agent intravenous input process. Based on the above-mentioned distinguishing technical features, the problem to be solved by the present application can include: how to reduce the difficulty of the patient to be tested to perform the Valsalva action without affecting the accuracy of the detection result. Specifically, without clear instructions, the patient to be tested is easy to maintain the holding breath state or inhale in advance due to the uncertainty of whether to breathe, which is easy to cause the patient to appear the dizziness, hypoxia and non-standard Valsalva action after performing the Valsalva action. The present application issues clear instructions by monitoring the blowing pressure, so that the patient to be tested is more likely to cooperate and reduce the difficulty of performing the Valsalva action standard.

[0031] According to one preferred embodiment, the blowing unit comprises a pressure sensor, a pressure communication port, a blowing assembly and a pipeline, the pressure sensor is arranged in the pipeline connected with the blowing assembly to collect blowing pressure; the pressure sensor sends data information of blowing pressure to the control unit through the pressure communication port. The present application monitors blowing pressure and its change condition through the control unit and the pressure sensor, which is more conducive to timely issuing instructions of implementation time of blowing, inhaling, holding breath and other actions, realizes standardization of Valsalva action and improves effectiveness of shooting of the heart's ultrasound image.

[0032] According to one preferred embodiment, the electrocardio change condition measured by the electrocardio monitoring unit comprises change of heart rate, P wave and QRS complex; the electrocardio monitoring unit compares electrocardiogram before and after the patient to be tested implements Valsalva action to measure the electrocardio change condition. The present application can more timely judge heart rate change and timely judge inflection point of heart rate by comparing electrocardiogram before and after Valsalva action.

[0033] According to one preferred embodiment, the device further comprises a display unit, based on change of blowing pressure, the control unit controls the display unit to display change of blowing pressure in different pressure ranges in a marked change manner to remind the patient to be tested or medical staff of current blowing condition. This is conducive to the patient to be tested and medical staff to know current blowing state according to the marked change, so as to actively cooperate to complete the diagnosis process.

[0034] According to one preferred embodiment, the control unit synchronously displays blowing pressure and its change, timing information and electrocardio change condition through the display unit. The present application integrates three types of information in the same display unit, so that medical staff can know changes of various types of information at the same time, and timely implement various operations according to received prompt instructions, realizing standardization of diagnosis operation of patent foramen ovale.

[0035] According to one preferred embodiment, when heart rate of the patient to be tested slows down and reaches a preset heart rate threshold, the control unit issues prompt information of inputting contrast agent to the contrast agent injection unit. When the preset heart rate threshold is reached, the inflection point of heart rate change is reached, and accurate judgment and reminding of this time can avoid medical staff missing the best moment of shooting of ultrasound image, so as to obtain blood flow change condition of the heart at the best moment, avoid missing the best shooting moment and provide the most accurate diagnosis basis.

[0036] According to a preferred embodiment, during the preparation of the contrast agent, the contrast agent is mixed with air in a repeated plunger manner to form a uniform mixture of air and contrast agent. The advantages of doing so are: it can improve the image quality of the shot, making the outline of the structure in the body clearer; it reduces the need for contrast agent, reducing the side effects and allergic reactions that patients may experience; the air-mixed contrast agent is more evenly distributed and may also reduce the viscosity, so that the mixture is easier to inject and transport.

[0037] The present application provides a processor for constructing an artificial intelligence data set related to cardiac contrast agent injection from a third aspect. The processor is provided with an artificial intelligence semantic training model. The processor first sends a blowing instruction to the patient to be tested through the prompt unit connected thereto (which can be a speech synthesizer or a display screen). This instruction can be a voice prompt telling the patient to perform a blowing test, or a text prompt displayed on the screen for the patient to read. When the patient blows according to the instruction, the blowing unit connected to the processor measures and sends the pressure data generated by the patient's blowing. At the same time, the electrocardiogram monitoring unit also monitors and sends the patient's electrocardiogram changes. In this step, the prompt unit generates labeled prompt information according to the instructions sent by the processor. This includes the pressure reading of blowing, the graph or numerical value of electrocardiogram changes, and any abnormal situation warning. The artificial intelligence semantic training model analyzes the received electrocardiogram change data. The artificial intelligence semantic training model learns to judge the inflection point of the electrocardiogram change rate according to the labeled information of the electrocardiogram change rate. The labeled information of the electrocardiogram change rate contains a large amount of electrocardiogram change data and corresponding labels, which are used to train the artificial intelligence semantic training model to identify and predict heart rate changes. By analyzing the electrocardiogram change rate, the artificial intelligence semantic training model can timely judge the inflection point of the heart rate. When the artificial intelligence semantic training model identifies the inflection point of the heart rate, it can remind medical personnel to input the contrast agent at the best time.

[0038] Through the connection of the processor with the blowing unit and the electrocardiogram monitoring unit, real-time data collection and labeling are realized, ensuring the high accuracy and integrity of the data, thereby improving the quality of the data set. Combining blowing pressure and electrocardiogram changes, comprehensive data from multiple dimensions helps to construct a more comprehensive and complex data set, providing a basis for training a more powerful AI model.

[0039] According to a preferred embodiment, during the insufflation process, the processor monitors the magnitude of the sudden change in the heart rate variability at the time point when the insufflation pressure reaches the first predetermined threshold. To determine the magnitude of the sudden change, the processor first starts collecting the electrocardiogram data and the insufflation pressure data from the time point when the insufflation starts, usually at a frequency of 100 times per second to ensure the accuracy and delicacy of the data. Next, the processor calculates the initial change magnitude of the heart rate variability during the time period from the start of the insufflation to the insufflation pressure reaching the first predetermined threshold. This process uses statistical methods such as mean and standard deviation to calculate the heart rate variability amplitude during this time period, which is defined as the initial change magnitude.

[0040] After the insufflation pressure reaches the first predetermined threshold, the processor monitors the changes in the heart rate variability in real time and compares the current change magnitude of the heart rate variability with the initial change magnitude. If the current sudden change magnitude of the heart rate variability exceeds the predetermined proportion (e.g., 1.5 times or 2 times) of the initial change magnitude, the processor confirms that a significant sudden change has occurred. At this time, the processor uses a sudden change detection algorithm (such as CUSUM or Bayesian sudden change detection) to reduce the impact of short-term fluctuations and ensure the accuracy of the judgment. Once a significant sudden change is identified, the processor records the exact time point of the event and marks it as the change inflection point of the heart rate variability, which is stored in the memory for subsequent analysis and verification.

[0041] During the insufflation process, the change magnitude of the insufflation pressure is also a key monitoring indicator. When the insufflation starts, the patient performs the insufflation operation according to the instructions, and the insufflation unit monitors and records the changes in the insufflation pressure in real time through high-precision sensors. Throughout the process, the processor collects the insufflation pressure data at a high frequency (such as 100 times per second) and transmits these data to the processor for analysis through wireless or wired connection. The processor receives and analyzes the insufflation pressure data in real time, while using smoothing algorithms (such as moving average or Kalman filtering) to reduce data noise and ensure the accuracy of the analysis results.

[0042] During the time period from the start of the insufflation to the insufflation pressure reaching the first predetermined threshold, the processor calculates and records the initial change magnitude of the insufflation pressure. When the insufflation pressure reaches the predetermined threshold, the processor records the time point and switches to the heart rate variability monitoring mode, continuing to monitor the insufflation pressure and electrocardiogram data at a high frequency. After that, the processor continuously monitors and records the changes in the insufflation pressure for subsequent analysis and verification.

[0043] When the processor determines the inflection point of the change in the electrocardio rate, i.e., the time point of the mutation amplitude, the processor generates a labeling prompt information. This information includes the mutation time point, electrocardio rate data, mutation amplitude, and blowing pressure data, etc. The processor transmits these information to the prompt unit, which timely conveys to the medical staff through screen display, voice broadcast, vibration reminder, etc. At the same time, the processor records and stores all generated labeling prompt information in the internal or external memory, ensuring the integrity and traceability of the data, and backing up to the cloud storage or external hard disk regularly to prevent data loss. This series of detailed steps ensures the accuracy and efficiency of the processor in monitoring the blowing pressure and electrocardio rate, as well as the timeliness and reliability of the labeling prompt information, providing strong support for the medical process.

[0044] According to a preferred embodiment, during the blowing process, when the blowing pressure reaches a first predetermined threshold, the processor triggers the corresponding timer to start timing. First, the processor monitors the blowing pressure data collected by the blowing unit in real time, and the collection frequency is usually set to multiple times per second (such as 100 times per second) to ensure the continuity and accuracy of the data. Then, a first predetermined threshold (for example, 10 cmH2O) is set in advance and displayed and confirmed on the control panel or software interface of the processor. When the processor monitors that the real-time blowing pressure reaches or exceeds this predetermined threshold, a timing instruction is generated to make the timer connected to it start timing and record the exact time point when the blowing pressure reaches the first predetermined threshold.

[0045] In the next step, the processor generates instruction information to guide the patient to be tested to perform the breath-holding operation. First, according to the clinical requirements or experimental design, a predetermined duration (for example, 30 seconds) is set and displayed and confirmed on the control panel or software interface of the processor. The processor monitors the timing progress of the blowing pressure reaching the first predetermined threshold through the timer in real time, and compares the time of the timer with the predetermined duration. When the time recorded by the timer reaches the predetermined duration, the processor generates instruction information to require the patient to start breath-holding and specifies the specific duration of breath-holding. In addition, the processor also generates a breath-holding duration prompt information containing the specific time (for example, 10 seconds) and conveys it to the patient or medical staff through the prompt unit, etc.

[0046] To ensure accurate time control of data collection, the processor employs a high-precision timer to reduce time errors and record the time points of all key events in real time, including the time when the blowing pressure reaches the predetermined threshold, the start of the timing, and the arrival of the predetermined duration. All collected data, including time points and corresponding blowing pressure, heart rate changes, and other information, are recorded and stored in internal or external memory for subsequent time series analysis. Through the timer and the setting of the predetermined duration, accurate time control of data collection is ensured, providing a reliable basis for time series analysis in the data set. The processor adjusts the data collection process flexibly according to the dynamic changes of blowing pressure and predetermined duration, improving the diversity and coverage of data.

[0047] According to a preferred embodiment, the processor is connected to the memory in a wired or wireless manner. For example, the processor is connected to the memory through a high-speed bus interface such as SPI, I2C, or PCIe, ensuring the speed and stability of data transmission. The memory is used to store various data information collected and generated by the processor. These data information can include various received data such as heart rate changes and their change inflection points, breath-holding duration, blowing-related time and pressure, and contrast agent injection unit injection time, as well as various instruction data. The processor formats the collected data information into a structured data format suitable for storage such as JSON, XML, or CSV. The processor generates a timestamp for each record, establishes a time index, and facilitates subsequent data retrieval in chronological order. The processor marks the change inflection points of the heart rate, the breath-holding duration, the key time points of the blowing pressure, and the contrast agent injection time, and establishes an event index. The processor integrates the contrast agent injection time information into the overall data record and stores it in association with the heart rate, blowing pressure, and other information. The processor marks the injection time information for subsequent data analysis and retrieval, ensuring data integrity and traceability, and providing sufficient raw data for subsequent data analysis and model training. BRIEF DESCRIPTION OF DRAWINGS

[0048] FIG. 1 is a schematic diagram of the hardware connection relationship of the patent foramen ovale auxiliary diagnosis device provided by the present application;

[0049] FIG. 2 is a schematic diagram of the logic connection relationship of the patent foramen ovale auxiliary diagnosis device provided by the present application;

[0050] FIG. 3 is a schematic diagram of the display unit provided by the present application;

[0051] FIG. 4 is a schematic diagram of the steps of the method for administering contrast agent provided by the present application;

[0052] FIG. 5 is a logic diagram of the patent foramen ovale diagnosis method based on Valsalva action;

[0053] Fig. 6 is a schematic diagram of one of the PFO diagnostic images of patient A;

[0054] Fig. 7 is a schematic diagram of another PFO diagnostic image of patient A;

[0055] Fig. 8 is a schematic diagram of one of the PFO diagnostic images of patient B;

[0056] Fig. 9 is a schematic diagram of the processor connected to construct an artificial intelligence data set related to the injection of a cardiac contrast agent according to the present application.

[0057] Fig. 10 is a schematic diagram of another PFO auxiliary diagnostic device in the prior art.

[0058] List of reference signs 100: air blowing unit; 110: pressure sensor; 120: pressure communication port; 130: air blowing assembly; 140: pipeline; 200: electrocardiogram monitoring unit; 210: electrocardiogram communication port; 220: electrode; 230: induction module; 300: contrast agent injection unit; 310: contrast agent; 320: injector; 330: execution assembly; 340: injection cavity; 350: injection communication port; 400: ultrasonic image imaging unit; 410: ultrasonic probe; 420: image communication port; 500: control unit; 510: prompting unit; 520: control communication port; 530: processor; 540: artificial intelligence semantic training model; 550: memory; 560: timer; 600: patient to be tested; 700: display unit. DETAILED DESCRIPTION

[0059] The following will be described in detail in conjunction with the accompanying drawings.

[0060] Valsalva test is a clinical physiological test that makes patients perform strong closed mouth action, i.e. after deep inhalation, tightly close the glottis, and then perform exhalation action, resist the tightly closed epiglottis during exhalation, increase the intrathoracic pressure to affect blood circulation and autonomic nervous function state, and then achieve the purpose of diagnosis and treatment.

[0061] In the present application, a physiological saline and air mixture is selected as the contrast agent 310.

[0062] The physiological saline and air mixture can be used as a contrast agent 310 for echocardiography, especially in evaluating right ventricular function and detecting patent foramen ovale (PFO). This contrast agent 310 is commonly referred to as "saline contrast" or "acoustic contrast", which is used to enhance ultrasound images by rapidly injecting physiological saline containing microbubbles.

[0063] In the present application, as shown in Fig. 4, the "saline contrast" or "acoustic contrast" administration method includes the following steps:

[0064] S11: Preparation step: Take a certain amount of normal saline, for example, take 9 milliliters of normal saline. The specific volume of normal saline is obtained as needed, which is not limited here.

[0065] S12: Mixing step: 9 milliliters of normal saline is mixed with 1 milliliter of air to form a solution containing microbubbles. This can be achieved by repeatedly pushing and pulling the three-way connection of two 10 milliliter syringes 20-30 times to mix the normal saline and air thoroughly.

[0066] Preferably, during the preparation of the contrast agent 310, the contrast agent 310 is uniformly mixed with air by repeatedly pulling the syringe 320 to form a uniform mixture of air and contrast agent 310.

[0067] S13: Intravenous injection step: The microbubble mixture is quickly injected into the vein, usually into the elbow vein.

[0068] S14: Monitoring step: Immediately after injection, use the ultrasound device to observe and evaluate the contrast effect and the structure of the heart.

[0069] Example 1

[0070] Currently, transesophageal echocardiography (TEE) combined with right heart acoustic contrast (ASCE) and sufficient stimulation test is the gold standard imaging method for diagnosing patent foramen ovale, but there is no uniform standard for the operation process and diagnostic criteria in China. Medical staff instruct the patient 600 to blow, hold breath and determine the duration of each action according to work experience, which makes the duration of the Valsalva action of the patient 600 not uniform, and the timing of the medical staff injecting the contrast agent 310 into the vein and the timing of shooting the heart ultrasound image are all determined according to work experience, which causes some of the heart ultrasound images cannot accurately reflect the blood flow dynamics of the heart, thus forming misdiagnosis for PFO diagnosis.

[0071] The present application hopes to accurately judge and remind the timing of heart shunt and the timing of application of contrast agent 310 by providing a patent foramen ovale auxiliary diagnosis device based on Valsalva action, so as to generate accurate ultrasound images of intracardiac shunt and provide accurate diagnostic basis for patent foramen ovale.

[0072] In view of the deficiencies of the prior art, the present application provides a patent foramen ovale auxiliary diagnosis device based on Valsalva action, as shown in FIG. 1, which comprises a blowing unit 100, an electrocardiogram monitoring unit 200 and a contrast agent injection unit 300. As shown in FIG. 2, the blowing unit 100, the electrocardiogram monitoring unit 200 and the contrast agent injection unit 300 are respectively connected in communication with the control unit 500.

[0073] As shown in FIG. 1 and FIG. 2, the blowing unit 100 is used to measure the blowing pressure of the patient 600 to be measured. Preferably, the blowing unit 100 comprises a pressure sensor 110, a pressure communication port 120, a blowing assembly 130 and a pipe 140, the pressure sensor 110 is arranged in the pipe 140 connected with the blowing assembly 130 to collect the blowing pressure. The pressure sensor 110 sends the data information of the blowing pressure to the control unit 500 through the pressure communication port 120. The present application monitors the blowing pressure and its change condition through the control unit 500 and the pressure sensor 110, which is more conducive to timely issuing the correct instruction of the implementation time of blowing, inhaling, holding breath and other actions, realizing the standardization of Valsalva action and improving the effectiveness of the shooting of the ultrasound image of the heart.

[0074] As shown in FIG. 1 and FIG. 2, the ECG monitoring unit 200 is used to measure the ECG change condition of the patient 600 to be measured. The ECG monitoring unit 200 comprises an ECG communication port 210, a sensing module 230, a lead wire and an electrode 220. The sensing module 230 is located in the monitoring and display assembly box and is connected with the electrode 220 through the lead wire. The sensing module 230 is mainly responsible for sensing the ECG signal of the patient 600 to be measured. It is usually composed of one or more electrodes 220 through the lead wire, and these electrodes 220 are in contact with the body of the patient 600 to be measured to detect the change of the ECG signal. The electrode 220 can be pasted on the radial artery of the patient 600 to be measured through the adhesive tape. The electromyographic signal collected through the electrode 220 constructs the electrocardiogram, and the ECG waveform is displayed on the display unit 700. The sensing module 230 is in communication connection with the control unit 500 through the ECG communication port 210.

[0075] For example, when the heart contracts, the electromyographic signal generated is transmitted to the sensing module 230 through the electrode 220 and the lead wire. The sensing module 230 receives and amplifies these electrical signals and converts them into digital signals. Then, the digital signals are transmitted to the ECG communication port 210 and transmitted to the display unit 700 through the connected control unit 500, so as to be monitored and observed by medical staff.

[0076] As shown in FIG. 1 and FIG. 2, the contrast agent injection unit 300 is used to input the contrast agent 310 into the patient 600 to be tested. The contrast agent injection unit 300 comprises the contrast agent 310 and the injector 320. Preferably, the injector 320 is two 10ml screw injectors 320 connected in a three-way manner so as to manually inject the contrast agent 310 by the medical staff. In the case that the injector 320 can be controlled by the control unit 500, the injector 320 further comprises an execution assembly 330 capable of performing push and pull. The contrast agent 310 is filled in the injection cavity 340 of the injector 320. The execution assembly 330 is communicatively connected with the control unit 500 through the injection communication port 350. The execution assembly 330 is, for example, a displacement sensor arranged on the push rod of the injector 320. The displacement sensor sends displacement data to the control unit 500 through the injection communication port 350, so that the control unit 500 can determine the injection action and the injection stopping moment of the injector 320 according to the displacement data.

[0077] As shown in FIG. 1 and FIG. 2, the ultrasound image imaging unit 400 is used to image the blood flow force condition in the heart of the patient 600 to be tested to monitor the right-to-left intracardiac shunt condition at the atrial level. The ultrasound image imaging unit 400 comprises an ultrasound image shooting body, an ultrasound probe 410 and an image communication port 420. The ultrasound image shooting body feeds back information to the control unit 500 through the image communication port 420 whether the shooting action is completed.

[0078] As shown in FIG. 2, in the case that the blowing pressure of the patient 600 to be tested fed back by the blowing unit 100 reaches the first predetermined threshold and remains for a predetermined time length, the control unit 500 starts to request the heart rate variation of the patient 600 to be tested output by the electrocardiogram monitoring unit 200. The present application uses the control unit 500 and the pressure sensor 110 to collect the data and time length of the blowing pressure, thereby providing accurate judgment opportunity for collecting the electrocardiogram variation condition and avoiding the patient 600 to be tested repeatedly performing the Valsalva action.

[0079] For example, in order to ensure the accuracy and effectiveness of the examination process, the patient 600 to be tested needs to receive detailed guidance before the cardiac angiography examination, in order to correctly perform the Valsalva action. Specifically, the patient 600 to be tested needs to be guided by a medical professional to perform this specific operation: first take a deep breath, then try to exhale with force, but at the same time keep the airway closed, even if the gas cannot actually be expelled outside the body. This forced exhalation action similar to the effort to defecate or hold breath, so that the gas has the pressure and flow rate of movement, so that the gas generates pressure in the pipeline 140, when the gas pressure in the pipeline 140 reaches the pressure of 40 mmHg, the maintenance time of the gas with pressure above 40 mmHg is started to be calculated. Generally, the patient 600 to be tested needs to make the pressure in the pipeline 140 maintain above 40 mmHg for 4 to 5 seconds. When the pressure in the pipeline 140 maintains above 40 mmHg for 4 to 5 seconds, the control unit 500 starts to request the heart rate change of the patient 600 to be tested output by the electrocardio monitoring unit 200. Through this action, the cardiovascular system of the patient 600 to be tested will have a series of physiological reactions, which are very important for the diagnosis of patent foramen ovale (PFO).

[0080] Preferably, the control unit 500 is a processor, a dedicated integrated chip or a server capable of performing the information processing steps of the present application. The control unit 500 includes a prompting unit 510, a control communication port 520, a processor 530, a memory 550 and a timer 560. The prompting unit 510, the control communication port 520, the memory 550 and the timer 560 are respectively connected with the processor 530 in a wired or wireless manner and realize the transmission of information. The control communication port 520 is used for communication and belongs to the communication port. The prompting unit 510 includes a sound prompting component and a light component. The prompting unit 510 can receive the instruction information issued by the processor 530 and issue sound or light prompting information corresponding to the instruction information. The memory 550 is used to store the data information of the processor 530. The timer 560 can be a timing program in the processor 530, or a specific clock. The processor 530 is used to specifically perform the tasks of data analysis, instruction generation, etc. in the control unit 500, and is the core component of the control unit 500.

[0081] In advanced medical imaging techniques, ensuring the precise injection of contrast agent 310 is a crucial step that not only affects the quality of the imaging but also directly relates to the safety and comfort of the patient 600 under examination. In this context, the design of the control unit 500 takes into full consideration the real-time monitoring and analysis of electrocardiogram (ECG) data and how to optimize the timing of the contrast agent 310 injection based on these data. Valsalva maneuver, as a simple physiological maneuver, can cause significant changes in the rate of electrocardiographic changes, providing a unique window for monitoring and assessing cardiac function. When the patient 600 under examination performs the Valsalva maneuver, the heart rate usually experiences a rapid rise followed by a slow decline, during which one or more inflection points of slowing heart rate may occur. These change inflection points reflect the heart's response to physiological stress and are ideal indicators for precise contrast agent 310 injection.

[0082] When the electrocardiographic rate of change of the patient 600 under examination is determined to have a change inflection point related to the Valsalva maneuver, the control unit 500 sends a prompt to the contrast agent injection unit 300 to input the contrast agent 310. Preferably, the change inflection point refers to the inflection point of slowing heart rate.

[0083] For example, upon receiving the electrocardiogram data of the patient 600 under examination, the control unit 500 is responsible for real-time analysis of the electrocardiogram data of the patient 600 under examination. When the patient 600 under examination performs the Valsalva maneuver, the control unit 500 analyzes the electrocardiographic rate of change in real time and identifies the inflection point of slowing heart rate through an algorithm. This judgment is based on a deep analysis of electrocardiogram data, including heart rate change trend, heart rate change amplitude, and heart rate change rate, etc. Once the inflection point of slowing heart rate is detected, the control unit 500 immediately sends a prompt to the contrast agent injection unit 300. This prompt contains all the necessary instructions and parameters to ensure that the contrast agent 310 is injected precisely at the optimal time. The contrast agent injection unit 300 injects immediately upon receiving the bolus prompt and cannot be stationary or stationary. The automation and intelligence of this process significantly improve the efficiency and safety of the use of contrast agent 310, while also reducing unnecessary repeated examinations due to poor injection timing. The present invention can standardize the injection timing of the contrast agent 310.

[0084] The electrocardiographic changes measured by the electrocardiographic monitoring unit 200 include changes in heart rate, P wave, and QRS complex. The electrocardiographic monitoring unit 200 compares the electrocardiogram before and after the patient 600 under examination performs the Valsalva maneuver to measure the electrocardiographic changes. The present invention captures the physiological response of the heart caused by the Valsalva maneuver by comparing the electrocardiogram before and after the Valsalva maneuver, providing important information for subsequent diagnosis and treatment.

[0085] According to a preferred embodiment, the control unit 500 integrates an artificial intelligence semantic training model 540, which, once trained, analyzes and determines the change inflection point related to the Valsalva maneuver from the heart rate variability of the patient 600 under test sent by the electrocardiographic monitoring unit 200.

[0086] Preferably, the artificial intelligence semantic training model 540 can be obtained by training a deep neural network (DNN). A deep neural network is a network structure composed of multiple layers of artificial neurons that simulate the way neurons work in the human brain, capable of learning and recognizing complex data patterns.

[0087] During the training process of the artificial intelligence semantic training model 540, a key step is the use of a large amount of electrocardiogram (ECG) data as training samples. These samples contain various patterns of normal and abnormal electrocardiograms, especially data related to the heart rate variability and change inflection point associated with the Valsalva maneuver. Through these data, the artificial intelligence semantic training model 540 learns how to identify the trend of heart rate changes, especially to identify the change inflection point, which is crucial for subsequent medical decisions.

[0088] The training process of the artificial intelligence semantic training model 540 adopts a supervised learning method, where each training sample has a corresponding label, i.e., the correct interpretation of the electrocardiogram data represented by the sample (e.g., whether a change inflection point associated with the Valsalva maneuver has occurred). The artificial intelligence semantic training model 540 gradually improves its accuracy and reliability in analyzing electrocardiogram data by continuously adjusting the weights and biases in the network to minimize the difference between the predicted results and the true labels.

[0089] In addition, the artificial intelligence semantic training model 540 can also employ specific techniques to optimize the training process, such as convolutional neural networks (CNN) for extracting local features in electrocardiogram data, or recurrent neural networks (RNN) to handle the temporal dependencies in sequential data. The application of these techniques further enhances the ability of the artificial intelligence semantic training model 540 in handling complex electrocardiogram data, enabling it to more accurately identify and analyze the heart rate variability and change inflection point associated with the Valsalva maneuver.

[0090] Although the intelligent network model such as the neural network model and the reinforcement deep learning model can form the artificial intelligence semantic training model 540 for identifying the change inflection point related to the Valsalva action, the disadvantage of such formation is that the change inflection point related to the Valsalva action learned and identified by the intelligent network model may be misjudged. Especially when the electrocardio change rate of the patient to be tested is different from that of ordinary patients, for a special electrocardio change rate, the artificial intelligence semantic training model 540 is prone to automatically generate inaccurate identification results, which misleads the doctor to misdiagnosis. Therefore, based on this defect, the artificial intelligence semantic training model 540 retrieves the training set data and identifies the change inflection point related to the Valsalva action of the electrocardio change rate based on the training set, and the result will be more accurate. When the physiological characteristics of the patient are special, or the artificial intelligence semantic training model 540 cannot find approximate data from the training set, the artificial intelligence semantic training model 540 can issue a prompt information of human intervention through the prompt unit 510, so that the medical staff can make a judgment on the change inflection point related to the Valsalva action. In this way, the judgment result automatically generated by the artificial intelligence semantic training model 540 can be fully avoided, the occurrence of false results is avoided, and the misleading of medical staff is also avoided.

[0091] Specifically, when the blowing instruction is issued, the control unit 500 forms a labeling prompt information through the prompt unit 510, and the artificial intelligence semantic training model 540 in the control unit 500 labels the electrocardio change rate of the patient 600 to be tested sent by the electrocardio monitoring unit 200, and then forms a training set of the electrocardio change rate. When the patient performs the Valsalva action, the electrocardio data will show a specific change pattern, especially the change of heart rate. During the Valsalva action, the heart rate usually increases first, and then rapidly decreases after the action ends. These changes can be observed through the R-R interval of the electrocardio signal, that is, the time interval between two consecutive heartbeats. The inflection point data usually refers to the starting and ending points of the heart rate change.

[0092] Table 1: Exemplary data of training set table.

[0093] In Table 1, the "@" symbol is used to connect the time point and the corresponding heart rate value, indicating that these data are measured at the same time point. This representation method helps to clearly show the heart rate data at different time points, making the data reading more intuitive. For example, "70bpm@09:05" means that at 09:05, the heart rate of the patient is 70 times per minute. This format is commonly used in medical records and data analysis to ensure that the correlation between time and related measurement values is clear at a glance.

[0094] In Table 1, the initial heart rate and time represent the heart rate and time recorded before the Valsalva action.

[0095] Heart rate increase inflection point represents the inflection point and time of the increase of heart rate during the Valsalva action.

[0096] Heart rate decrease inflection point represents the inflection point and time of the decrease of heart rate after the end of the Valsalva action.

[0097] Final heart rate and time represent the recorded heart rate and time after the end of the Valsalva action.

[0098] Heart rate change rate (ΔHR / ΔT) represents the change rate of heart rate during the Valsalva action, expressed in the change of heart rate per minute.

[0099] Annotation information represents the description of the heart rate change feature, which is used to assist the artificial intelligence semantic training model to understand the data.

[0100] The present application trains the artificial intelligence semantic training model 540 to call the training set to analyze the electrocardiogram change rate and timely judge the change inflection point of heart rate, so as to realize the good effect of accurately reminding the contrast agent input time.

[0101] The present application further optimizes the analysis of electrocardiogram change rate by using artificial intelligence algorithm. Through the learning of a large amount of sample data, the artificial intelligence semantic training model 540 can more accurately identify the electrocardiogram change feature related to patent foramen ovale, thereby improving the accuracy and efficiency of diagnosis.

[0102] In response to the received start information of the contrast agent intravenous input of the contrast agent injection unit 300, the control unit 500 issues a blowing instruction to the to-be-tested patient 600 through the prompting unit 510, and in the case where the blowing pressure of the to-be-tested patient 600 measured by the blowing unit 100 is determined to be lowered to a second predetermined threshold, the control unit 500 issues a recording instruction of the blood flow dynamics in the heart through the prompting unit 510, so that the ultrasonic image imaging unit 400 starts to shoot the image of the blood flow dynamics in the heart.

[0103] In view of the defects of judging the shooting opportunity according to work experience and the defect that the blowing condition of the to-be-tested patient 600 cannot be directly observed, the present application can determine the shooting opportunity of the ultrasonic image based on the change of the blowing condition, so as to timely capture the shunt condition in the heart and avoid missing the best shooting opportunity in the shooting action.

[0104] According to a preferred embodiment, in the case that the control unit 500 issues the blowing instruction through the prompting unit 510, the blowing unit 100 monitors the blowing pressure of the pipeline 140 and its change, and the control unit 500 triggers the timing instruction that causes the timer 560 connected thereto to start timing based on the information condition that the blowing pressure reaches the first predetermined threshold. In the case that the time when the blowing pressure reaches the first predetermined threshold reaches a predetermined length of time, the control unit 500 generates the instruction information that causes the patient 600 to be tested to hold breath and the timing prompting information of the length of time of holding breath. The control unit 500 of the present application can monitor the change condition and length of time of the blowing pressure through the timer 560 and the pressure sensor 110, and can also remind the patient 600 to be tested to cooperate to realize the blowing pressure standardization, i.e., to standardize the Valsalva action of the patient 600 to be tested, thereby reducing the case that the Valsalva action of the patient 600 to be tested is not standardized and improving the efficiency of the diagnosis process of the patent foramen ovale.

[0105] According to a preferred embodiment, in response to the end information of the contrast agent intravenous input of the contrast agent injection unit 300, the control unit 500 issues the exhalation instruction to the patient 600 to be tested through the prompting unit 510, and determines whether the blowing pressure of the patient 600 to be tested measured by the blowing unit 100 is lowered to the second predetermined threshold. In the case that there is no explicit instruction, the patient 600 to be tested is easy to maintain the holding breath state or inhale in advance due to the uncertainty of whether breathing is allowed, which is easy to cause the patient 600 to be tested to appear the dizziness, hypoxia and non-standard Valsalva action after the Valsalva action is performed. The present application issues the explicit instruction by monitoring the condition of the blowing pressure, so that the patient 600 to be tested is more easy to cooperate and the difficulty of performing the standard Valsalva action is reduced.

[0106] For example, in the process that the patient 600 to be tested performs the Valsalva action, the patient 600 to be tested blows through the hose, and the blowing pressure needs to reach 40 mmHg (the first predetermined threshold). After reaching, the patient 600 to be tested needs to hold breath for 10 seconds (by the preset length of time), and the effective Valsalva action holding breath heart rate appears the process of descending→ascending→maintaining→descending, and the heart rate change of the patient 600 to be tested is compared by recording the electrocardiogram. The medical staff pushes the contrast agent 310 in the heart rate maintaining period of the patient 600 to be tested (about 7 seconds of holding breath), the patient 600 to be tested exhales (resumes breathing) after the contrast agent 310 is pushed, and the doctor observes the heart shunt condition of the patient 600 to be tested through the ultrasonic imaging unit 400 after exhaling, and diagnoses whether there is the patent foramen ovale and the level of shunt according to the shunt condition.

[0107] The application improves the diagnosis accuracy and operation safety of patent foramen ovale by elaborating the selection and administration of the cardiac contrast agent 310 and the method of accurately setting the inflation pressure and the heart rate change rate, and has important application value for clinical diagnosis.

[0108] According to a preferred embodiment, the device further comprises a display unit 700. Based on the change of the inflation pressure, the control unit 500 controls the display unit 700 to display the change of the inflation pressure in different pressure ranges in a marked manner, so as to remind the patient 600 to be tested or the medical staff of the current inflation situation. In this way, the patient 600 to be tested and the medical staff can know the current inflation state according to the marked change, so as to actively cooperate to complete the diagnosis process.

[0109] According to a preferred embodiment, the control unit 500 synchronously displays the inflation pressure and its change, the timing information and the electrocardiogram change through the display unit 700. The application integrates the three types of information in the same display unit 700, so that the medical staff can know the changes of various types of information at the same time, and timely implement various operations according to the received prompt instructions, so as to realize the standardization of the diagnosis operation of patent foramen ovale.

[0110] For example, as shown in FIG. 3, the display information of the display unit 700 is divided into three rows, the uppermost row is a red light, a yellow light and a green light. The middle row displays the pressure value and the timing time, and the lowermost row displays the electrocardiogram.

[0111] The pressure sensor 110 monitors the inflation pressure when the patient 600 to be tested inflates, and displays the pressure value on the display unit 700. When the pressure is greater than 40 mmHg, the green light is on, when the pressure is 36-40 mmHg, the yellow light is on, and when the pressure is less than 35 mmHg, the red light is on. When there is no pressure, the light is off. When the pressure is greater than 40 mmHg, the timer 560 is triggered to count for 10 seconds, and the value is displayed on the display unit 700.

[0112] According to a preferred embodiment, when the heart rate of the patient 600 to be tested slows down and reaches a preset heart rate threshold, the control unit 500 sends a prompt information of inputting the contrast agent 310 to the contrast agent injection unit 300. When the preset heart rate threshold is reached, the inflection point of the heart rate change is reached. The accurate judgment and prompt of this opportunity can avoid the medical staff missing the best moment of ultrasonic image shooting, so as to obtain the blood flow change of the heart at the best moment, avoid missing the best shooting opportunity, and provide the most accurate diagnosis basis.

[0113] Two heart ultrasound images of patient A are shown in Fig. 6 and Fig. 7. Patient A has been injected with contrast agent 310. Patient A is a 10-year-old child, who cooperated with the Valsalva maneuver as instructed, "rest 3 levels, 3 levels after Valsalva maneuver", breath-holding time 13 seconds, and heart rate changed. The contrast agent 310 in the heart ultrasound image is clear and easy to identify. In the heart ultrasound images of Fig. 6 and Fig. 7, there is no case of contrast agent 310 flowing from right to left. Patient A does not have an open patent foramen ovale.

[0114] Fig. 8 shows a heart ultrasound image of patient B. Patient B has been injected with contrast agent 310. Patient B is 18 years old, who cooperated with the Valsalva maneuver as instructed, "rest 1 level, 1 level after Valsalva maneuver", breath-holding time 15 seconds, and heart rate changed. The contrast agent 310 in the heart ultrasound image is clear and easy to identify. In the heart ultrasound image of Fig. 8, there is a trace of contrast agent 310 flowing from right to left. Patient B has an open patent foramen ovale.

[0115] According to the experimental results of the auxiliary diagnosis of patent foramen ovale based on Valsalva maneuver, the uniform foaming treatment of the contrast agent 310 in the present application makes the contrast agent 310 in the ultrasound image clear, thereby providing clear diagnostic reference information. The medical staff of the present application can timely take clear heart ultrasound images under the appropriate instruction reminder, without requiring the patient to repeat the Valsalva maneuver when the qualified image is not taken, and also reducing the difficulty of patient cooperation.

[0116] Embodiment 2

[0117] This embodiment is a further improvement of embodiment 1, and the repeated contents will not be described again.

[0118] This embodiment also provides a method for auxiliary diagnosis of patent foramen ovale based on Valsalva maneuver, as shown in Fig. 5, comprising the following steps.

[0119] S000: In the preparation process of the contrast agent 310, the contrast agent 310 is uniformly mixed with air by repeatedly pulling the syringe 320 to form a uniform mixture of air and contrast agent 310.

[0120] S100: When the blowing pressure of the patient 600 fed back by the blowing unit 100 reaches a first predetermined threshold and remains for a predetermined length of time, the control unit 500 starts to request the heart rate variation of the patient 600 output by the electrocardiogram monitoring unit 200. The present application uses the control unit 500 and the pressure sensor 110 to collect the data and length of time of the blowing pressure, so as to provide accurate judgment opportunity for collecting the heart rate variation, and avoid the patient 600 repeatedly performing the Valsalva action.

[0121] S110: When the control unit 500 issues the blowing instruction through the prompting unit 510, the blowing unit 100 monitors the blowing pressure of the pipeline 140 and its variation, and the control unit 500 triggers the timing instruction connected thereto to start the timer 560 based on the information condition that the blowing pressure reaches the first predetermined threshold; when the time of the blowing pressure reaching the first predetermined threshold reaches the predetermined length of time, the control unit 500 generates the instruction information of the patient 600 holding breath and the timing prompt information of the holding breath length.

[0122] S200: When it is determined that the heart rate variation of the patient 600 appears the inflection point related to the Valsalva action, the control unit 500 issues the prompt information of inputting the contrast agent 310 to the contrast agent injection unit 300. The present application analyzes the heart rate variation and judges the timing of the inflection point, so that the contrast agent injection unit 300 can timely inject the contrast agent 310 into the vein, avoid the injection timing of the contrast agent 310 being missed, and unify the injection timing of the contrast agent 310.

[0123] S210: When the heart rate of the patient 600 slows down and reaches a preset heart rate threshold, the control unit 500 issues the prompt information of inputting the contrast agent 310 to the contrast agent injection unit 300. When the preset heart rate threshold is reached, the inflection point of the heart rate variation is reached, and the accurate judgment and reminder of this timing can avoid the medical staff missing the best moment of the ultrasonic image shooting, so as to obtain the blood flow variation of the heart at the best moment, avoid missing the best shooting opportunity, and provide the most accurate diagnosis basis. The preset heart rate threshold is, for example, 5-10 times / min.

[0124] S220: A dose of 1ml air + 9ml physiological saline (contrast agent 310) is input into the vein.

[0125] S300: In response to the start information of the contrast agent vein input of the contrast agent injection unit 300, the control unit 500 issues a blowing instruction to the patient 600 to be tested through the prompting unit 510, and in the case where the blowing pressure of the patient 600 to be tested measured by the blowing unit 100 is lower than the second predetermined threshold, the control unit 500 issues a recording instruction of the blood flow dynamics in the heart through the prompting unit 510. In view of the defects of the judgment of the shooting opportunity according to the working experience and the defects that the blowing condition of the patient 600 to be tested cannot be directly observed, the present application can determine the shooting opportunity of the ultrasound image based on the change of the blowing condition, so as to timely capture the shunt condition in the heart and avoid missing the best shooting opportunity in the shooting action.

[0126] S310: In response to the end information of the contrast agent vein input of the contrast agent injection unit 300, the control unit 500 issues an exhalation instruction to the patient 600 to be tested through the prompting unit 510, and determines whether the blowing pressure of the patient 600 to be tested measured by the blowing unit 100 is lower than the second predetermined threshold.

[0127] S320: The control unit 500 sends a reminding instruction of the blood flow dynamics imaging in the heart to the ultrasound image imaging unit 400.

[0128] Embodiment 3

[0129] This embodiment is a further elaboration of Embodiment 1 and Embodiment 2, and the repeated contents will not be described again.

[0130] This embodiment provides a processor for constructing an artificial intelligence data set related to cardiac contrast agent injection, as shown in FIG. 9, the processor 530 is provided with an artificial intelligence semantic training model 540. When the processor 530 issues a blowing instruction to the patient 600 to be tested through the prompting unit 510, the processor 530 receives the blowing pressure of the patient 600 to be tested sent by the blowing unit 100 and the heart rate change condition of the patient 600 to be tested sent by the electrocardiogram monitoring unit 200. In the case where the prompting unit 510 forms the labeled prompting information, the artificial intelligence semantic training model 540 in the processor 530 labels the heart rate change rate of the patient 600 to be tested sent by the electrocardiogram monitoring unit 200, and prepares for constructing the training set of the heart rate change rate. The artificial intelligence semantic training model 540 calls the training set to analyze the heart rate change rate and timely judge the change inflection point of the heart rate change rate, so as to accurately remind the input time of the contrast agent 310.

[0131] In order to enable the artificial intelligence semantic training model 540 to learn to identify the change inflection point of the heart rate change rate, it is necessary to accurately label various data in the data acquisition process and express relevant information at appropriate time points.

[0132] The ECG monitoring unit 200 is immediately activated and starts monitoring the ECG variability in real-time as soon as the patient under test starts performing the Valsalva maneuver. The processor 530 receives the ECG signal through the ECG monitoring unit 200 and performs a preliminary analysis using the AI semantic training model 540. This process starts at the time point (T start ) when the Valsalva maneuver is initiated and records the initial data of the ECG variability.

[0133] For example, electrodes are attached to the surface of the patient's body to acquire the ECG signal. The ECG monitoring unit 200 collects the raw ECG signal from the electrodes. The ECG monitoring unit 200 transmits the ECG signal to the processor 530 through wired connection or wireless means such as Bluetooth or Wi-Fi. The processor 530 pre-processes the received signal to remove interference and noise. The processor 530 converts the ECG signal from analog form to digital form for subsequent analysis. The processor 530 extracts key features such as P-wave, QRS complex, etc. from the ECG signal using the AI semantic training model 540. The ECG variability is calculated and the initial data is recorded from T start .

[0134] For example, when the processor 530 is a microcontroller (MCU), the GPIO pin (General Purpose Input / Output) is the basic interface of the microcontroller for reading and writing digital signals. The ECG monitoring unit 200 is connected to the GPIO pin. The ADC module is connected to the GPIO pin for converting analog signals to digital signals for the microcontroller to process. The SPI is connected to the ADC module for high-speed communication between the microcontroller and peripherals.

[0135] First, the start signal of the Valsalva maneuver is read through the GPIO pin and monitoring is started. Then, the ADC (Analog-to-Digital Converter) module is used to receive the analog ECG signal from the ECG monitoring unit 200 and convert it to a digital signal. The ECG signal data is transmitted to the internal MCU through the Serial Peripheral Interface (SPI) or I2C protocol, and the DSP library inside the MCU is used for noise filtering and signal preprocessing. Next, the MCU extracts features and calculates the variability using the pre-trained model stored in the Flash, and monitors the feature changes in real-time through the interrupt mechanism, generating an alarm. The processed data is stored through the external EEPROM or SD card module, generating a detailed report, and the data is transmitted to the external system through the UART or USB interface for further analysis.

[0136] For example, when the processor 530 is a digital signal processor (DSP), the external trigger signal is input to the trigger input port of the DSP. The analog output pin of the electrocardiogram monitoring unit 200 is connected to the input pin of the ADC module. The digital output port of the ADC module is connected to the high-speed parallel interface (EMIF) input port of the DSP. The DSP monitoring is first started by the external trigger signal. The signal is collected from the electrocardiogram monitoring unit 200 using the high-speed ADC module, and then the signal data is transmitted to the DSP memory through the high-speed parallel interface (such as EMIF). The signal preprocessing is performed using the built-in filter and Fourier transform library of the DSP, and the complex feature extraction and rate of change calculation are performed using the high-performance computing capability of the DSP.

[0137] For example, when the processor 530 is a single-board computer, the start signal source is connected to the GPIO input pin of the single-board computer. The output pin of the electrocardiogram monitoring unit 200 is connected to the I2C or SPI interface input pin of the single-board computer. The Python script and GPIO pin are used to monitor the start signal. The single-board computer connects the electrocardiogram monitoring unit 200 through the I2C or SPI interface, acquires signal data, and transmits the data to the processor of the Raspberry Pi using the memory mapping technology. The signal preprocessing uses the NumPy and SciPy libraries for noise filtering and signal processing, and runs a pre-trained machine learning model (such as TensorFlow or PyTorch) for feature extraction and rate of change calculation.

[0138] When the insufflation pressure reaches the first predetermined threshold (e.g., 40 mmHg), the insufflation unit 100 detects and immediately transmits this information to the processor 530. At this time, the processor 530 records the time point (T threshold ) when the insufflation pressure reaches the predetermined threshold, and switches to a more intensive monitoring mode, focusing on the sudden change amplitude of the electrocardiogram rate. The electrocardiogram rate data during the period from the start of insufflation (T start ) to the insufflation pressure reaching the standard (T threshold ) is recorded.

[0139] The processor 530 analyzes the electrocardiogram rate data during the period from T start to T threshold , looking for sudden change amplitudes. When the monitored sudden change amplitude exceeds the maximum change amplitude during the previous monitoring period, and the exceeding degree reaches or exceeds the preset percentage (e.g., 30%, 50%, 70%), the time point (T inflection ) when the sudden change amplitude occurs is determined as the inflection point of the electrocardiogram rate change related to the Valsalva action.

[0140] In the determination of the inflection point of the electrocardiogram rate change (T inflection) After that, the processor 530 immediately sends prompt information for the injection to the contrast agent injection unit 300. The medical staff prepares to inject the contrast agent into the vein of the patient to be tested according to the indication of the prompt unit 510. Record the time point (T prepare ) when the prompt information is sent.

[0141] The processor 530 sends an injection start signal to the contrast agent injection unit 300 after confirming that the heart rate meets the preset condition. While the medical staff performs the injection action, the processor 530 continues to monitor the heart rate in real time to ensure that the injection timing is synchronized with the change inflection point (T inflection ) of the heart rate.

[0142] After the injection is completed, the processor 530 issues an exhalation instruction to the patient to be tested 600 through the prompt unit 510, and instructs the ultrasonic image imaging unit 400 to start recording the blood flow dynamics in the heart. Record the time point (T exhale ) when the exhalation instruction is sent and the time point (T imaging ) when the ultrasonic image imaging unit starts capturing and recording the change of the blood flow in the heart.

[0143] The format of the annotation of the heart rate of the patient to be tested 600 sent by the electrocardiogram monitoring unit 200 by the artificial intelligence semantic training model 540 is shown in Table 2.

[0144] Table 2: Example of annotation information of heart rate

[0145] Through the above detailed time point annotation and information description, the artificial intelligence semantic training model 540 can learn and identify the change inflection point of the heart rate, and can accurately determine the best time for the input of the contrast agent 310.

[0146] The process of the artificial intelligence semantic training model 540 learning to determine the change inflection point of the heart rate is as follows.

[0147] Normalize the electrocardiogram signal:

[0148] Where X(t) is the original electrocardiogram signal, μ is the mean, and σ is the standard deviation.

[0149] Extract features such as R-R interval, P wave, etc. from the normalized signal. Assume that the R-R interval is used as the main feature: RR i = R i -R i-1 (2).

[0150] Where R i i represents the time point of the i-th R wave.

[0151] The mutation amplitude of the heart rate variability is calculated using a sliding window. Assuming the window size is w, the heart rate variability of the data in the window is:

[0152] The mutation detection can be performed using a difference method. The difference method is:

[0153] When |d(t)| is greater than a certain threshold δ, it is determined to be a change inflection point. If |d(t)|>δ, it is determined that a change point is detected at time t.

[0154] The feature sequence is trained using a long short-term memory network (LSTM). Assuming the input features are X=[RR1, RR2, …, RR n ], the artificial intelligence semantic training model 540 outputs a prediction

[0155] The loss function of the artificial intelligence semantic training model 540 is defined as the mean square error:

[0156] The parameters θ of the artificial intelligence semantic training model 540 are optimized by gradient descent:

[0157] where η is the learning rate.

[0158] The inflection point of the heart rate variability predicted by the artificial intelligence semantic training model 540 is:

[0159] When the predicted value exceeds the first preset threshold α, it is determined to be a change inflection point of the heart rate variability, that is, if a change inflection point of the heart rate variability is detected at time t.

[0160] In practical applications, the artificial intelligence semantic training model 540 monitors the electrocardiogram data in real time, identifies the change inflection point, and judges the best injection time. The formula for determining the injection time is: set the threshold value β of the injection time, then

[0161] The specific injection time calculation is as follows:

[0162] If a change inflection point of the heart rate variability is detected at time t and the injection condition is met: t 注射 =t (9).

[0163] The process of learning to judge the injection time of the contrast agent 310 by the artificial intelligence semantic training model 540 according to the labeled information of the insufflation pressure is as follows.

[0164] The received insufflation pressure data is smoothed:

[0165] P(t) represents pressure data, and t represents time. smoothed (t) represents the pressure data after smoothing at time t, and w is the width of the smoothing window.

[0166] The normalization process involves mapping the pressure data to the range [0,1].

[0167] P norm (t) represents the pressure data after normalization at time t, P min and P max These represent the minimum and maximum values ​​of the pressure data, respectively.

[0168] The formula for calculating the rate of change of blowing pressure is:

[0169] ΔP(t) represents the rate of change of the blowing pressure at time t, and Δt represents the time step.

[0170] The time when the blowing pressure reaches the second predetermined threshold is: T pressure_low =min{t|P norm (t)≤P threshold} (13).

[0171] T pressure_low This indicates the time it takes for the blowing pressure to drop to the second set threshold. (P) threshold This indicates the second predetermined pressure threshold.

[0172] The formula used by the AI ​​semantic training model 540 to predict the optimal shooting time based on air pressure and injection information is as follows:

[0173] is the predicted imaging alert instruction time, and f(·) is the prediction function.

[0174] The formula for the AI ​​semantic training model 540 to issue an imaging alert command after confirming that the blowing pressure has reached the second predetermined threshold is: if P norm (t)≤P threshold ,then T imaging_remind =t (15).

[0175] Validate the model's accuracy using an untrained dataset and calculate the prediction error:

[0176] Where M is the number of validation samples. It is the prediction time for the i-th sample. That is the actual time.

[0177] The formula for further optimization according to the feedback of medical staff is:

[0178] θ new is the updated model parameter, θ old is the model parameter before updating, Δθ feedback is the parameter adjusted according to the feedback information.

Claims

1. An artificial intelligence data set construction method related to cardiac contrast agent injection, the method comprising: The Valsalva pressure of the patient to be tested (600) is measured by the blowing unit (100); The electrocardio change of the patient to be tested (600) is measured by the electrocardio monitoring unit (200); The method further comprises: The control unit (500) determines the injection timing of the contrast agent injection unit (300) and issues a recording instruction for the blood flow dynamics in the heart; Wherein, when the blowing pressure reaches the first predetermined threshold, the control unit (500) monitors the mutation amplitude of the electrocardio change rate in real time; when the mutation amplitude exceeds the change amplitude before the time point from the start of blowing to the time point when the blowing pressure reaches the first predetermined threshold, the time point when the mutation amplitude occurs is determined as the change inflection point of the electrocardio change rate; After determining the change inflection point of the electrocardio change rate, the control unit (500) sends prompt information for the preparation for injection to the contrast agent injection unit (300); After confirming that the electrocardio change rate meets the preset condition, the control unit (500) sends an injection start signal to the contrast agent injection unit (300); The artificial intelligence semantic training model (540) in the control unit (500) labels the electrocardio change rate of the patient to be tested (600) sent by the electrocardio monitoring unit (200), and then forms a training set of the electrocardio change rate. 2.The artificial intelligence data set construction method of claim 1, wherein, The method further comprises: The control unit (500) analyzes the electrocardio change rate data from the start of blowing to the pressure reaching the standard period, and finds the mutation amplitude; Once the monitored mutation amplitude exceeds the maximum change amplitude during the previous monitoring period, and the exceeding degree reaches or exceeds the preset percentage, the time point when the mutation amplitude occurs is determined as the change inflection point of the electrocardio change rate related to the Valsalva action. 3.The artificial intelligence data set construction method according to claim 1 or 2, characterized in that, The method further comprises: when the blowing pressure of the patient to be tested (600) fed back by the blowing unit (100) reaches the first predetermined threshold and remains for a predetermined time length, the control unit (500) starts to request the electrocardio change rate of the patient to be tested (600) output by the electrocardio monitoring unit (200). 4.The artificial intelligence data set construction method according to any one of claims 1-3, characterized in that, The method further comprises: In response to the received start information of the contrast agent intravenous input of the contrast agent injection unit (300), the control unit (500) sends blowing instructions to the patient to be tested (600) through the prompt unit (510), and the blowing unit (100) monitors the blowing pressure of the pipeline (140) and its change; The control unit (500) triggers the timing instruction that causes the timer (560) connected thereto to start timing based on the information condition that the blowing pressure reaches the first predetermined threshold; in the case that the time when the blowing pressure reaches the first predetermined threshold reaches a predetermined time length, the control unit (500) generates instruction information that causes the patient to be tested (600) to hold breath and timing prompt information of the holding breath time length. 5.The artificial intelligence data set construction method according to any one of claims 1-4, characterized in that, The method of confirming that the electrocardio change rate meets the preset condition comprises: When the heart rate of the patient to be tested (600) slows down and reaches a preset heart rate threshold, the control unit (500) sends prompt information for inputting the contrast agent (310) to the contrast agent injection unit (300). 6.The artificial intelligence data set construction method according to any one of claims 1-5, characterized in that, The method further comprises: After the injection is completed, the control unit (500) issues an exhale instruction to the patient (600) to be tested through the prompting unit (510), and instructs the ultrasound imaging unit (400) to start recording the blood flow dynamics in the heart; Wherein, in response to the end information of the contrast agent vein input of the contrast agent injection unit (300), the control unit (500) issues an exhale instruction to the patient (600) to be tested through the prompting unit (510), and determines whether the measured blowing pressure of the patient (600) to be tested by the blowing unit (100) is lower than a second predetermined threshold value; In the case where the blowing pressure is lower than the second predetermined threshold value, the control unit (500) issues a recording instruction of the blood flow dynamics in the heart through the prompting unit (510).

7. An artificial intelligence data set construction apparatus related to a cardiac contrast agent injection, the apparatus comprising: A blowing unit (100) for measuring the blowing pressure of the patient (600) to be tested; An electrocardio monitoring unit (200) for measuring the electrocardio change condition of the patient (600) to be tested; A contrast agent injection unit (300) for inputting a contrast agent (310) to the patient (600) to be tested; It is characterized in that it further comprises: A control unit (500) for determining the injection timing of the contrast agent injection unit (300) and issuing a recording instruction of the blood flow dynamics in the heart; Wherein, the control unit (500) is configured to: When the blowing pressure reaches the first predetermined threshold value, the control unit (500) immediately monitors the mutation amplitude of the electrocardio change rate; when the mutation amplitude exceeds the change amplitude before the time point from the start of blowing to the time point when the blowing pressure reaches the first predetermined threshold value, the time point when the mutation amplitude occurs is determined as the change inflection point of the electrocardio change rate; after determining the change inflection point of the electrocardio change rate, the control unit (500) sends a prompt information of the preparation for injection to the contrast agent injection unit (300); After confirming that the electrocardio change rate meets the preset condition, the control unit (500) sends an injection start signal to the contrast agent injection unit (300); The artificial intelligence semantic training model (540) in the control unit (500) labels the electrocardio change rate of the patient (600) to be tested sent by the electrocardio monitoring unit (200), and further forms a training set of the electrocardio change rate.

8. The artificial intelligence data set construction apparatus according to claim 7, wherein, The control unit (500) analyzes the electrocardio change rate data from the start of blowing to the pressure meeting period, and finds the mutation amplitude; Once the monitored mutation amplitude exceeds the maximum change amplitude in the previous monitoring period, and the exceeding degree reaches or exceeds the preset percentage, the time point when the mutation amplitude occurs is determined as the change inflection point of the electrocardio change rate related to the Valsalva action.

9. The artificial intelligence data set construction device according to claim 7 or 8, characterized in that, After the injection is completed, the control unit (500) issues an exhale instruction to the patient (600) to be tested through the prompting unit (510), and instructs the ultrasound imaging unit (400) to start recording the blood flow dynamics in the heart; Wherein, in response to the end information of the contrast agent intravenous input of the contrast agent injection unit (300), the control unit (500) issues an exhale instruction to the patient (600) to be tested through the prompting unit (510), and determines whether the insufflation pressure of the patient (600) to be tested measured by the insufflation unit (100) is lower than a second predetermined threshold value; In the case where the insufflation pressure is lower than the second predetermined threshold value, the control unit (500) issues a recording instruction of the blood flow dynamic situation in the heart through the prompting unit (510). 10.The artificial intelligence data set construction apparatus according to any one of claims 7 to 9, characterized in that, In the case where the insufflation pressure of the patient (600) to be tested fed back by the insufflation unit (100) reaches a first predetermined threshold value and remains for a predetermined time length, the control unit (500) starts to request the heart rate of the patient (600) to be tested output by the electrocardiogram monitoring unit (200). 11.The artificial intelligence data set construction apparatus according to any one of claims 7 to 10, characterized in that, In response to the start information of the contrast agent intravenous input of the contrast agent injection unit (300) received, the control unit (500) issues an insufflation instruction to the patient (600) to be tested through the prompting unit (510), and the insufflation unit (100) monitors the insufflation pressure of the pipeline (140) and its change; The control unit (500) triggers a timing instruction that causes the timer (560) connected thereto to start timing based on the information condition that the insufflation pressure reaches the first predetermined threshold value; in the case where the time when the insufflation pressure reaches the first predetermined threshold value reaches a predetermined time length, the control unit (500) generates instruction information that causes the patient (600) to be tested to hold breath and timing prompt information of the breath holding time length.

12. A patent foramen ovale auxiliary diagnosis device based on Valsalva action, characterized in that, an insufflation unit (100) for measuring the insufflation pressure of the patient (600) to be tested; an electrocardiogram monitoring unit (200) for measuring the electrocardiogram change of the patient (600) to be tested; a contrast agent injection unit (300) for inputting a contrast agent (310) to the patient (600) to be tested; characterized in that, in the case where the insufflation pressure of the patient (600) to be tested fed back by the insufflation unit (100) reaches a first predetermined threshold value and remains for a predetermined time length, the control unit (500) starts to request the heart rate of the patient (600) to be tested output by the electrocardiogram monitoring unit (200); when it is determined that the heart rate of the patient (600) to be tested appears a change inflection point related to Valsalva action, the control unit (500) issues prompt information of the input of the contrast agent (310) to the contrast agent injection unit (300), and synchronizes the contrast agent injection occasion with the change inflection point of the heart rate. In response to the received start information of the contrast agent vein input of the contrast agent injection unit (300), the control unit (500) issues a blowing instruction to the patient (600) to be tested through the prompting unit (510), and in the case where the blowing pressure of the patient (600) to be tested measured by the blowing unit (100) is determined to be lower than a second predetermined threshold value, the control unit (500) issues a recording instruction of the blood flow dynamic situation in the heart through the prompting unit (510) to record the blood flow dynamic situation in the heart. The determination of the change inflection point of the heart rate change rate associated with the Valsalva action in time is performed in the following manner: When the mutation amplitude exceeds the change amplitude before the time point from the start of blowing to the time point when the blowing pressure reaches the first predetermined threshold value, the time point when the mutation amplitude occurs is determined as the inflection point; the control unit (500) analyzes the heart rate change rate data from the start of blowing to the pressure reaching standard, and finds the mutation amplitude; when the mutation amplitude is monitored to exceed the maximum change amplitude during the previous monitoring period, and the exceeding degree reaches or exceeds the preset percentage, the time point when the mutation amplitude occurs is determined as the inflection point of the change of the heart rate change rate associated with the Valsalva action.

13. The Valsalva action-based patent foramen ovale auxiliary diagnosis device according to claim 12, characterized in that, When the blowing instruction is issued, the control unit (500) forms a labeling prompt information through the prompting unit (510), and the artificial intelligence semantic training model (540) in the control unit (500) labels the heart rate change rate of the patient (600) to be tested sent by the electrocardio monitoring unit (200), thereby forming a training set of the heart rate change rate.

14. The Valsalva-based patent foramen ovale auxiliary diagnostic device according to claim 12 or 13, characterized in that In the case where the blowing instruction is issued by the control unit (500) through the prompting unit (510), the blowing unit (100) monitors the blowing pressure of the pipeline (140) and its change, The control unit (500) triggers a timing instruction that causes the timer (560) connected thereto to start timing based on the information condition that the blowing pressure reaches the first predetermined threshold value; In the case where the time when the blowing pressure reaches the first predetermined threshold value reaches a predetermined length of time, the control unit (500) generates instruction information that causes the patient (600) to be tested to hold breath and timing prompt information of the holding breath duration.

15. The Valsalva-based patent foramen ovale aid diagnostic apparatus according to any one of claims 12 to 14, characterized in that, The blowing unit (100) comprises a pressure sensor (110), a pressure communication port (120), a blowing assembly (130) and a pipeline (140), The pressure sensor (110) is arranged in the pipeline (140) connected with the blowing assembly (130) to collect the blowing pressure; The pressure sensor (110) sends the data information of the blowing pressure to the control unit (500) through the pressure communication port (120).

Images