Intelligent monitoring method and device for hemorrhagic transformation after thrombolysis and thrombectomy treatment of cerebral infarction

Abstract

This invention relates to the field of artificial intelligence technology, and in particular to an intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage conversion after cerebral infarction. The device comprises: a wearable monitoring component for continuous monitoring of the monitored individual to obtain a dataset of disturbance coefficients of the monitored individual under electromagnetic wave signals along five propagation paths; and a main control device for data communication with the wearable monitoring component, used for automatic monitoring based on the disturbance coefficient dataset continuously detected by the wearable monitoring component, and for issuing early warnings based on the monitoring results.

Description

Technical Field

[0001] This invention relates to artificial intelligence, and more particularly to intelligent monitoring of post-thrombolytic and thrombectomy hemorrhage conversion based on artificial intelligence. Background Technology

[0002] Cerebrovascular disease is a common ailment among middle-aged and elderly people, characterized by its rapid onset, severe symptoms, and dangerous progression. It is characterized by high incidence, high mortality, high disability, and high recurrence rates. Post-infarction emergency treatments mainly include thrombolysis and endovascular therapy. However, these advanced technologies and treatments increase the risk of secondary hemorrhagic transformation. For example, after thrombolytic therapy for acute ischemic stroke, three major complications typically occur: hemorrhage, vascular re-occlusion, and reperfusion injury. Among these, hemorrhagic transformation is the most dangerous, affecting the efficacy and prognosis of acute ischemic stroke. Therefore, hemorrhagic transformation has become a key focus in the treatment and monitoring of ischemic stroke. Secondary hemorrhagic transformation refers to bleeding caused by the restoration of blood flow to the ischemic area after acute ischemic stroke, directly impacting the prognosis of stroke patients. Hemorrhagic transformation can lead to varying degrees of space-occupying lesions in brain tissue, cerebral edema, and other neurological deterioration, endangering the patient's life and representing one of the challenges in medical research. Studies show that the incidence of hemorrhagic transformation after thrombolysis is 10%-48%; the incidence after endovascular treatment is 46%-49.5%. Currently, doctors, patients, and their families in China have concerns and fears about secondary hemorrhagic transformation after endovascular treatments such as thrombolysis and mechanical thrombectomy. The main mechanisms of hemorrhagic transformation include post-infarction ischemic injury, reperfusion injury, coagulation disorders, and blood-brain barrier disruption. Existing medical monitoring methods have very limited accuracy in predicting hemorrhagic transformation after treatment, and often the condition has already worsened and clinical symptoms have appeared by the time it is detected.

[0003] Currently, the detection of secondary hemorrhagic transformation after treatment mainly relies on clinical experience. Further refinement methods are now employed, relying on head imaging examinations such as CT / MRI to determine the occurrence of secondary hemorrhagic transformation. The judgment is based on the initial head CT / MRI after a stroke not detecting hemorrhage, but a subsequent head CT / MRI after endovascular treatment revealing intracranial hemorrhage; this confirms hemorrhagic transformation. Existing technologies propose using CT images for model training to detect hemorrhagic areas. For example, Chinese invention patent CN106296653B discloses a method and system for segmenting hemorrhagic areas in brain CT images based on semi-supervised learning. This method identifies hemorrhagic areas by converting CT images from two-dimensional space to three-dimensional space and then performing image segmentation. For example, Chinese invention patent application CN110503630A discloses a method for classifying, locating, and predicting cerebral hemorrhage based on a three-dimensional deep learning model. It obtains three-dimensional CT images by performing three-dimensional modeling on two-dimensional CT images, and uses three-dimensional convolutional neural networks, target detection networks, and three-dimensional conditional generative adversarial networks to quickly and accurately determine whether a patient has cerebral hemorrhage. It can also accurately locate the bleeding point and generate three-dimensional CT images based on the patient's physical indicators, thereby predicting the patient's condition development.

[0004] While methods like CT / MRI are highly accurate in detecting secondary hemorrhagic transformation, they all involve acquiring CT / MRI images after the transformation has occurred, and then making judgments based on these images. That is, CT / MRI imaging after endovascular treatment typically occurs when the patient experiences neurological deterioration and worsening clinical symptoms, i.e., after secondary hemorrhagic transformation has taken place. However, neurological deterioration is a dynamic process, and head imaging methods cannot provide continuous, dynamic monitoring at the bedside. Traditional methods suffer from significant lag and cannot meet the demands of new treatment technologies. Therefore, there is an urgent need to innovate a simple, rapid, dynamic, continuous, and accurate monitoring device for secondary hemorrhagic transformation. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent monitoring device and method for hemorrhage transformation after thrombolysis and thrombectomy for cerebral infarction, which partially solves or alleviates the above-mentioned deficiencies in the prior art, enables dynamic and continuous bedside monitoring, and can easily, quickly and accurately monitor secondary hemorrhage transformation.

[0006] To partially address or alleviate the aforementioned deficiencies in the prior art, a first aspect of the present invention provides an intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation in cerebral infarction, comprising:

[0007] A wearable monitoring device is used to continuously monitor a person under surveillance to obtain a dataset of disturbance coefficients of the person under surveillance under electromagnetic wave signals in a preset frequency band and with five preset propagation paths. The person under surveillance is a patient who has undergone thrombolysis and thrombectomy for cerebral infarction. The wearable monitoring device includes: a wearable component for wearing on the head of the person under surveillance, and four electrode pads disposed on the wearable component. When the person under surveillance wears the wearable component, the four electrode pads are respectively located at four designated monitoring points on the head of the person under surveillance. The four designated monitoring points are: a fourth monitoring point located on the forehead, a third monitoring point located above the left ear, a first monitoring point located above the right ear, and a second monitoring point located at the back of the head. A first propagation path is formed between the first monitoring point and the third monitoring point; a second propagation path is formed between the first monitoring point and the fourth monitoring point; a third propagation path is formed between the first monitoring point and the second designated monitoring point; a fourth propagation path symmetrical to the second propagation path is formed between the fourth monitoring point and the third designated monitoring point; and a fifth propagation path symmetrical to the third propagation path is formed between the third monitoring point and the second monitoring point.

[0008] The main control device communicates with the wearable monitoring component to automatically monitor the perturbation coefficient dataset obtained through continuous monitoring by the wearable monitoring component and to issue early warnings based on the monitoring results. The main control device includes: a sample library construction module, used to acquire perturbation coefficient datasets obtained through continuous monitoring of electromagnetic wave signals along the first, second, third, fourth, and fifth propagation paths for subjects without hemorrhage after treatment surgery, and perturbation coefficient datasets obtained through continuous monitoring of electromagnetic wave signals along the first, second, third, fourth, and fifth propagation paths for subjects with secondary hemorrhage after treatment surgery, to construct a training sample library; and a model training module, used to learn from the training samples in the training sample library using machine learning algorithms to construct feature parameter mappings respectively. A mapping relationship is established between the classification labels of secondary hemorrhage and a classifier that can be automatically monitored based on the mapping relationship. The feature parameters include: the maximum absolute difference of perturbation coefficients between pairs of symmetrical propagation paths, the difference in the degree of fluctuation of perturbation coefficients between pairs of symmetrical propagation paths, the degree of fluctuation of whole-brain perturbation coefficients obtained from continuous monitoring over multiple days, and the difference in perturbation coefficients between adjacent two days. A data acquisition module is used to periodically or in real-time acquire perturbation coefficient datasets of the monitored subject's brain under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths from the wearable monitoring component. A monitoring module is used to input the perturbation coefficient dataset acquired by the data acquisition module into the classifier for automatic monitoring and to provide early warning prompts based on the monitoring results.

[0009] Generally, if secondary hemorrhage transformation occurs on one side, its perturbation coefficient will definitely change. Therefore, the absolute change in the perturbation coefficient on that side is characterized by the maximum value of the absolute difference between the perturbation coefficient on that side and the symmetrical side.

[0010] In some embodiments of the present invention, the formula for calculating the fluctuation degree P of the disturbance coefficient is as follows:

[0011] Where F1 is the maximum amplitude in the upward trend, F2 is the minimum amplitude in the upward trend, F3 is the maximum amplitude in the downward trend, and F4 is the minimum amplitude in the downward trend; a and b are the weighting coefficients for the upward and downward trends, respectively; t1 is the start time of the upward trend, t2 is the duration of the upward trend; t3 is the start time of the downward trend, and t4 is the duration of the downward trend.

[0012] Typically, the perturbation coefficient of regions where secondary hemorrhage transformation occurs fluctuates more significantly than that of healthy symmetrical regions (which also fluctuate, but usually less so we set the fluctuation level P=0). Therefore, by constructing a mapping relationship between the difference in the fluctuation level of the perturbation coefficient between symmetrical propagation path pairs and the classification label, we can, on the one hand, reduce the probability of misclassification caused by fluctuations in the perturbation coefficient due to other diseases; on the other hand, we can introduce more features to avoid the problem of large errors in the learning process due to too few features.

[0013] In some embodiments of the present invention, when the disturbance coefficient monitored over multiple days exhibits two changing trends, and either of these trends occurs at least once, Among them, F1 i F2 i Let a be the maximum and minimum amplitude values ​​in the i-th upward trend out of n upward trends. i F3 is the weighting coefficient for the i-th upward trend. j F4 j b represents the maximum and minimum amplitude values ​​in the j-th downward trend out of m downward trends; j is the weighting coefficient for the j-th downward trend.

[0014] In some embodiments of the present invention, the model training module is specifically used to construct a mapping relationship between the maximum absolute difference of perturbation coefficients between symmetric propagation path pairs and the difference of perturbation coefficient fluctuation degree P, based on the perturbation coefficient dataset within 36 hours after treatment surgery in the training sample, to obtain a first classifier for automatically monitoring the monitored individuals within 36 hours after treatment surgery based on the mapping relationship; and to construct a mapping relationship between the fluctuation degree of whole-brain perturbation coefficients obtained from continuous monitoring over multiple days and the difference of perturbation coefficients between adjacent two days, based on the perturbation coefficient dataset after 36 hours after treatment surgery in the training sample, to obtain a second classifier for automatically monitoring the monitored individuals after 36 hours after treatment surgery based on the mapping relationship.

[0015] In some embodiments of the present invention, the main control device further includes: a database for pre-storing a dataset of perturbation coefficients obtained by continuous monitoring of a number of subjects who have undergone treatment surgery without bleeding transformation under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths, and a dataset of perturbation coefficients obtained by continuous monitoring of a number of subjects who have undergone treatment surgery with secondary bleeding transformation under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths.

[0016] In some embodiments of the present invention, the main control device further includes: a condition acquisition module, which communicates with the medical information system of a medical institution to acquire the patient's medical records, including the time of the thrombolysis and thrombectomy procedure for cerebral infarction and historical cases; and a control module, which, when the time interval between the current time and the completion of the procedure is less than 36 hours, acquires a pre-set continuous monitoring duration and the frequency of continuous monitoring per day, and controls the monitoring module to perform automatic monitoring using the first classifier; or, when the time interval between the current time and the completion of the procedure is more than 36 hours, reduces the frequency of continuous monitoring per day and controls the monitoring module to perform automatic monitoring using the second classifier.

[0017] In some embodiments of the present invention, the feature parameters further include: the variance of the perturbation coefficient in the upward trend and / or the downward trend. Variance characterizes the volatility of the perturbation coefficient; therefore, by constructing a mapping relationship between the variance of the perturbation coefficient and the classification label, the probability of misjudgment caused by fluctuations in the perturbation coefficient due to other medical conditions can be further reduced.

[0018] In some embodiments of the present invention, the preset frequency band includes an ultra-low frequency band and a low frequency band, and both the ultra-low frequency band and the low frequency band are divided into at least two sub-frequency bands; the multiple sub-frequency bands obtained by dividing the preset frequency band include: a first sub-frequency band with a frequency range of 50Hz-100Hz, a second sub-frequency band with a frequency range of 100Hz-500Hz, a third sub-frequency band with a frequency range of 500Hz-1kHz, a fourth sub-frequency band with a frequency range of 1kHz-10kHz, a fifth sub-frequency band with a frequency range of 10kHz-100kHz, and a sixth sub-frequency band with a frequency range of 100kHz-300kHz; wherein, the perturbation coefficient dataset corresponding to the ultra-low frequency band 50Hz-500Hz accounts for the largest proportion of the total training samples.

[0019] In some embodiments of the present invention, the main control device further includes: the human-computer interaction module, used to display monitoring results.

[0020] In some embodiments of the present invention, the main control device further includes: a database for pre-storing a dataset of perturbation coefficients obtained by continuous monitoring of a number of subjects who have undergone treatment surgery without bleeding transformation under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths, and a dataset of perturbation coefficients obtained by continuous monitoring of a number of subjects who have undergone treatment surgery with secondary bleeding transformation under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths.

[0021] A second aspect of the present invention provides a method for intelligent monitoring of hemorrhage conversion after thrombolysis and thrombectomy for cerebral infarction, comprising the following steps:

[0022] A training sample library was constructed by acquiring perturbation coefficient datasets [R11; R12; R13; R14; R15] obtained by continuously monitoring subjects without hemorrhage transformation after treatment surgery under electromagnetic wave signals of five preset propagation paths, and perturbation coefficient datasets [R21; R22; R23; R24; R25] obtained by continuously monitoring subjects with secondary hemorrhage transformation after treatment surgery under electromagnetic wave signals of five preset propagation paths. The five propagation paths are respectively located in: the middle region between the first monitoring point and the third monitoring point, the left anterior region between the first monitoring point and the fourth monitoring point, the left posterior region between the first monitoring point and the second monitoring point, the right anterior region between the fourth monitoring point and the third monitoring point, and the right posterior region between the third monitoring point and the second monitoring point, wherein the left anterior region is symmetrical to the right anterior region, and the left posterior region is symmetrical to the right posterior region.

[0023] Machine learning algorithms are used to learn from training samples in the training sample library to construct mapping relationships between feature parameters and secondary hemorrhage transformation classification labels, and a classifier that can automatically monitor based on the mapping relationship is obtained. The feature parameters include: the maximum absolute difference of perturbation coefficients between pairs of symmetrical regions, and the difference in the degree of fluctuation P of perturbation coefficients between pairs of symmetrical regions; the degree of fluctuation of the whole-brain perturbation coefficient obtained from continuous monitoring over multiple days, and / or the difference in perturbation coefficients between adjacent two days. The classifier includes a first classifier based on the mapping relationship between the maximum absolute difference of the perturbation coefficients between pairs of symmetrical regions and the difference in the degree of fluctuation P of perturbation coefficients and the secondary hemorrhage transformation classification label; and a second classifier based on the mapping relationship between the degree of fluctuation of the whole-brain perturbation coefficients and / or the difference in perturbation coefficients between adjacent two days and the secondary hemorrhage transformation classification label.

[0024] The perturbation coefficient dataset [R1; R2; R3; R4; R5] of electromagnetic wave signals under preset frequency bands and five preset propagation paths is acquired periodically or in real time after the patient undergoes treatment surgery. The perturbation coefficient dataset is then input into the classifier for automatic monitoring, and early warning prompts are issued based on the monitoring results.

[0025] A third aspect of the present invention is to provide a computer program product, comprising a computer program that, when executed by a processor, implements the above-described intelligent monitoring method for hemorrhage conversion after thrombolysis and thrombectomy for cerebral infarction.

[0026] The beneficial effects of this invention are as follows: Traditional short-term detection methods for secondary hemorrhagic transformation, such as CT, ultrasound, and MRI, require clinicians to make judgments based on clinical experience after certain symptoms have appeared. Therefore, they are inherently time-consuming. Furthermore, the randomness of secondary hemorrhagic transformation greatly complicates the clinician's judgment. Automated detection through machine learning requires a large amount of effective training data. However, traditional detection methods are neither capable of long-term detection nor are they time-consuming, making it currently impossible to obtain effective training data. This invention relates to an intelligent monitoring device for hemorrhagic transformation after thrombolysis and thrombectomy for cerebral infarction. It continuously monitors subjects with secondary hemorrhagic transformation and those without hemorrhagic transformation after surgery under electromagnetic wave signals at a preset frequency and along five specified propagation paths. The device obtains a dataset of perturbation coefficients and performs machine learning to construct feature parameters, such as the maximum absolute difference in perturbation coefficients between symmetrical propagation paths, the degree of fluctuation, and a mapping relationship between the classification labels for secondary hemorrhagic transformation (i.e., those potentially considered as having secondary hemorrhagic transformation) and non-secondary hemorrhagic transformation. This allows for automatic monitoring based on this mapping relationship and provides early warnings based on the automatic monitoring.

[0027] Typically, in monitored individuals experiencing secondary hemorrhage transformation, the perturbation coefficient of the hemorrhage area is higher than that of the corresponding side. However, in actual monitoring, it's impossible to immediately determine that secondary hemorrhage transformation has occurred simply because the perturbation coefficient is greater than that of the corresponding side, or to conclude that secondary hemorrhage transformation has occurred based on the difference between the perturbation coefficients of the two symmetrical sides at a single moment. This is because the abnormality could be due to unstable equipment connections or other medical conditions. Furthermore, secondary hemorrhage transformation is persistent (i.e., it usually lasts for a certain period of time). Therefore, compared to directly comparing a single moment... The perturbation coefficients between symmetrical regions are treated as a whole in this application (i.e., the changing trend of the perturbation system over time is considered as a whole) for model training. This allows the construction of a mapping relationship between the classification label of (suspected) secondary hemorrhage transformation and feature parameters (e.g., the difference in the degree of fluctuation of the perturbation coefficients between symmetrical regions). This enables intelligent monitoring based on the mapping relationship, and then automatic monitoring, i.e., detection of suspected secondary hemorrhage transformation in monitored individuals. The system then issues an early warning to prompt a final diagnosis through traditional detection methods such as CT / MRI.

[0028] As is well known, training requires a large amount of training data. However, subjects typically undergo impedance spectrum testing at hospitals or clinics after experiencing certain symptoms. This means that the amount of training sample data is limited. To expand the amount of training sample data, the method of this invention, on the one hand, switches the propagation path of electromagnetic wave signals by changing the working mode of the electrode group. Compared to frequently adjusting the electrode placement, this avoids increasing the complexity of training sample data acquisition due to frequent adjustments, reducing the efficiency of training sample data acquisition, and preventing the introduction of excessive noise due to the inability to guarantee that each adjustment achieves the ideal state. On the other hand, to expand the training sample with limited data, ACGAN is used to amplify the training sample data, thereby improving the discrimination accuracy of the stroke classifier.

[0029] On the other hand, CT scans, ultrasound, and MRI can accurately detect the location of lesions and even the severity of the condition. Therefore, if machine learning is performed using image data acquired through these methods, the accuracy will certainly be higher. However, as mentioned earlier, these methods are inherently time-consuming and only detect for a short period, failing to capture the entire developmental stage of hemorrhagic transformation. This limits the amount of data available for machine learning. Furthermore, these methods collect all brain data, requiring significant time for data cleaning and screening during machine learning, typically accounting for 80% of the time. Therefore, using traditional detection methods not only introduces a lot of unnecessary interference data but also greatly increases the time spent on data cleaning and screening, thus reducing the efficiency of machine learning. In this invention, by using five preset propagation paths and collecting perturbation coefficient datasets under electromagnetic wave excitation along these paths, it can comprehensively cover areas where secondary hemorrhagic transformation may occur while ensuring sufficient data acquisition without introducing excessive interference data, significantly reducing the efficiency of machine learning. On the other hand, this application treats the perturbation coefficient dataset obtained by continuous monitoring, that is, the trend of perturbation coefficient changes over time, as a whole, and seeks the mapping relationship between the feature parameters (e.g., the degree of fluctuation) corresponding to the trend and the secondary hemorrhage transformation classification label, instead of constructing the mapping relationship between the perturbation coefficient and the secondary hemorrhage transformation classification label at each moment. This greatly reduces the amount of data that needs to be preprocessed and labeled in the early stage of training, thereby greatly improving the efficiency of machine learning. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0031] Figure 1 This is a functional block diagram of an intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage conversion in an exemplary embodiment of the present invention.

[0032] Figure 2 This is a functional block diagram of the bioelectrical impedance measurement device in an exemplary embodiment of the present invention, which is a smart monitoring device for post-thrombolysis and thrombectomy hemorrhage conversion after cerebral infarction.

[0033] Figure 3This is a schematic diagram illustrating four designated monitoring points and five preset propagation paths;

[0034] Figure 4 This is a block diagram illustrating the structural composition of an electronic device according to an exemplary embodiment of the present invention;

[0035] Figure 5 A flowchart illustrating an exemplary embodiment of the present invention is provided, showing a method for intelligent monitoring of post-stroke hemorrhage conversion after thrombolysis and thrombectomy for cerebral infarction.

[0036] Figure 6 This is a trend graph of perturbation coefficient changes obtained from a dataset of perturbation coefficients obtained by subjects who were continuously monitored for 5 hours after treatment surgery without secondary hemorrhage transformation.

[0037] Figure 7 This is a trend graph of perturbation coefficient changes obtained from the perturbation coefficient dataset of the first subject who developed secondary hemorrhage after treatment surgery and was continuously monitored for 5 hours.

[0038] Figure 8 This is a trend graph of the perturbation coefficient obtained from the perturbation coefficient dataset of a second subject who developed secondary hemorrhage after treatment surgery and was continuously monitored for 5 hours.

[0039] Figure 9 This is a trend graph of the perturbation coefficient obtained from the perturbation coefficient dataset of a third subject who developed secondary hemorrhage after treatment surgery and was continuously monitored for 5 hours.

[0040] Figure 10a Case 1: An early warning system for continuous monitoring of wards over multiple days;

[0041] Figure 10b For based on Figure 10a A trend chart of the change in the disturbance coefficient obtained from continuous monitoring over multiple days;

[0042] Figure 11a Case 2 serves as an early warning for the continuous monitoring of wards over multiple days;

[0043] Figure 11b For based on Figure 11a A trend chart of the change in the disturbance coefficient obtained from continuous monitoring over multiple days;

[0044] Figure 12a Case 3: Early warning system for continuous monitoring of wards over multiple days;

[0045] Figure 12b For based on Figure 12a The trend of the disturbance coefficient obtained from continuous monitoring over multiple days. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In this document, suffixes such as "module," "component," or "unit" used to represent elements are only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "component," or "unit" can be used interchangeably. In this document, the terms "upper," "lower," "left," "right," "inner," "outer," "front," "rear," "one end," "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In this document, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," and "connected," etc., should be interpreted broadly. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; it can also refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. In this document, "full frequency band" refers to the frequency range of the electromagnetic wave signal applied to the subject's brain, from 50Hz to 300kHz. Below 500Hz is the ultra-low frequency band, 500Hz-10kHz is the low frequency band, and above 100kHz is the high frequency band; correspondingly, 10kHz-100kHz is the mid-frequency band.

[0047] In this article, the "disturbance coefficient (DC)" refers to the brain impedance disturbance coefficient under electromagnetic wave excitation. Normally, the disturbance coefficient in a healthy person is around 140, and the disturbance coefficients in the left and right hemispheres are essentially equal, with an absolute difference of less than 0.3. However, once secondary hemorrhagic transformation occurs, the disturbance coefficients in the left and right hemispheres will no longer be the same, because the disturbance coefficient in the region where secondary hemorrhagic transformation occurs will change. Therefore, the difference in the disturbance coefficients between symmetrical regions is usually used to determine whether secondary hemorrhagic transformation has occurred. For example, when the disturbance coefficients corresponding to symmetrical regions are monitored, if the difference between the two is greater than a preset threshold, an abnormality is determined, i.e., suspected secondary hemorrhagic transformation has occurred. However, clinical practice has shown that although the duration of secondary hemorrhage is relatively short, ranging from 30 to 100 minutes, and usually occurs within 36 hours after surgical intervention (thrombectomy, thrombolysis, hematoma removal, etc.) (of course, it may also occur after 36 hours), the perturbation coefficient gradually changes as the secondary hemorrhage progresses. This results in a dynamic change in the absolute difference between the perturbation coefficients of the left and right hemispheres, which will continue for a certain period of time.

[0048] In this document, "continuous monitoring" or "continuous surveillance" refers to a total monitoring duration T0 greater than 30 minutes for each monitoring session. Since secondary cerebral hemorrhage is random and each episode lasts 30-100 minutes, each monitoring session must be at least 30 minutes long, although multiple continuous monitoring sessions can be performed daily. For example, in some embodiments, during continuous monitoring, if the absolute difference between the perturbation coefficient of the propagation path in the left anterior region of Ti (0 < i < T0) and its symmetrical side (i.e., the right anterior region) at a certain moment is greater than a preset absolute difference (e.g., 5), the monitoring time is automatically recalculated from the start of the upward or downward trend of the perturbation coefficient on the corresponding side until the upward or downward trend ends. This ensures that the continuous monitoring time covers the complete change process of the perturbation system. See [link to relevant documentation]. Figure 7 and Figure 8 For example, in some embodiments, when two trends exist simultaneously: an upward trend and a downward trend, and the time interval between the two trends is less than a preset time interval (e.g., when the trend first rises and then falls, the time interval between the end of the upward trend and the beginning of the downward trend is less than a preset 1-10 minute interval), the trend is further defined as follows: Figure 9 As shown; or, when the trend first declines and then rises, if the time interval between the end of the downward trend and the start of the upward trend is less than the preset 1min-10min, the monitoring time is automatically recalculated from the start of the first trend of change of the disturbance coefficient (e.g., upward or downward trend) until the end of the second trend.

[0049] In this article, "continuous monitoring" refers to continuous monitoring over multiple days, with multiple consecutive monitoring sessions each day. For example, from the day after the patient's surgery until the day of discharge, continuous monitoring is conducted daily at a predetermined frequency.

[0050] Example 1: See Figure 1 This is a functional block diagram of an intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage conversion in cerebral infarction, an exemplary embodiment of the present invention. Specifically, the intelligent monitoring device includes:

[0051] A bioelectrical impedance measurement device is used for continuous monitoring of a monitored individual. It measures the electrical impedance of the monitored individual at multiple target frequencies across the entire frequency band under electromagnetic wave signals propagating through five preset paths. The device then processes the electrical impedance data to obtain the corresponding perturbation coefficients. Specifically, it includes a wearable monitoring component used to continuously transmit electromagnetic waves of a preset frequency band to the monitored individual's brain along the five preset propagation paths, and to acquire a dataset of perturbation coefficients [R1; R2; R3; R4; R5] under the electromagnetic wave signals. The monitored individual is a patient who has undergone thrombolysis and thrombectomy for cerebral infarction.

[0052] The main control device communicates with the wearable monitoring component via data (e.g., wireless or wired communication) to automatically monitor based on the disturbance coefficient dataset obtained through continuous monitoring by the wearable monitoring device, and to issue early warnings based on the monitoring results.

[0053] In some embodiments, the wearable monitoring component specifically includes: a wearable component, four electrode pads disposed on the wearable component, and a control circuit electrically connected to the four electrode pads;

[0054] The control circuit acquires the target attribute control signal and the current attribute information of each electrode pad. Based on the target attribute control signal, it changes the attribute terminals electrically connected to the two electrode pads along each propagation path. Each time the target attribute control signal changes, the electrode pads acting as transmitting and receiving electrodes also change, thus altering the propagation path or direction of the electromagnetic wave signal. Therefore, without changing the placement of each electrode pad on the subject's brain, only different target attribute control signals are needed to change the attribute terminals electrically connected to each electrode pad, i.e., changing the working mode of the electrode group. This allows the propagation path of the electromagnetic wave signal to cover five specific areas to be detected on the subject. This not only enables three-dimensional acquisition of the dynamic changes in the perturbation coefficient of the monitored subject's brain but also ensures a large amount of sample data without changing the electrode positions. Furthermore, the four electrode pads, positioned at specific monitoring points, can comprehensively cover all areas of the monitored subject's brain. Moreover, when an abnormality is detected (e.g., suspected secondary hemorrhage transformation), the abnormality area can be precisely located (e.g., left anterior region, right anterior region, left posterior region, or right posterior region).

[0055] Traditionally, to obtain sufficient data, electromagnetic wave signals need to be applied to the subject from different angles (or positions) to obtain bioelectrical impedance. This requires changing the position of the electrodes on the subject's brain to acquire bioelectrical impedance data at different locations. However, this method of changing electrode positions not only makes the operation cumbersome and reduces data acquisition efficiency, but also makes it impossible to guarantee the accuracy and stability of each electrode placement, thus significantly increasing the amount of interference data introduced. Therefore, this embodiment uses a switching electrode group operating mode to acquire more impedance spectrum data, thereby directly obtaining a more accurate perturbation coefficient dataset.

[0056] In some embodiments, see Figure 5 and Figure 6 The second and fourth electrodes are located in the frontal region (the designated fourth monitoring point) and posterior region (the designated second monitoring point), respectively, with the line connecting them parallel to the sagittal axis. The first and third electrodes are located symmetrically on the left and right sides of the brain, with the line connecting them parallel to the coronal axis. The area between the first and second electrodes is the left posterior region to be monitored; the area between the first and fourth electrodes is the left anterior region; the area between the second and third electrodes is the right posterior region; the area between the third and fourth electrodes is the right anterior region; and the area between the first and third electrodes is the central region.

[0057] See Figure 2 and Figure 3When the monitored person wears the wearable device, the four electrode pads are located at four designated monitoring points on the monitored person's head. The four designated monitoring points are: the fourth monitoring point 4 located on the forehead, the third monitoring point 3 located above the left ear, the first monitoring point 1 located above the right ear, and the second monitoring point 2 located at the back of the head. Among them, the first monitoring point 1 and the third monitoring point 3 form a first propagation path (i.e., the first propagation path is located in the middle region); the first monitoring point 1 and the fourth monitoring point 4 form a second propagation path (i.e., the second propagation path is located in the right front region); the first monitoring point 1 and the second designated monitoring point 2 form a third propagation path (i.e., the third propagation path is located in the right back region); the fourth monitoring point 4 and the third designated monitoring point 3 form a fourth propagation path symmetrical to the second propagation path (i.e., the fourth propagation path is located in the left front region); and the third monitoring point 3 and the second monitoring point form a fifth propagation path symmetrical to the third propagation path (i.e., the fifth propagation path is located in the left back region).

[0058] Specifically, the step of switching the working mode of the electrode in each propagation path includes: acquiring the target attribute control signal triggered by the user, wherein the target attribute control signal includes target attribute information of four electrode pieces, wherein the target attribute information includes: a transmitting end, a receiving end, and a grounding end; and replacing the target attribute end connected to each electrode piece according to the target attribute control signal and the current target attribute information of each electrode piece, wherein the target attribute end includes: a transmitting end, a receiving end, and a grounding end.

[0059] For example, when it is necessary to collect a perturbation coefficient dataset from the left posterior region of the brain, the electromagnetic wave signal is propagated through the propagation path between the first and second electrode pads; when it is necessary to collect a perturbation coefficient dataset from the left anterior region of the brain, the electromagnetic wave signal is propagated through the propagation path between the first and fourth electrode pads; when it is necessary to collect a perturbation coefficient dataset from the right posterior region of the brain, the electromagnetic wave signal is propagated through the propagation path between the second and third electrode pads; when it is necessary to collect a perturbation coefficient dataset from the right anterior region of the brain, the electromagnetic wave signal is propagated through the propagation path between the third and fourth electrode pads; when it is necessary to collect a dataset representing the perturbation coefficients of the whole brain, the electromagnetic wave signal is propagated through the propagation path between the first and third electrode pads.

[0060] Typically, to achieve more precise monitoring, electromagnetic wave signals are applied to the entire brain of the subject to obtain more comprehensive bioimpedance data, leading to a more accurate perturbation coefficient dataset. For example, in addition to acquiring the five propagation paths mentioned above, a perturbation coefficient dataset along the propagation path between the second and fourth electrode pads is also required. However, experiments have shown that the perturbation coefficient dataset between the second and fourth electrode pads has little impact on the accuracy of model training and does not positively affect the accuracy of monitoring results for secondary hemorrhage transformation. Therefore, this application only uses the aforementioned five propagation paths to propagate electromagnetic signals and acquire the subject's perturbation coefficient dataset.

[0061] Furthermore, to increase the data volume, the electromagnetic wave signal can propagate forward or backward along a preset propagation path. Specifically, as mentioned above, the target attribute of the target attribute terminal connected to each of the electrode plates can be changed through a target attribute control signal. For example, the electromagnetic wave signal can propagate forward along the propagation path between the first electrode plate and the second electrode plate; of course, the electromagnetic wave signal can also propagate backward along the propagation path between the second electrode plate and the first electrode plate.

[0062] In some embodiments, the step of replacing the target attribute terminal connected to each electrode piece according to the target attribute control signal and the current target attribute information of each electrode piece specifically includes: obtaining the current target attribute information corresponding to each electrode piece; determining whether the current target attribute information corresponding to each electrode piece is the same as the target attribute information corresponding to the target attribute control signal; if they are different, replacing the target attribute terminal connected to each electrode piece according to the target attribute information in the target attribute control signal; if they are the same, no operation is performed.

[0063] In some embodiments, if the common electrode pad of two adjacent propagation paths is configured as the transmitter, then the other electrode pad on the corresponding propagation path is configured as the receiver. For example, when it is necessary to monitor two symmetrical regions (i.e., the left anterior region and the right anterior region, and the left posterior region and the right posterior region) simultaneously and continuously, the common electrode pad of the two propagation paths in the symmetrical region pair can be configured as the transmitter (e.g., the common electrode pad of the left anterior region and the right anterior region, i.e., the fourth electrode pad located on the forehead, is configured as the transmitter, while the two electrode pads located above the left and right ears are configured as the receivers).

[0064] In some embodiments, secondary hemorrhagic transformation typically occurs in the lesion area; therefore, during continuous monitoring, monitoring can be performed only on the area corresponding to the lesion and its corresponding lateral region. For example, if the original lesion is located in the left anterior region based on the case information, correspondingly, continuous monitoring of the left and right anterior regions is performed after treatment surgery. Of course, in some extreme cases, such as when the lesion covers a large area, or for greater precision, continuous monitoring of multiple regions or the entire brain may be performed simultaneously, i.e., electromagnetic signals are propagated along five propagation pathways or two symmetrical pairs of propagation pathways to obtain the corresponding perturbation coefficient dataset.

[0065] Of course, in some embodiments, different monitoring modes can be pre-configured according to the area that needs to be monitored, and under different monitoring modes, the control circuit automatically configures the target attributes of the electrode sheet.

[0066] In some embodiments, the wearable device may have an approximately ring-shaped structure. Furthermore, to allow the electrode pads to adhere to the skin of the monitored person's brain, the wearable component may be made of a flexible material with a degree of elasticity.

[0067] In some embodiments, the main control device includes:

[0068] The sample library construction module is used to acquire perturbation coefficient datasets [R11; R12; R13; R14; R15] obtained by continuous monitoring of electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths of subjects without hemorrhage after treatment surgery, and perturbation coefficient datasets [R21; R22; R23; R24; R25] obtained by continuous monitoring of electromagnetic wave signals in subjects with secondary hemorrhage after treatment surgery from the first to the fifth propagation paths, in order to construct a training sample library; specifically, this sample library construction module can be directly obtained from the database;

[0069] The model training module is used to learn from training samples using machine learning algorithms to construct mapping relationships between feature parameters and suspected secondary hemorrhage transformation classification labels (and non-secondary hemorrhage transformation classification labels), and to obtain a classifier that can automatically monitor based on the mapping relationship; wherein, the feature parameters include: the maximum value of the absolute difference of perturbation coefficient between symmetrical propagation path pairs obtained from each continuous monitoring, and the difference of the perturbation coefficient fluctuation degree P between symmetrical propagation path pairs; and the fluctuation degree of the whole brain perturbation coefficient obtained from multiple days of continuous monitoring, and / or the perturbation coefficient difference between two adjacent days;

[0070] The data acquisition module is used to periodically or in real-time acquire the perturbation coefficient dataset [R1; R2; R3; R4; R5] of the brain of the monitored person under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths from the wearable monitoring component;

[0071] The monitoring module is used to input the perturbation coefficient dataset [R1; R2; R3; R4; R5] acquired by the data acquisition module into the classifier for automatic monitoring, and to issue early warning prompts based on the monitoring results.

[0072] In some embodiments, the formula for calculating the fluctuation degree P of the disturbance coefficient is as follows:

[0073]

[0074] Wherein, F1 is the maximum amplitude in the upward trend of the time series of disturbance coefficients collected along the current propagation path (i.e., the disturbance coefficient dataset obtained from continuous monitoring; or, the disturbance coefficient dataset obtained from continuous monitoring over multiple days), and F2 is the minimum amplitude in this upward trend; F3 is the maximum amplitude in the downward trend of the time series of disturbance coefficients collected along the current propagation path (i.e., the disturbance coefficient dataset obtained from continuous monitoring; or, the disturbance coefficient dataset obtained from continuous detection over multiple days), and F4 is the minimum amplitude in this downward trend; a and b are the weighting coefficients of the upward and downward trends, respectively (specifically, a and b are empirical values; or, a is the ratio of the area of ​​the upward trend in the disturbance change curve to the area of ​​the upward region and the total area of ​​the downward trend; b is...). The ratio of the area of ​​the downward trend to the total area; t1 is the starting time of the upward trend in the time series of the disturbance coefficients collected by the current propagation path (e.g., the starting moment of the upward trend obtained through continuous monitoring; or, the starting day of the upward trend obtained through continuous monitoring over multiple days), t2 is the duration of the upward trend; t3 is the starting time of the downward trend in the time series of the disturbance coefficients collected by the current propagation path (e.g., the starting moment of the downward trend obtained through continuous monitoring; or, the starting day of the downward trend obtained through continuous monitoring over multiple days), t4 is the duration of the downward trend; if there is no fluctuation or almost no fluctuation in the time series of the disturbance coefficients collected by the current propagation path (e.g., the maximum amplitude is lower than the preset threshold 5), then P = 0. Of course, if the disturbance coefficients continuously monitored over multiple days not only have two trends, but each trend also exists at least twice, P is the sum of the products of the fluctuation degree of each trend and its weight coefficient, that is: Among them, F1 i F2 i Let a be the maximum and minimum amplitude values ​​in the i-th upward trend out of n upward trends. i F3 is the weighting coefficient for the i-th upward trend. j F4 j b represents the maximum and minimum amplitude values ​​in the j-th downward trend out of m downward trends; j is the weighting coefficient for the j-th downward trend.

[0075] Typically, secondary hemorrhage transformation lasts for a certain period and exhibits a certain fluctuation pattern. Therefore, to avoid misjudgment due to sudden increases and / or decreases in the perturbation coefficient caused by misoperation, equipment malfunction, or other medical conditions (e.g., judging solely based on the difference between the current and previous perturbation coefficients in the same region, or solely based on the difference in the current perturbation coefficients in symmetrical regions), this application treats the time series of continuously monitored perturbation coefficients as a whole when training the model to construct its fluctuation pattern. For example, for subjects monitored within 36 hours post-surgery, the mapping relationship between the fluctuation rate difference between symmetrical regions and the secondary hemorrhage transformation classification label is combined; similarly, for subjects monitored after 36 hours post-surgery, the mapping relationship between the fluctuation degree of the whole-brain perturbation coefficients obtained from continuous monitoring and the secondary hemorrhage transformation classification label is combined.

[0076] Of course, to further reduce the misclassification rate, the variance of the perturbation coefficient was also introduced. For example, for the period within 36 hours after surgery, a mapping relationship was constructed between the maximum difference, the difference in fluctuation, and the variance of the perturbation coefficients between symmetrical regions and the classification label of secondary hemorrhage transformation, thus obtaining the first classifier; and for the period after 36 hours after surgery, a mapping relationship was constructed between the difference, the fluctuation, and the variance of the whole-brain perturbation coefficients between two consecutive days and the classification label of secondary hemorrhage transformation, thus obtaining the second classifier. Variance accurately describes the fluctuation of the disturbance coefficient. Therefore, introducing variance can improve the accuracy of the model. However, it also increases the amount of data processing in the early stages of the system, thereby increasing system energy consumption. The challenge is to find a balance between accuracy and energy consumption. In scenarios where system equipment performance is relatively low and accuracy requirements are relatively low, a higher-precision monitoring mode can be set through the main control module to prevent the introduction of variance during subsequent automatic monitoring. Conversely, in applications where system equipment performance is relatively high and accuracy requirements are relatively high, a high-precision monitoring mode can be set through the main control module to introduce variance during subsequent automatic monitoring. This allows for more flexible selection of appropriate methods for automatic monitoring based on the actual application scenario and equipment performance, offering high flexibility and wide applicability.

[0077] Secondary hemorrhage transformation typically occurs within 36 hours post-operatively and lasts for 30-100 minutes. However, in practice, it has been observed that it can occur even after 36 hours, and the development process of secondary hemorrhage transformation occurring after 36 hours is relatively longer compared to that occurring within 36 hours post-operatively. Therefore, the perturbation coefficient dataset in the training sample library includes perturbation coefficient datasets obtained from each continuous monitoring session within 36 hours post-operatively, as well as perturbation coefficient datasets obtained from multiple continuous monitoring sessions per day after 36 hours post-operatively. Two classifiers were trained separately for secondary hemorrhage transformation within and after 36 hours.

[0078] For example, for secondary hemorrhage transformation within 36 hours, the following mapping relationship is constructed through model training: the maximum value of the absolute difference of the perturbation coefficient between symmetrical propagation path pairs, and the difference of the fluctuation degree P of the perturbation coefficient between symmetrical propagation path pairs, which are mapped to the classification labels of suspected secondary hemorrhage transformation and non-secondary hemorrhage transformation, thereby obtaining a first classifier that can be automatically monitored based on this mapping relationship.

[0079] For secondary hemorrhage transformation after 36 hours, the model training constructs a mapping relationship between the changing trend of the whole-brain perturbation coefficient (e.g., the perturbation coefficient between the first and third designated monitoring points) obtained from continuous monitoring over multiple days, and / or the difference in perturbation coefficients between adjacent two days (e.g., the difference between the mean perturbation coefficients obtained from continuous monitoring over two days) and the classification labels of suspected secondary hemorrhage transformation and non-secondary hemorrhage transformation, thereby obtaining a second classifier that can automatically monitor based on this mapping relationship.

[0080] In some embodiments, the perturbation coefficient between the first designated monitoring point and the third designated monitoring point is used to characterize the whole-brain perturbation coefficient; of course, the average of the perturbation coefficients of two symmetrical regions can also be used to characterize the whole-brain perturbation coefficient.

[0081] In some embodiments, the bioelectrical impedance measurement device further includes an electromagnetic wave generator for sending electromagnetic waves of a corresponding preset frequency band to the electrode sheet. Specifically, the preset frequency band includes an ultra-low frequency band and a low frequency band, and both the ultra-low frequency band and the low frequency band are divided into at least two sub-frequency bands; the multiple sub-frequency bands obtained by dividing the preset frequency band include: a first sub-frequency band with a frequency range of 50Hz-100Hz, a second sub-frequency band with a frequency range of 100Hz-500Hz, a third sub-frequency band with a frequency range of 500Hz-1kHz, a fourth sub-frequency band with a frequency range of 1kHz-10kHz, a fifth sub-frequency band with a frequency range of 10kHz-100kHz, and a sixth sub-frequency band with a frequency range of 100kHz-300kHz; wherein, the perturbation coefficient dataset corresponding to the ultra-low frequency band of 50Hz-500Hz accounts for the largest proportion of the total training samples.

[0082] In some embodiments, the main control device further includes: a medical condition acquisition module, which communicates with the medical information system of a medical institution to acquire the medical condition of the monitored person, the medical condition including: treatment and surgery time;

[0083] The control module is used to identify, based on the case situation, that when the time interval between the day the monitored person is monitored and the completion of the treatment surgery is less than 36 hours, set the continuous monitoring duration to be greater than 30-100 minutes, and the monitoring frequency is 8-14 times per day.

[0084] In some embodiments, the control module is further configured to match a corresponding classifier based on the case details. For example, if it is identified that the time interval between the day the monitored person is monitored and the completion of the treatment surgery is less than 36 hours, the monitoring module is controlled to perform automatic monitoring based on the first classifier; if it is identified that the time interval between the day the monitored person is monitored and the completion of the treatment surgery is greater than 36 hours, the monitoring module is controlled to perform automatic monitoring based on the second classifier.

[0085] Of course, if the time between treatment and surgery for the monitored person exceeds 36 hours, the frequency of daily monitoring can be reduced. Specifically, for example, monitoring can be performed 3-8 times per day.

[0086] In some embodiments, if the time between the monitored patient's surgical procedure and the procedure exceeds 36 hours, although the probability of secondary hemorrhage transformation decreases, the development process may be slower. Therefore, identification based solely on the perturbation coefficient dataset monitored each time may result in missed detections. Thus, it is necessary to identify perturbation coefficient datasets monitored over multiple days. For example, the average of the perturbation coefficient datasets monitored each day can be calculated to obtain a perturbation coefficient dataset monitored over multiple days. Preferably, if the differences between the perturbation coefficient datasets monitored multiple times on a given day are small (e.g., the difference between the averages of multiple consecutive monitoring data points on a given day is less than a preset threshold, such as 2-3), the largest average can be used as the average perturbation coefficient for that day.

[0087] Because multiple continuous monitoring sessions are conducted daily, the amount of data obtained is enormous. Therefore, in order to reduce the system's data processing load (especially the processing of invalid data, such as monitoring three times a day, but each monitoring result indicating that the monitored person has not experienced secondary bleeding, which can be considered an invalid data processing process), a multi-day perturbation coefficient dataset of the monitored person (or the average perturbation coefficient obtained from daily monitoring) is generated.

[0088] In some embodiments, the control module is further configured to automatically recalculate the monitoring time from the start of the upward or downward trend of the corresponding side's perturbation coefficient when the absolute difference between the perturbation coefficient of any propagation path in any symmetrical region pair and the propagation path of its symmetrical side (i.e., the right front region) is greater than a preset absolute difference (e.g., 5), or when the absolute difference between the perturbation coefficient of the propagation path in the middle region at the current time and the previous time is greater than a preset absolute difference (e.g., 5), until the upward or downward trend ends.

[0089] Furthermore, the control module is also used to automatically recalculate the monitoring time from the start time of the first change trend of the disturbance coefficient when the upward or downward trend of the above-mentioned disturbance coefficient has ended, and when a downward or upward trend occurs again within a preset time interval (e.g., 1 min-10 min), until the second change trend ends.

[0090] Furthermore, the main control device also includes a human-machine interface module for displaying monitoring and detection results. It is also used by operators to set monitoring parameters, such as continuous monitoring duration, monitoring cycle (or frequency), and monitoring mode.

[0091] Furthermore, the main control device also includes a data augmentation module, used to augment the perturbation coefficients in the training sample library using a conditional generative adversarial network.

[0092] Example 2: See Figure 5 The above is a flowchart of an exemplary embodiment of the present invention regarding a method for intelligent monitoring of bleeding conversion after thrombolysis and thrombectomy for cerebral infarction. Specifically, the method includes the following steps:

[0093] S101, acquire the normal perturbation coefficient dataset obtained by continuous monitoring of the brain of subjects without hemorrhage transformation in the preset frequency band under electromagnetic wave signals of multiple target frequencies and five preset propagation paths, and the hemorrhage perturbation coefficient dataset obtained by continuous monitoring of subjects with hemorrhage transformation after treatment surgery (i.e., subjects with secondary hemorrhage) under electromagnetic wave signals of multiple target frequencies and five preset propagation paths in the preset frequency band, in order to construct a training sample library.

[0094] In some embodiments, electromagnetic wave signals can be applied to five specific propagation pathways in the brains of normal subjects and hemorrhage-transformed subjects using various devices or equipment for measuring (cranial) bioelectrical impedance, thereby directly obtaining the corresponding perturbation coefficient data from the device or equipment for measuring bioelectrical impedance.

[0095] In some embodiments, participate Figure 2 and Figure 3The five specific transmission paths mentioned above are located in the following areas: the middle region between the first and third monitoring points (i.e., the middle region between the left and right sides of the brain); the left anterior region between the first and fourth monitoring points (i.e., the left anterior region between the left ear and the forehead); the left posterior region between the first and second monitoring points (i.e., the left posterior region between the left ear and the back of the brain); the right anterior region between the fourth and third monitoring points (i.e., the right anterior region between the right ear and the forehead); and the right posterior region between the third and second monitoring points (i.e., the right posterior region between the right ear and the back of the brain). The left anterior region is symmetrical to the right anterior region, and the left posterior region is symmetrical to the right posterior region. Correspondingly, the transmission paths in the left anterior region are symmetrical to the transmission paths in the right anterior region, and the transmission paths in the left posterior region are symmetrical to the transmission paths in the right posterior region.

[0096] In some embodiments, the preset frequency band ranges from 50Hz to 300kHz. This involves acquiring the impedance spectrum of electromagnetic wave signals at various target frequencies across the entire frequency band, thereby obtaining their corresponding perturbation coefficients.

[0097] In some embodiments, the preset frequency band is divided into multiple sub-frequency bands. Accordingly, the electromagnetic wave signals (e.g., alternating current electric field) of multiple sub-frequency bands are periodically applied to the subject through the device or equipment for measuring bioelectrical impedance described above, thereby obtaining the impedance corresponding to each sub-frequency band, and then obtaining the impedance spectrum and its perturbation coefficient corresponding to each sub-frequency band.

[0098] In some embodiments, to obtain more disturbance coefficient data for the ultra-low frequency band, the entire frequency band is divided into multiple sub-bands, wherein the number of sub-bands for the ultra-low frequency band is greater than the number of sub-bands corresponding to the mid-frequency band and the high-frequency band. For example, the ultra-low frequency band is divided into at least two sub-bands. Further, the low-frequency band can also be divided into at least two sub-bands. For example, the entire frequency band (preset band) 50Hz-300kHz is divided into six sub-bands: 1) First sub-band (ultra-low frequency band): 50Hz~100Hz (greater than or equal to 50Hz and less than 100Hz); 2) Second sub-band (ultra-low frequency band): 100Hz-500Hz (greater than or equal to 100Hz and less than 500Hz); 3) Third sub-band (low-frequency band): 500Hz-1000Hz (greater than or equal to 500Hz and less than 1000Hz); 4) Fourth sub-band (low-frequency band): 500Hz-1000Hz (greater than or equal to 500Hz and less than 1000Hz); 5) Fourth sub-band (low-frequency band): 500Hz-1000Hz (greater than or equal to 500Hz and less than 1000Hz); 6) Fifth sub-band (low-frequency band): 50Hz-1000Hz (greater than or equal to 500Hz and less than 1000Hz); 7) Sixth sub-band (low-frequency band): 50Hz-1000Hz (greater than or equal to 500Hz and less than 1000Hz); 8) Sixth sub-band (low-frequency band): 50Hz-1000Hz (greater than or equal to 500Hz and less than 100 The frequency bands are divided into six sub-bands: 1 kHz to 10 kHz (greater than or equal to 1 kHz and less than 10 kHz); 5) Fifth sub-band (mid-frequency band): 10 kHz to 100 kHz (greater than or equal to 10 kHz and less than 100 kHz); 6) Sixth sub-band (high-frequency band): 100 kHz to 300 kHz (greater than or equal to 100 kHz and less than or equal to 300 kHz). The first, second, third, fourth, fifth, and sixth sub-bands are applied to the subject periodically. In some embodiments, each sub-band corresponds to an initial frequency, and within each sub-band, the corresponding initial frequency is used as the target frequency. Then, the frequency is adjusted according to a preset frequency adjustment threshold to obtain the corresponding target frequency. Furthermore, the preset frequency adjustment thresholds corresponding to each of the multiple sub-bands gradually increase in order of increasing frequency. That is, the preset frequency adjustment threshold of the low-frequency sub-band is less than the preset frequency adjustment threshold of the mid-frequency band, and the preset frequency adjustment threshold of the mid-frequency band is less than the preset frequency adjustment threshold of the high-frequency band. Preferably, the preset frequency adjustment thresholds among the multiple sub-bands corresponding to the low-frequency band are also gradually increased in order from low to high, so that the lower the frequency of the sub-band, the more impedance spectrum data is obtained, and the more accurate the disturbance coefficient change curve is obtained based on the impedance spectrum data.For example, for the first sub-band: 50Hz~100Hz (ultra-low frequency band), 50Hz is set as the corresponding initial frequency, and the corresponding frequency adjustment threshold is 5Hz; for the second sub-band: 100Hz-500Hz (ultra-low frequency band), 100Hz is set as the corresponding initial frequency, and the corresponding frequency adjustment threshold is 55Hz; for the third sub-band: 500Hz-1000Hz (low frequency band), 500Hz is set as the corresponding initial frequency, and the corresponding frequency adjustment threshold is 100Hz; for the fourth sub-band: 1kHz~10kHz (low frequency band), 1kHz is set as the corresponding initial frequency, and the corresponding frequency adjustment threshold is 1kHz; for the fifth sub-band: 10kHz~100kHz (mid frequency band), 10kHz is set as the corresponding initial frequency, and the corresponding frequency adjustment threshold is 20kHz; for the sixth sub-band: 100kHz~300kHz (high frequency band), 100kHz is set as the corresponding initial frequency, and the corresponding frequency adjustment threshold is 100kHz. Of course, in other embodiments, within a sub-band, when the frequency adjustment value (i.e., the target frequency) obtained after multiple adjustments based on a preset frequency adjustment threshold starting from the initial frequency does not belong to the current sub-band, this frequency adjustment value is used as the initial frequency of the corresponding sub-band. Alternatively, the larger of the frequency adjustment value and the original initial frequency of the corresponding sub-band is used as the new initial frequency of the sub-band. For example, for the second sub-band: starting from 100Hz (initial frequency), frequency adjustment is performed at preset time intervals and a preset frequency adjustment threshold of 55Hz. Then: the target frequency after the first adjustment is 100 + 55 = 155Hz, and the target frequency after the second adjustment is 155 + 55 = 155Hz. After the 7th adjustment, the frequency is 485 + 55 = 540 Hz, which is greater than 500 Hz. This exceeds the second sub-band's 100 Hz to 500 Hz range and corresponds to the third sub-band's 500 Hz to 1 kHz range. It is also greater than the original target frequency of the third sub-band, 500 Hz. Therefore, this frequency adjustment value (i.e., the target frequency) of 540 Hz is used as the new initial frequency for the third sub-band. Starting from this new initial frequency, frequency adjustment is performed according to a preset time interval and a preset frequency adjustment threshold of 500 Hz, and so on, until the last sub-band completes frequency adjustment (i.e., one cycle is completed). In some embodiments, the percentage of sample data corresponding to the ultra-low frequency band in the sample training library is the highest. For example, the percentage of the total sample data corresponding to the perturbation coefficients of the first to second sub-bands is higher than the percentage of the perturbation coefficient data of any one of the third to sixth sub-bands. Most preferably, the perturbation coefficient data corresponding to the ultra-low frequency band accounts for at least 50% of the total sample data.

[0099] S102 uses ACGAN to augment the training samples.

[0100] In this embodiment, ACGAN is used for data augmentation, which further increases the amount of training sample data.

[0101] Of course, in other embodiments, other conditional generative adversarial networks, such as GANs, can be used to augment the training samples.

[0102] S103, using machine learning algorithms to learn the training samples amplified in step S102, in order to construct a mapping relationship between feature parameters and suspected secondary hemorrhage transformation classification labels, and obtain a classifier that can be identified based on this mapping relationship.

[0103] In some embodiments, the characteristic parameter includes the absolute difference in perturbation coefficients between pairs of symmetrical regions and the difference in the degree of fluctuation P of perturbation coefficients between pairs of symmetrical regions.

[0104] In some embodiments,

[0105] In some embodiments, the formula for calculating the fluctuation degree P of the disturbance coefficient is as follows: Where F1 is the maximum amplitude in the upward trend, F2 is the minimum amplitude in the upward trend, F3 is the maximum amplitude in the downward trend, and F4 is the minimum amplitude in the downward trend; a and b are the weighting coefficients for the upward and downward trends, respectively; t1 is the start time of the upward trend, t2 is the duration of the upward trend; t3 is the start time of the downward trend, and t4 is the duration of the downward trend; if the disturbance coefficient does not fluctuate, then P = 0.

[0106] In other embodiments, the characteristic parameter also includes the variance of the perturbation coefficient in the upward and / or downward trends.

[0107] In other embodiments, the characteristic parameter further includes: the variance of the upward and / or downward trend of the disturbance coefficient in the intermediate region, or the degree of fluctuation of the disturbance coefficient in the intermediate region over a continuous monitoring period.

[0108] In some embodiments, the machine learning algorithm includes a Bayesian neural network model, a convolutional neural network model, or a random forest, wherein the ratio of the training set to the test set in the training sample library is 8:5; and wherein the training set is divided into training data and validation data in a 7:3 ratio.

[0109] In some embodiments, the classifier is tested using a test set.

[0110] S104, acquires in real time the disturbance coefficient dataset obtained by continuous monitoring of electromagnetic wave signals propagating at at least one target frequency and through five preset propagation paths after the patient undergoes treatment surgery.

[0111] In some embodiments, as described above, an electromagnetic wave signal, such as an alternating electric field, of at least one target frequency band and different propagation paths is applied to the subject to be tested by a bioelectrical impedance measuring device or apparatus, thereby directly obtaining the corresponding impedance to be tested from the bioelectrical impedance measuring device or apparatus, and obtaining the corresponding perturbation coefficient dataset.

[0112] Preferably, a dataset of disturbance coefficients of electromagnetic wave signals of the subject under test at multiple target frequencies in the full frequency band (or in each sub-band) and along five predetermined propagation paths is obtained.

[0113] S105, Input the disturbance coefficient to be tested obtained in step S104 into the classifier for automatic detection.

[0114] In some embodiments, when the classifier performs automatic detection based on the perturbation coefficient to be tested, the final result is the probability of classifying the subject to be tested as having secondary hemorrhage transformation or non-secondary hemorrhage transformation. When the probability of being classified as having secondary hemorrhage transformation is greater than a preset probability threshold (e.g., greater than 85%), the subject to be tested is determined to be a patient with secondary hemorrhage transformation.

[0115] Accordingly, when the method of this embodiment uses a machine learning algorithm to learn from the training samples in the training sample library (i.e., execute step S103), it constructs a mapping relationship between the feature parameters and the classification labels of secondary hemorrhage transformation and non-secondary hemorrhage transformation, respectively, and obtains a classifier that can be used to monitor secondary hemorrhage transformation based on the mapping relationship. Then, the classifier is used to automatically monitor the subjects to be tested, i.e., execute steps S104-S105 in the above embodiment 1.

[0116] In some embodiments, the classifier is tested using a test set, and the test results are shown in Table 1 below: The classifier has a precision of 98% and a recall of 87% for suspected secondary hemorrhage transformation; and a precision of 76% and a recall of 97% for non-secondary hemorrhage transformation.

[0117] Table 1. Classifier Performance on Test Set

[0118] Of course, to increase the reliability of machine learning, the training samples can be augmented using conditional generative adversarial networks.

[0119] In normal individuals, the perturbation coefficients between the left and right hemispheres are not significantly different, with the absolute value of the bilateral perturbation coefficients being less than 0.03. However, when no hemorrhagic transformation occurs post-surgery, the perturbation coefficient curve of the subject remains almost unchanged, such as... Figure 6As shown, therefore, the difference in perturbation coefficients between the left and right hemispheres remains unchanged; however, when postoperative hemorrhage transformation occurs, the perturbation coefficient of the area experiencing hemorrhage transformation will show an increasing or decreasing trend, such as... Figure 7 or Figure 8 Even rises and falls exist, such as Figure 9 This results in a difference in the perturbation coefficient between this region and the other side of the region that is symmetrical to it. Naturally, the difference in the perturbation coefficient between the two will also show an upward or downward trend.

[0120] However, using only the difference in perturbation coefficients between symmetrical regions to determine the transformation of bleeding may not be accurate enough. Therefore, this embodiment uses machine learning to obtain the mapping relationship between the various feature parameters of the perturbation coefficient and the classification label to obtain a classifier. Then, the classifier is used to perform more accurate monitoring of secondary bleeding transformation based on the various feature parameters of the perturbation coefficient.

[0121] After secondary hemorrhagic transformation occurs, the perturbation coefficient change curve of the corresponding area of ​​the patient will continue to rise or fall for a period of time, or rise first and then fall, and will soon fluctuate. Correspondingly, the difference in perturbation coefficient between symmetrical areas will also fluctuate. If we judge based solely on the difference in perturbation coefficient between symmetrical areas at the current moment, the accuracy is too low. Moreover, the fluctuation rate of the perturbation coefficient change curve reflects the development process of hemorrhagic transformation, which can be monitored more accurately.

[0122] To demonstrate the advanced nature and effectiveness of the method of this invention, the applicant conducted clinical trials, and the results are as follows:

[0123] Source of participants: Cases collected at Southwest Hospital between March and August 2022 (see [link to study]). Figure 6 (Introduction of some cases), among which, 3 cases developed hemorrhagic transformation after treatment surgery, and all of them were continuously monitored by the monitoring equipment applied for in this application. After the warning prompts were issued, the doctor ordered the corresponding CT / MRI examination.

[0124] See Figure 10a and Figure 10bCase 1 – The patient underwent surgery on August 7, 2022. After the surgery, the method of this invention was used to continuously monitor the subject postoperatively. Based on the whole-brain perturbation coefficient dataset obtained from the continuous monitoring from August 11 to 13 (compared to August 11, the perturbation coefficient on August 12 increased by 2 points but less than 5 points, while on August 13 the patient's perturbation coefficient increased by 7 points, that is, it started to rise on August 12, and the fluctuation level at this time was (97-90) / 1=7), the system automatically monitored the patient. Based on the monitoring results, the system indicated suspicious bleeding. Therefore, the doctor ordered a head CT scan on August 13 for follow-up examination, which found that the patient's intracranial hemorrhage had increased, but the amount was small, and the corresponding treatment continued.

[0125] Continuous monitoring was conducted after treatment, and the perturbation coefficient obtained on the 14th was input into the corresponding classifier for automatic monitoring (compared to the 13th, the patient's perturbation coefficient increased by 7 points on the 14th, that is, the above upward trend continued to rise, but the degree of fluctuation remained unchanged, with the degree of fluctuation being: (104-90) / 2=7). The monitoring results indicated that the bleeding had further worsened. Therefore, the doctor ordered a head CT scan on August 14, 2022 for follow-up examination, which revealed that the intracranial hemorrhage had increased compared to the 13th.

[0126] Continuous monitoring was conducted, and the perturbation coefficient obtained from the monitoring on the 15th was input into the corresponding classifier for automatic monitoring (the patient's perturbation coefficient only increased by 2 points on the 15th. Although the trend continued to rise, the degree of fluctuation decreased to (104-88) / 5=3.6), that is, no abnormality was detected.

[0127] See Figure 11a and Figure 11b Case 2 – After surgery on August 2, 2022, the patient was continuously monitored using the monitoring device of this invention. The system automatically monitored the patient based on the whole-brain perturbation coefficient dataset obtained from continuous monitoring from August 14 to 16 (compared to August 14, the perturbation coefficient of the right anterior region increased by 5 points on August 15, indicating an upward trend starting on August 14, with a fluctuation level of 5; while on August 16, the perturbation coefficient of the patient's right anterior region continued to increase by 10 points, with a fluctuation level of 15 / 2 = 7.5). The system indicated suspected bleeding based on the monitoring results and advised close observation. Therefore, continuous monitoring was continued.

[0128] The perturbation coefficient dataset obtained on the 17th was input into the corresponding classifier for automatic monitoring (the patient's perturbation coefficient increased by 9 points on the 17th, meaning that the above upward trend will continue to rise, and the degree of fluctuation also shows an upward trend: (78-59) / 3=8). The system indicated that the bleeding had worsened further based on the monitoring results. Therefore, the doctor ordered a head MRI examination on August 17, 2022 for follow-up examination, which revealed a small amount of bleeding in the surgical area, local edema, and air accumulation in the right frontal region.

[0129] Continue monitoring and input the perturbation coefficients obtained from the monitoring on the 18th and 19th into the corresponding classifier for automatic monitoring (the patient's perturbation coefficient only increased by 2 points on the 18th, while the perturbation coefficient tended to stabilize on the 19th, and the degree of fluctuation gradually decreased). The system did not detect any abnormalities.

[0130] See Figure 12a and Figure 12b Case 3 – Following surgery on August 11, 2022, the patient was continuously monitored using the intelligent monitoring device of this invention. Based on the whole-brain perturbation coefficient dataset obtained from continuous monitoring from the 13th to the 16th (compared to the 13th, the perturbation coefficient increased by 2 points on the 14th, but less than 5 points; compared to the 14th, the patient's perturbation coefficient increased by 6 points on the 15th, indicating an upward trend starting on the 14th, with a fluctuation level of 6 / 1 = 6; compared to the 15th, the perturbation coefficient decreased by 2 points on the 16th, with the fluctuation level decreasing to 6 × a + 4 × b = 4.75, where a is the area ratio of the upward trend and b is the area ratio of the downward trend); if… Figure 12b The area of ​​the upward trend (the area obtained by using the 15-day disturbance coefficient as the baseline) is considered as a triangle (the area of ​​the triangle is 3), and the area of ​​the downward trend is considered as a trapezoid (the area of ​​the trapezoid is 5). Then a is 3 / 8 and b is 5 / 8. The system did not indicate any abnormalities (or secondary hemorrhage transformation did not occur).

[0131] Continue monitoring and input the perturbation coefficient obtained on the 17th into the corresponding classifier for automatic monitoring (compared to the 16th, the patient's perturbation coefficient increased by 12 points on the 17th, that is, the perturbation coefficient continued to rise, and the fluctuation level also turned into an upward trend: (6×a1+4×b+12×a2=8.375, where a is the area ratio of all upward trends, and b is the area ratio of all downward trends; if... Figure 12b The area of ​​the first upward trend is considered as a triangle (the area of ​​the triangle is 3), the area of ​​the downward trend is considered as a trapezoid (the area of ​​the trapezoid is 5), and the area of ​​the second upward trend is considered as a trapezoid (its area is 8). Then a1 is 3 / 16, b is 5 / 16, and a2 is 8 / 16. Automatic monitoring is performed, and the system indicates bleeding based on the monitoring results. Therefore, the doctor ordered a head MRI on August 17, 2022, which detected contusions and lacerations of the bilateral frontal, parietal, left temporal lobes and bilateral cerebellar hemispheres, subacute subdural hematoma in the left temporal and occipital regions, and subarachnoid hemorrhage.

[0132] Continued monitoring was conducted, and the perturbation coefficient obtained on the 18th was input into the corresponding classifier for automatic monitoring (the patient's perturbation coefficient remained unchanged on the 18th compared to the 17th). The system did not indicate any abnormalities based on the monitoring results.

[0133] Therefore, this invention not only enables preliminary identification of secondary hemorrhage transformation, achieving a breakthrough in early monitoring and continuous surveillance of secondary hemorrhage transformation, but also monitors the dynamic development process of secondary hemorrhage transformation and provides early warnings. Of course, the intelligent monitoring device and method described above can also be used for clinical monitoring of neurological diseases.

[0134] A fourth aspect of the present invention provides an electronic device, including a memory 502, a processor 501, and a computer program stored in the memory 502 and executable on the processor 501, wherein the processor 501 executes the program to implement the steps of the method described above. For ease of explanation, only the parts related to the embodiments of this specification are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of this specification. This electronic device can be any electronic device, including PCs, cloud servers, and even mobile phones, tablets, PDAs (Personal Digital Assistants), POS (Point of Sales) terminals, in-vehicle computers, desktop computers, etc.

[0135] Specifically, Figure 4 The illustrated block diagram of an electronic device configuration related to the technical solutions provided in the embodiments of this specification shows that bus 500 may include any number of interconnected buses and bridges, linking various circuits including one or more processors represented by processor 501 and memory represented by memory 502. Bus 500 may also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Communication interface 503 provides an interface between bus 500 and receiver and / or transmitter 504, which may be separate independent receivers or transmitters or a single element such as a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 501 is responsible for managing bus 500 and general processing, while memory 502 may be used to store data used by processor 501 during operation.

[0136] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a computer-readable storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, or network device, etc.) to execute the methods described above according to the embodiments of this disclosure.

[0137] The computer-readable storage medium may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0138] Program code for performing the operations of this disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0139] The aforementioned computer-readable medium carries one or more programs, which, when executed by a device, cause the computer-readable medium to perform the steps in the above embodiments. Because the occurrence of secondary hemorrhage transformation has a certain degree of randomness (although it mostly occurs some time after surgery, the specific timeframe is unclear; and it can even occur 36 hours after surgery).

[0140] Those skilled in the art will understand that the above modules can be distributed in the device as described in the embodiments, or they can be modified accordingly and placed in one or more devices that are unique to this embodiment. The modules in the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0141] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a computer terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0142] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0143] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A smart monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation after cerebral infarction, characterized in that, include: Wearable monitoring components are used to continuously monitor a person under monitoring in order to obtain a dataset of disturbance coefficients of the person under monitoring under electromagnetic wave signals in a preset frequency band and with five preset propagation paths. The person under supervision was a patient who had undergone thrombolysis and thrombectomy for cerebral infarction. The wearable monitoring component includes: a wearable component for wearing on the head of the monitored person, and four electrode pads disposed on the wearable component; when the monitored person wears the wearable component, the four electrode pads are respectively located at four designated monitoring points on the monitored person's head, the four designated monitoring points being: a fourth monitoring point located on the forehead, a third monitoring point located above the left ear, a first monitoring point located above the right ear, and a second monitoring point located at the back of the head, with a first propagation path formed between the first monitoring point and the third monitoring point; a second propagation path formed between the first monitoring point and the fourth monitoring point; a third propagation path formed between the first monitoring point and the second monitoring point; a fourth propagation path symmetrical to the second propagation path formed between the fourth monitoring point and the third monitoring point; and a fifth propagation path symmetrical to the third propagation path formed between the third monitoring point and the second monitoring point. A main control device communicates with the wearable monitoring component to automatically monitor the data set of disturbance coefficients obtained through continuous monitoring by the wearable monitoring component, and provides early warnings based on the monitoring results; the main control device includes: The sample library construction module is used to obtain the perturbation coefficient datasets obtained by continuously monitoring the electromagnetic wave signals of the first, second, third, fourth, and fifth propagation paths of subjects without hemorrhage after treatment surgery, as well as the perturbation coefficient datasets obtained by continuously monitoring the electromagnetic wave signals of the first, second, third, fourth, and fifth propagation paths of subjects with secondary hemorrhage after treatment surgery, in order to construct a training sample library. The model training module is used to learn from the training samples in the training sample library using machine learning algorithms to construct mapping relationships between feature parameters and secondary hemorrhage transformation classification labels, and to obtain a classifier that can automatically monitor based on the mapping relationship; wherein, the feature parameters include: the maximum value of the absolute difference of perturbation coefficients between symmetrical propagation path pairs, the difference in the degree of fluctuation of perturbation coefficients between symmetrical propagation path pairs; and the degree of fluctuation of whole-brain perturbation coefficients obtained from continuous monitoring over multiple days, and the difference in perturbation coefficients between two adjacent days; The data acquisition module is used to periodically or in real time acquire the perturbation coefficient dataset of the brain of the monitored person under the electromagnetic wave signals of the first, second, third, fourth and fifth propagation paths from the wearable monitoring component; The monitoring module is used to input the perturbation coefficient dataset acquired by the data acquisition module into the classifier for automatic monitoring, and to provide early warning prompts based on the monitoring results; The formula for calculating the fluctuation degree P of the disturbance coefficient is as follows: ， in, This represents the largest value during an upward trend. This represents the minimum amplitude within an upward trend. This represents the largest value in a downward trend. This represents the minimum amplitude during a downward trend. and These are the weighting coefficients for upward and downward trends, respectively. The starting time of the upward trend. The duration of the upward trend; The starting time of the downward trend. The duration of the downward trend.

2. The intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation after cerebral infarction according to claim 1, characterized in that, When the disturbance coefficient monitored over multiple days exhibits two changing trends, and either trend occurs at least once, , in, , Let $\mathbf{i}$ be the maximum and minimum magnitude values ​​in the $i$-th upward trend out of $n$ upward trends. The weighting coefficient for the i-th upward trend; These represent the maximum and minimum amplitude values ​​in the j-th downward trend out of m downward trends; is the weighting coefficient for the j-th downward trend.

3. The intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation after cerebral infarction according to claim 1, characterized in that, The model training module is specifically used to construct a mapping relationship between the maximum value of the absolute difference of the perturbation coefficient between symmetric propagation path pairs and the difference of the perturbation coefficient fluctuation degree P, based on the perturbation coefficient dataset within 36 hours after the treatment surgery in the training sample, and to the classification label of secondary bleeding, so as to obtain a first classifier for automatic monitoring of the monitored person within 36 hours after the treatment surgery based on the mapping relationship. And a second classifier is used to construct the fluctuation degree of the whole brain perturbation coefficient obtained by continuous monitoring over multiple days based on the perturbation coefficient dataset 36 hours after treatment surgery in the training sample, and the mapping relationship between the perturbation coefficient difference between two adjacent days and the secondary hemorrhage transformation classification label, so as to obtain the second classifier for automatic monitoring of the monitored person 36 hours after treatment surgery based on the mapping relationship.

4. The intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation after cerebral infarction according to claim 3, characterized in that, The main control device also includes: The medical condition acquisition module communicates with the medical information system of the medical institution to acquire the medical records of the monitored person, including the treatment and surgery time. The control module is configured to, based on the treatment surgery time, identify when the time interval between the current time and the completion of the treatment surgery is less than 36 hours, acquire a pre-set continuous monitoring duration and the frequency of continuous monitoring per day, and control the monitoring module to perform automatic monitoring using the first classifier; or, when based on the treatment surgery time, identify when the time interval between the current time and the completion of the treatment surgery is more than 36 hours, reduce the frequency of continuous monitoring per day, and control the monitoring module to perform automatic monitoring using the second classifier.

5. The intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation after cerebral infarction according to claim 1, characterized in that, The characteristic parameters also include the variance of the disturbance coefficient in the upward trend and / or the downward trend.

6. The intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation after cerebral infarction according to any one of claims 1 to 5, characterized in that, The preset frequency band includes an ultra-low frequency band and a low frequency band, and both the ultra-low frequency band and the low frequency band are divided into at least two sub-frequency bands; the multiple sub-frequency bands obtained by dividing the preset frequency band include: a first sub-frequency band with a frequency range of 50Hz-100Hz, a second sub-frequency band with a frequency range of 100Hz-500Hz, a third sub-frequency band with a frequency range of 500Hz-1kHz, a fourth sub-frequency band with a frequency range of 1kHz-10kHz, a fifth sub-frequency band with a frequency range of 10kHz-100kHz, and a sixth sub-frequency band with a frequency range of 100kHz-300kHz; wherein, the ultra-low frequency band is 50Hz. The perturbation coefficient dataset corresponding to 500Hz accounts for the largest proportion of the total training samples.

7. The intelligent monitoring device for post-thrombolysis and thrombectomy hemorrhage transformation after cerebral infarction according to claim 1, characterized in that, The main control device further includes: a human-computer interaction module for displaying monitoring results; and / or a database for pre-storing a dataset of disturbance coefficients obtained by continuous monitoring of several subjects without hemorrhage after treatment surgery under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths, and a dataset of disturbance coefficients obtained by continuous monitoring of several subjects with secondary hemorrhage after treatment surgery under electromagnetic wave signals in the first, second, third, fourth, and fifth propagation paths.

Images