Diagnosis and treatment system for intelligent classification diagnosis of vertigo disease based on deep learning

Abstract

The invention relates to the technical field of disease diagnosis, and particularly discloses a diagnosis and treatment system for intelligent classified diagnosis of vertigo diseases based on deep learning, and the system comprises a nystagmus sample collection module which is used for obtaining training set data; the model training module is used for obtaining an eye seismic waveform range; the model diagnosis module is used for obtaining a dizziness diagnosis result of the patient; the diagnosis result confirmation module is used for evaluating whether the dizziness diagnosis result is accurate or not based on the dizziness diagnosis result of the single target person; misleading reason analysis: based on the inaccurate diagnosis signal, analyzing whether a single target person misleads the diagnosis model or not, and if misleading occurs, generating a type misleading signal; the model autonomous learning module is used for carrying out learning updating on a single target person in an eye seismic waveform range on the basis of the type misleading signal; according to the method, the diagnosis accuracy is remarkably improved, and the misdiagnosis or missed diagnosis phenomenon caused by individual differences is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of disease diagnosis, and in particular to a diagnosis and treatment system for vertigo diseases based on intelligent classification and diagnosis based on deep learning. Background Art

[0002] Vertigo is a common clinical symptom with complex and diverse causes, including vestibular diseases, neurological diseases, etc. The traditional diagnosis of vertigo relies on the doctor's experience and the patient's description of symptoms, but this method has problems such as strong subjectivity and high misdiagnosis rate. In recent years, with the development of medical imaging technology and artificial intelligence, especially the application of deep learning in medical imaging diagnosis, new means have been provided for the diagnosis of vertigo.

[0003] The SRM Vertigo Treatment System is an advanced device for vertigo disease diagnosis and treatment. Its built-in camera can capture the patient's nystagmus data. However, the existing SRM system needs to export the captured nystagmus data and then perform manual analysis or learning. Due to individual differences and the complexity of the disease, traditional diagnostic methods often have the risk of misdiagnosis and missed diagnosis, which is not only cumbersome to operate, but also inefficient. Therefore, there is an urgent need for a system that can automatically learn, classify and diagnose vertigo diseases directly in the background. Summary of the invention

[0004] The purpose of the present invention is to provide a diagnosis and treatment system for vertigo disease based on intelligent classification and diagnosis based on deep learning to solve the technical problems in the above background.

[0005] The purpose of the present invention can be achieved by the following technical solutions: The present invention provides a diagnosis and treatment system for vertigo disease based on intelligent classification and diagnosis based on deep learning, comprising: Nystagmus sample collection module: used to collect nystagmus sample waveform curves from patients with past vertigo diseases, and divide the nystagmus sample waveform curves to obtain training set data; Model training module: used to input the training set data into the convolutional neural network to train the diagnostic model and obtain the nystagmus waveform range; Model diagnosis module: used to collect the nystagmus waveform curve of the target person, input the nystagmus waveform curve into the diagnosis model, compare it with the nystagmus waveform range, and obtain the patient's vertigo diagnosis result; Diagnosis result confirmation module: based on the dizziness diagnosis result of a single target person, evaluates whether the dizziness diagnosis result is accurate, and generates an inaccurate diagnosis signal when the dizziness diagnosis result is inaccurate; Misleading cause analysis: Based on the inaccurate diagnosis signal, analyze whether the single target person has misled the diagnostic model. If so, generate a type misleading signal; Model autonomous learning module: Based on the type misleading signal, obtain the nystagmus misleading waveform curve of the dizziness type corresponding to the type misleading signal. At the same time, based on the nystagmus waveform range of the dizziness type in the diagnostic model, obtain the dizziness characterization curve, and compare and analyze the nystagmus misleading waveform curve with the dizziness characterization curve, calculate the misleading coefficient, and thus learn and update the nystagmus waveform range for a single target person.

[0006] As a further solution of the present invention: the process of obtaining the training set data is: The nystagmus sample waveform curves of patients with different vertigo types and past vertigo diseases are extracted, and each nystagmus sample waveform curve is combined with its corresponding vertigo type and marked as training set data.

[0007] As a further solution of the present invention: the process of obtaining the nystagmus waveform range is: The waveform curves of nystagmus samples of each type of vertigo are respectively input into the convolutional neural network for diagnostic model training. After the convolutional layer features are extracted, the pooling layer divides the area and extracts the nystagmus information and outputs the corresponding target value. The target value is subtracted from the actual value of the nystagmus to obtain the deviation value, and it is judged whether the deviation value is within the allowable range. When it is within the allowed range, the weights and thresholds after convolutional neural network training are obtained to obtain the nystagmus waveform range of each vertigo type; When it is not within the allowable range, the error between the deviation value and the allowable range is calculated to obtain the error value; Calculate the gradient of the error value to obtain the error gradient value; The weights are updated according to the error gradient value, the updated weights are substituted into the convolutional neural network, and the calculation is re-started from the convolutional layer.

[0008] As a further solution of the present invention: the process of obtaining the patient's dizziness diagnosis result is: Capture the nystagmus data of the target person in real time to obtain the nystagmus waveform curve; The nystagmus waveform curve is input into the diagnostic model, the features of the nystagmus waveform curve are extracted in the convolution layer, the features are divided in the pooling layer, and the divided features are identified in the fully connected layer according to the dizziness type corresponding to the training set data to obtain the dizziness diagnosis result of the target person.

[0009] As a further solution of the present invention: the process of obtaining the inaccurate diagnosis signal is as follows: Based on a single target person, obtaining the dizziness diagnosis result of the target person by the model diagnosis module, and determining whether the dizziness diagnosis result is accurate; When the dizziness diagnosis result is inaccurate, an inaccurate diagnosis signal is generated, and the nystagmus waveform curve corresponding to the dizziness diagnosis result is marked as an inaccurate curve.

[0010] As a further solution of the present invention: the process of generating the type misleading signal is: Based on the inaccurate diagnosis signal, the inaccurate curves of each vertigo type of a single target person are extracted respectively; Based on any type of dizziness, all inaccurate curves are compared and analyzed to evaluate whether all inaccurate curves are synchronized. The specific evaluation method is as follows: Based on all inaccurate curves, extract several time points and the values ​​corresponding to each inaccurate curve, and mark them as comparison values; Then, based on the comparison values ​​of all the inaccurate curves at the same time point, all the comparison values ​​are sorted according to the magnitude of the values, and based on the sorting of the comparison values, the first quartile and the third quartile are obtained; Calculate the difference between the first quartile and the third quartile, and take the absolute value to obtain the floating representation value at a single moment; Preset a single moment floating threshold, and compare and analyze the single moment floating representation value corresponding to each moment point with the single moment floating representation threshold; If the single moment floating characterization value is less than or equal to the single moment floating characterization threshold, a stable moment signal is generated; If the single moment floating characterization value is greater than the single moment floating characterization threshold, an unstable moment signal is generated, and the single moment floating characterization value is marked as an unstable moment floating characterization value; The number of times all unstable moment signals are generated in all time points is extracted, and the ratio is calculated with the total number of all time points to obtain the unstable ratio; At the same time, all the floating characterization values ​​of all unstable moments in all time points are extracted, and the difference between each unstable moment floating characterization value and the floating characterization threshold of a single moment is calculated to obtain the floating difference of the unstable moment, and then the ratio of the floating difference of the unstable moment and the floating characterization threshold of a single moment is calculated to obtain the floating ratio of the unstable moment, and all the floating ratios of the unstable moments are summed up to obtain the total floating ratio of the unstable moments; By formula: , calculate and obtain the synchronous characterization value TB, where a1 and a2 are preset proportional factors, and both are greater than 0, BW is the unstable ratio, and BF is the total floating ratio at the unstable moment; Preset a synchronization threshold, and compare and analyze the synchronization characterization value with the synchronization threshold; If the synchronization characterization value is less than or equal to the synchronization threshold, a type misleading signal is generated.

[0011] As a further solution of the present invention: the process of obtaining the nystagmus misleading waveform curve is: Based on the type misleading signal, the median of each time point of all inaccurate curves in the dizziness type corresponding to the type misleading signal is extracted; Then the medians of all time points are connected with a smooth curve to obtain the nystagmus misleading waveform curve.

[0012] As a further solution of the present invention: the process of obtaining the vertigo characterization curve is: Based on the nystagmus waveform range of the vertigo type in the diagnostic model, the average value of the nystagmus waveform range at all time points is extracted and connected with a smooth curve to obtain the vertigo characterization curve.

[0013] As a further solution of the present invention: the process of obtaining the misleading coefficient is: The overlapping part between the nystagmus misleading waveform curve and the vertigo characterization curve is extracted to obtain the overlapping curve, and then the ratio of the overlapping curve length to the nystagmus misleading waveform curve length is calculated to obtain the misleading coefficient.

[0014] As a further solution of the present invention: the process of learning and updating the nystagmus waveform range for a single target person is as follows: Based on the misleading coefficient, the nystagmus waveform range of the vertigo type in the current diagnostic model is narrowed, and the narrowed nystagmus waveform range replaces the current nystagmus waveform range, thereby completing the learning and updating of the nystagmus waveform range for a single target person.

[0015] Beneficial effects of the present invention: The present invention collects nystagmus samples, trains diagnostic models, diagnoses the dizziness of the target person, and performs personalized learning and updating when the diagnostic results are inaccurate. It conducts in-depth analysis on individual diagnostic errors and identifies nystagmus waveform features that may lead to misleading. By comparing these features with the standard waveform range in the model, the system can intelligently adjust and optimize the model parameters to make them more suitable for the individual's physiological characteristics. The present invention not only significantly improves the accuracy of diagnosis, but also more effectively reduces misdiagnosis or missed diagnosis caused by individual differences. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below in conjunction with the accompanying drawings.

[0017] Figure 1 It is a flowchart of the present invention; Figure 2 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Embodiment 1: See also Figure 1 , Figure 2 As shown, the diagnosis and treatment system for vertigo disease based on intelligent classification diagnosis based on deep learning according to the embodiment of the present invention includes: Nystagmus sample collection module: used to collect nystagmus sample waveform curves from patients with past vertigo diseases, and divide the nystagmus sample waveform curves to obtain training set data; It should be explained that patients with past vertigo disorders refer to patients who have been diagnosed with vertigo disorders in the past; In some embodiments, nystagmus sample waveform curves of patients with different vertigo types and past vertigo diseases are extracted, and each nystagmus sample waveform curve is combined with its corresponding vertigo type and marked as training set data, so as to train the diagnostic model based on the training set data; Model training module: used to input the training set data into the convolutional neural network to train the diagnostic model and obtain the nystagmus waveform range; In some embodiments, the waveform curves of nystagmus samples of each type of vertigo are respectively input into a convolutional neural network for training of a diagnostic model. After feature extraction by the convolutional layer, the pooling layer divides the region and extracts nystagmus information (nystagmus information is nystagmus direction information, and nystagmus direction information includes horizontal nystagmus signals and vertical nystagmus signals) and outputs a corresponding target value. The target value is subtracted from the actual value of nystagmus to obtain a deviation value, and it is determined whether the deviation value is within an allowable range. When it is within the allowed range, the weights and thresholds after convolutional neural network training are obtained to obtain the nystagmus waveform range of each vertigo type; When it is not within the allowable range, the error between the deviation value and the allowable range is calculated to obtain the error value; Calculate the gradient of the error value to obtain the error gradient value; Update the weights according to the error gradient value, substitute the updated weights into the convolutional neural network, and recalculate from the convolutional layer; Model diagnosis module: used to collect the nystagmus waveform curve of the target person, input the nystagmus waveform curve into the diagnosis model, compare it with the nystagmus waveform range, and obtain the patient's vertigo diagnosis result; It should be explained that the target person refers to individuals who currently need to be diagnosed with vertigo disorders; In some implementation schemes, an electronystagmus device is used to capture the nystagmus data of the target person in real time to obtain a nystagmus waveform curve; The nystagmus waveform curve is input into the diagnostic model, the features of the nystagmus waveform curve are extracted in the convolution layer, the features are divided in the pooling layer, and the divided features are identified in the fully connected layer according to the vertigo type corresponding to the training set data to obtain the vertigo diagnosis result of the target person; The technical solution of the present invention is mainly as follows: the technical solution of the embodiment of the present invention realizes the accurate diagnosis of vertigo disease by constructing a diagnostic model based on deep learning, collects training data by using a nystagmus sample collection module, obtains a diagnostic model through training by a model training module, and performs real-time diagnosis of the target person by using the model diagnosis module. The embodiment of the present invention not only improves the diagnostic accuracy of vertigo disease, but also reduces the risk of misdiagnosis and missed diagnosis.

[0020] Embodiment 2: Based on the first embodiment, in the diagnosis of vertigo, due to the individual differences of different target persons, in order to ensure the accuracy of the diagnosis of vertigo of the target person, the vertigo diagnosis results of the diagnostic model are analyzed, and based on the vertigo diagnosis results of a single target person, the diagnostic model for the target person is learned and updated. Figure 1 , Figure 2 As shown, the diagnosis and treatment system for vertigo disease based on intelligent classification diagnosis based on deep learning according to the embodiment of the present invention also includes: Diagnosis result confirmation module: based on the dizziness diagnosis result of a single target person, evaluate whether the dizziness diagnosis result is accurate; In some embodiments, based on a single target person, the dizziness diagnosis result of the target person is obtained by the model diagnosis module, and the dizziness diagnosis result is evaluated by professionals in the field to determine whether the dizziness diagnosis result is accurate; If the vertigo diagnosis result is inaccurate, an inaccurate diagnosis signal is generated, and the nystagmus waveform curve corresponding to the vertigo diagnosis result is marked as an inaccurate curve; If the vertigo diagnosis result is accurate, a diagnosis accuracy signal is generated; Misleading cause analysis: Based on the inaccurate diagnosis signal, analyze whether the single target person has misled the diagnostic model, thus leading to the generation of inaccurate diagnosis signals; In some embodiments, based on the diagnosis inaccuracy signal, inaccuracy curves of each vertigo type of a single target person are extracted respectively; Based on any type of dizziness, all inaccurate curves are compared and analyzed to evaluate whether all inaccurate curves are synchronized. The specific evaluation method is as follows: Based on all inaccurate curves, extract several time points and the values ​​corresponding to each inaccurate curve, and mark them as comparison values; It should be explained that the number of time points should not be too small. The more time points there are, the more it can reflect the synchronization between all inaccurate curves. Then, based on the comparison values ​​of all inaccurate curves at the same time point, all comparison values ​​are sorted according to the size of the values ​​(the sorting method can be ascending or descending), and based on the sorting of the comparison values, the first quartile and the third quartile are obtained; It should be explained that the first quartile is the data in the first quarter position (if there is no data in the first quarter position, the value is taken downward to the nearest data point position), and similarly, the third quartile is the data in the third quarter position (if there is no data in the third quarter position, the value is taken upward to the nearest data point position); Calculate the difference between the first quartile and the third quartile, and take the absolute value to obtain the floating representation value at a single moment; Preset a single moment floating threshold, and compare and analyze the single moment floating representation value corresponding to each moment point with the single moment floating representation threshold; If the floating characterization value at a single moment is less than or equal to the floating characterization threshold at a single moment, it means that at the moment point corresponding to the floating characterization value at the single moment, the comparison value difference of all inaccurate curves is small and stable within a certain floating range, and a stable moment signal is generated; If the single moment floating characterization value is greater than the single moment floating characterization threshold, it means that at the time point corresponding to the single moment floating characterization value, the comparison values ​​of all inaccurate curves are greatly different and are not stable within a certain floating range, then an unstable moment signal is generated, and the single moment floating characterization value is marked as an unstable moment floating characterization value; The number of times all unstable moment signals are generated in all time points is extracted, and the ratio is calculated with the total number of all time points to obtain the unstable ratio; At the same time, all the floating characterization values ​​of all unstable moments in all time points are extracted, and the difference between each unstable moment floating characterization value and the floating characterization threshold of a single moment is calculated to obtain the floating difference of the unstable moment, and then the ratio of the floating difference of the unstable moment and the floating characterization threshold of a single moment is calculated to obtain the floating ratio of the unstable moment, and all the floating ratios of the unstable moments are summed up to obtain the total floating ratio of the unstable moments; By formula: , calculate and obtain the synchronous characterization value TB, where a1 and a2 are preset proportional factors, and both are greater than 0, BW is the unstable ratio, and BF is the total floating ratio at the unstable moment; Preset a synchronization threshold, and compare and analyze the synchronization characterization value with the synchronization threshold; If the synchronization characterization value is less than or equal to the synchronization threshold, it means that the synchronization of all inaccurate curves of the dizziness type of the single target person is high, that is, it is determined that in the dizziness type of the single target person, there is a misleading dizziness diagnosis result of the diagnostic model, that is, a type misleading signal is generated; If the synchronization characterization value is greater than the synchronization threshold, it means that the synchronization of all inaccurate curves of the single target person's vertigo type is low, that is, it is determined that in the vertigo type of the single target person, there is no misleading of the vertigo diagnosis result of the diagnostic model, that is, a non-misleading signal is generated; Model autonomous learning module: based on the type misleading signal, obtain the nystagmus misleading waveform curve of the dizziness type corresponding to the type misleading signal. At the same time, based on the nystagmus waveform range of the dizziness type in the diagnostic model, obtain the dizziness characterization curve, and compare and analyze the nystagmus misleading waveform curve with the dizziness characterization curve, calculate the misleading coefficient, and thus learn and update the nystagmus waveform range for a single target person; Exemplarily, the process of obtaining the nystagmus misleading waveform curve is as follows: Based on the type misleading signal, extract the median of each time point of all inaccurate curves in the dizziness type corresponding to the type misleading signal (the median is the data at the half position in the sorting of each time point of all inaccurate curves. If there is no data at the half position, the average of the two data points of the nearest upward data point and the nearest downward data point is taken); Then the medians of all time points are connected with a smooth curve to obtain the nystagmus misleading waveform curve; The process of obtaining the vertigo characterization curve is as follows: Based on the nystagmus waveform range of the vertigo type in the diagnostic model, the average value of the nystagmus waveform range at all time points is extracted, and the average value is connected with a smooth curve to obtain a vertigo characterization curve; The process of obtaining the misleading coefficient is: Extract the overlapping part between the nystagmus misleading waveform curve and the vertigo characterization curve to obtain the overlapping curve, and then calculate the ratio of the length of the overlapping curve to the length of the nystagmus misleading waveform curve to obtain the misleading coefficient; The process of learning and updating the nystagmus waveform range for a single target person is as follows: Based on the misleading coefficient, the nystagmus waveform range of the vertigo type in the current diagnostic model is narrowed, and the narrowed nystagmus waveform range replaces the current nystagmus waveform range, thereby completing the learning and updating of the nystagmus waveform range for a single target person, thereby more accurately reflecting the vertigo characteristics of a single target person. Through the continuous personalized learning and updating of the diagnostic model, the diagnostic model can more accurately diagnose the vertigo disease of a single target person. The technical solution of the present invention is mainly as follows: the embodiment of the present invention collects nystagmus samples, trains diagnostic models, diagnoses the dizziness of the target person, and performs personalized learning and updating when the diagnostic results are inaccurate. It conducts in-depth analysis on individual diagnostic errors and identifies nystagmus waveform features that may cause misleading. By comparing these features with the standard waveform range in the model, the system can intelligently adjust and optimize the model parameters to make them more in line with the physiological characteristics of the individual. The embodiment of the present invention not only significantly improves the accuracy of diagnosis, but also more effectively reduces misdiagnosis or missed diagnosis caused by individual differences.

[0021] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A diagnosis and treatment system for vertigo based on intelligent classification and diagnosis based on deep learning, characterized in that: include: Nystagmus sample collection module: used to collect nystagmus sample waveform curves from patients with past vertigo diseases, and divide the nystagmus sample waveform curves to obtain training set data; Model training module: used to input the training set data into the convolutional neural network to train the diagnostic model and obtain the nystagmus waveform range; Model diagnosis module: used to collect the nystagmus waveform curve of the target person, input the nystagmus waveform curve into the diagnosis model, compare it with the nystagmus waveform range, and obtain the patient's vertigo diagnosis result; Diagnosis result confirmation module: based on the dizziness diagnosis result of a single target person, evaluates whether the dizziness diagnosis result is accurate, and generates an inaccurate diagnosis signal when the dizziness diagnosis result is inaccurate; Misleading cause analysis: Based on the inaccurate diagnosis signal, analyze whether the single target person has misled the diagnostic model. If so, generate a type misleading signal; Model autonomous learning module: Based on the type misleading signal, obtain the nystagmus misleading waveform curve of the dizziness type corresponding to the type misleading signal. At the same time, based on the nystagmus waveform range of the dizziness type in the diagnostic model, obtain the dizziness characterization curve, and compare and analyze the nystagmus misleading waveform curve with the dizziness characterization curve, calculate the misleading coefficient, and thus learn and update the nystagmus waveform range for a single target person.

2. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 1 is characterized in that: The process of obtaining training set data is as follows: The nystagmus sample waveform curves of patients with different vertigo types and past vertigo diseases are extracted, and each nystagmus sample waveform curve is combined with its corresponding vertigo type and marked as training set data.

3. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 1 is characterized in that: The process of obtaining the nystagmus waveform range is as follows: The waveform curves of nystagmus samples of each type of vertigo are respectively input into the convolutional neural network for diagnostic model training. After the convolutional layer features are extracted, the pooling layer divides the area and extracts the nystagmus information and outputs the corresponding target value. The target value is subtracted from the actual value of the nystagmus to obtain the deviation value, and it is judged whether the deviation value is within the allowable range. When it is within the allowed range, the weights and thresholds after convolutional neural network training are obtained to obtain the nystagmus waveform range of each vertigo type; When it is not within the allowable range, the error between the deviation value and the allowable range is calculated to obtain the error value; Calculate the gradient of the error value to obtain the error gradient value; The weights are updated according to the error gradient value, the updated weights are substituted into the convolutional neural network, and the calculation is re-started from the convolutional layer.

4. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 1 is characterized in that: The process of obtaining a diagnosis of vertigo in a patient is: Capture the nystagmus data of the target person in real time to obtain the nystagmus waveform curve; The nystagmus waveform curve is input into the diagnostic model, the features of the nystagmus waveform curve are extracted in the convolution layer, the features are divided in the pooling layer, and the divided features are identified in the fully connected layer according to the dizziness type corresponding to the training set data to obtain the dizziness diagnosis result of the target person.

5. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 1 is characterized in that: The process of obtaining the inaccurate diagnosis signal is: Based on a single target person, obtaining the dizziness diagnosis result of the target person by the model diagnosis module, and determining whether the dizziness diagnosis result is accurate; When the dizziness diagnosis result is inaccurate, an inaccurate diagnosis signal is generated, and the nystagmus waveform curve corresponding to the dizziness diagnosis result is marked as an inaccurate curve.

6. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 5 is characterized in that: The process of generating a type misleading signal is: Based on the inaccurate diagnosis signal, the inaccurate curves of each vertigo type of a single target person are extracted respectively; Based on any type of dizziness, all inaccurate curves are compared and analyzed to evaluate whether all inaccurate curves are synchronized. The specific evaluation method is as follows: Based on all inaccurate curves, extract several time points and the values ​​corresponding to each inaccurate curve, and mark them as comparison values; Then, based on the comparison values ​​of all the inaccurate curves at the same time point, all the comparison values ​​are sorted according to the magnitude of the values, and based on the sorting of the comparison values, the first quartile and the third quartile are obtained; Calculate the difference between the first quartile and the third quartile, and take the absolute value to obtain the floating representation value at a single moment; Preset a single moment floating threshold, and compare and analyze the single moment floating representation value corresponding to each moment point with the single moment floating representation threshold; If the single moment floating characterization value is less than or equal to the single moment floating characterization threshold, a stable moment signal is generated; If the single moment floating characterization value is greater than the single moment floating characterization threshold, an unstable moment signal is generated, and the single moment floating characterization value is marked as an unstable moment floating characterization value; The number of times all unstable moment signals are generated in all time points is extracted, and the ratio is calculated with the total number of all time points to obtain the unstable ratio; At the same time, all the floating characterization values ​​of all unstable moments in all time points are extracted, and the difference between each unstable moment floating characterization value and the floating characterization threshold of a single moment is calculated to obtain the floating difference of the unstable moment, and then the ratio of the floating difference of the unstable moment and the floating characterization threshold of a single moment is calculated to obtain the floating ratio of the unstable moment, and all the floating ratios of the unstable moments are summed up to obtain the total floating ratio of the unstable moments; By formula: , calculate and obtain the synchronous characterization value TB, where a1 and a2 are preset proportional factors, and both are greater than 0, BW is the unstable ratio, and BF is the total floating ratio at the unstable moment; Preset a synchronization threshold, and compare and analyze the synchronization characterization value with the synchronization threshold; If the synchronization characterization value is less than or equal to the synchronization threshold, a type misleading signal is generated.

7. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 5 is characterized in that: The process of obtaining the nystagmus misleading waveform curve is as follows: Based on the type misleading signal, the median of each time point of all inaccurate curves in the dizziness type corresponding to the type misleading signal is extracted; Then the medians of all time points are connected with a smooth curve to obtain the nystagmus misleading waveform curve.

8. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 1 is characterized in that: The process of obtaining the vertigo characterization curve is as follows: Based on the nystagmus waveform range of the vertigo type in the diagnostic model, the average value of the nystagmus waveform range at all time points is extracted and connected with a smooth curve to obtain the vertigo characterization curve.

9. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 1 is characterized in that: The process of obtaining the misleading coefficient is: The overlapping part between the nystagmus misleading waveform curve and the vertigo characterization curve is extracted to obtain the overlapping curve, and then the ratio of the overlapping curve length to the nystagmus misleading waveform curve length is calculated to obtain the misleading coefficient.

10. The diagnosis and treatment system for vertigo disease based on deep learning and intelligent classification diagnosis according to claim 1 is characterized in that: The process of learning and updating the nystagmus waveform range for a single target person is as follows: Based on the misleading coefficient, the nystagmus waveform range of the vertigo type in the current diagnostic model is narrowed, and the narrowed nystagmus waveform range replaces the current nystagmus waveform range, thereby completing the learning and updating of the nystagmus waveform range for a single target person.

Images