Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The technical scheme of the invention is described in detail below by specific examples. The following embodiments may be combined with each other, and some embodiments may not be repeated for the same or similar concepts or processes.
Fig. 1 exemplarily shows a schematic diagram of a brain disease prevention information push system according to an embodiment of the present invention, the system including:
the system comprises motion detection equipment, a communication assembly, a processor, a cloud server and an information receiving terminal;
The motion detection device comprises a waist wearable motion detection device, a knee wearable motion detection device and an ankle wearable motion detection device;
the processor is configured to:
acquiring the relative position relationship among a plurality of motion detection devices at a plurality of moments within a preset time period under the condition that a person to be detected wears the motion detection devices;
Determining pose data of the personnel to be tested according to the relative position relation;
the pose data of a plurality of moments in a preset time period are sent to a cloud server through a communication assembly;
the cloud server is used for:
receiving pose data of a person to be detected;
Retrieving brain CT images and brain electricity data of the person to be detected from the memory;
Obtaining brain disease prevention information of a person to be detected according to the pose data, the brain CT image and the brain electricity data;
And pushing the brain disease prevention information to an information receiving terminal.
According to the brain disease prevention information pushing system provided by the embodiment of the invention, the pose data of the personnel to be detected can be monitored in real time through wearing the motion detection equipment, a data base is provided for timely finding gait abnormalities, brain disease prevention information can be generated together with information such as CT images and brain electricity data, early brain diseases can be timely found and pushed to the information receiving terminal, and the personnel to be detected can be reminded of paying attention to the brain diseases.
According to one embodiment of the invention, the waist wearable motion detection device comprises a waist protector and a first ultrasonic sensor integrated in the waist protector, the knee wearable motion detection device comprises a knee protector and a second ultrasonic sensor integrated in the knee protector, the ankle wearable motion detection device comprises an ankle protector and a third ultrasonic sensor integrated in the ankle protector, and the frequencies of ultrasonic waves emitted by the first ultrasonic sensor, the second ultrasonic sensor and the third ultrasonic sensor are different from each other, so that the ultrasonic sensors can recognize echoes of ultrasonic waves emitted by the ankle wearable motion detection device.
According to one embodiment of the invention, the relative position relation among a plurality of motion detection devices is acquired at a plurality of moments within a preset time period when a person to be detected wears the motion detection device, the method comprises the steps of setting a plurality of measurement marks at preset positions in a test area, transmitting ultrasonic waves with a first frequency through a first ultrasonic sensor, receiving ultrasonic waves with the first frequency of the plurality of measurement marks, determining a first distance between a first ultrasonic sensor and the plurality of measurement marks according to a time difference between the ultrasonic waves with the first frequency and the ultrasonic waves with the first frequency, determining a first coordinate of the waist wearing type motion detection device in the test area according to the first distance, transmitting ultrasonic waves with a second frequency through a second ultrasonic sensor, receiving ultrasonic waves with the second frequency of the plurality of measurement marks, determining a second distance between a second ultrasonic sensor and the plurality of measurement marks according to a time difference between the ultrasonic waves with the second frequency and the ultrasonic waves with the second frequency, determining a first distance between the first ultrasonic sensor and the plurality of measurement marks according to the second distance, determining a third coordinate of the waist wearing type motion detection device in the test area, transmitting ultrasonic waves with the second frequency and the third ultrasonic waves with the third frequency and the third ultrasonic waves with the third ultrasonic wave with the third frequency and the third ultrasonic wave with the third frequency, a relative positional relationship between the plurality of motion detection devices is determined.
According to an embodiment of the present invention, based on a time difference between a receiving time of an ultrasonic echo and an emitting time of the ultrasonic wave, a distance between an ultrasonic sensor and each measurement mark may be determined, and a position of each measurement mark is known, so that, through the above analysis of the ultrasonic echo, a distance between the ultrasonic sensor and a plurality of known positions may be determined, and coordinates of the ultrasonic sensor may be obtained through a simultaneous equation set of coordinates of the distance and the known positions, the coordinates of the ultrasonic sensor being coordinates of the motion detection devices, and further, vectors between the respective motion detection devices may be determined based on the coordinates of the motion detection devices, and thus a relative positional relationship between the plurality of motion detection devices may be determined. The vectors between the motion detection devices can also be used as pose data of the personnel to be detected. The processor can send the pose data at a plurality of moments to the cloud server for analysis.
According to the embodiment of the invention, the cloud server can also call the brain CT image and the brain electricity data of the person to be tested for joint analysis when receiving the pose data of the person to be tested. Obtaining brain disease prevention information of a person to be detected according to the pose data, the brain CT image and the brain electricity data, wherein the brain disease prevention information comprises the steps of calling reference pose data of various brain diseases from a memory, determining gait discrimination information of the person to be detected according to the pose data and the reference pose data, detecting the brain CT image through an image detection model to obtain a first position area where brain tissues are located, determining brain CT discrimination information of the person to be detected according to the first position area, determining brain electricity discrimination information of the person to be detected according to the brain electricity data, and determining brain disease prevention information of the person to be detected according to the gait discrimination information, the brain CT discrimination information and the brain electricity discrimination information.
According to one embodiment of the present invention, the memory may include a history database, where a plurality of types of history data of a plurality of patients may be included, and reference pose data of an alzheimer disease patient, reference pose data of a parkinson disease patient, and reference pose data of a hydrocephalus patient may be obtained from the history database. The doctor can select pose data of a plurality of representative (for example, long-time and serious-illness) Alzheimer's disease patients at a plurality of moments when walking, make the change rates of the pose data of a plurality of Alzheimer's disease patients when walking consistent (namely, obtain the pose data of a plurality of Alzheimer's disease patients with consistent walking speed) through modes of slow release, fast forward and the like, and fuse the pose data of a plurality of Alzheimer's disease patients when walking (for example, the pose data comprises relative position relations among wearable motion detection devices, the relative position relations can be represented by position vectors of the same parts of the wearable motion detection devices, and fuse the position vectors of the same parts of the plurality of Alzheimer's disease patients through averaging, for example, obtain the reference pose data of the Alzheimer's disease patients. Similarly, reference pose data of a parkinson patient and reference pose data of a hydrocephalus patient can be obtained in a similar manner to reference pose data of an alzheimer patient, and are not described in detail herein.
According to one embodiment of the invention, gait discrimination information of a person to be detected is determined according to the pose data and the reference pose data, and the gait discrimination information of the person to be detected is obtained according to the gait similarity of the reference pose data and the reference pose data of various brain diseases.
According to one embodiment of the invention, the pose data of the person to be tested can be respectively compared with the reference pose data of the patient with Alzheimer's disease, the reference pose data of the patient with Parkinson's disease and the reference pose data of the patient with hydrocephalus so as to determine the gait discrimination information, for example, the gait discrimination information is information in a vector form, and the vector can have three components, namely, the gait similarity of the pose data of the person to be tested and the reference pose data of the patient with Alzheimer's disease, the gait similarity of the pose data of the person to be tested and the reference pose data of the patient with Parkinson's disease and the gait similarity of the pose data of the person to be tested and the reference pose data of the patient with hydrocephalus are respectively, so that the maximum similarity of the pose data of the person to be tested and the hydrocephalus can be determined based on the maximum value of the three similarities, namely, the gait characteristics of which disease is most likely to be present.
According to one embodiment of the invention, gait similarity of reference pose data and reference pose data of various brain diseases is determined through a pose similarity recognition model, and the method comprises the steps of inputting pose data of an Alzheimer disease diagnosis patient, reference pose data of the Alzheimer disease patient, reference pose data of a Parkinson disease patient and reference pose data of a hydrocephalus patient in a historical database into the pose similarity recognition model to obtain a first discrimination vector; inputting the pose data of the hydrocephalus diagnosis patient, the Alzheimer's disease patient reference pose data and the hydrocephalus patient reference pose data in the historical database into a pose similarity recognition model to obtain a second discrimination vector, inputting the pose data of the hydrocephalus diagnosis patient, the Alzheimer's disease patient reference pose data, the Parkinson's disease patient reference pose data and the hydrocephalus patient reference pose data in the historical database into a pose similarity recognition model to obtain a third discrimination vector, inputting the pose data of a control person without brain diseases, the Alzheimer's disease patient reference pose data, the Parkinson's disease patient reference pose data and the hydrocephalus patient reference pose data into a pose similarity recognition model to obtain a fourth discrimination vector, inputting the pose data of an experimental patient suffering from any one of the three cerebral diseases of Alzheimer's disease, parkinson's disease and hydrocephalus, the Alzheimer's disease patient reference pose data, the hydrocephalus patient reference pose data and the hydrocephalus patient reference pose data into a fifth discrimination vector, the method comprises the steps of obtaining a model of the position and the posture of a person to be tested, obtaining a model of the position and the posture of the person to be tested, obtaining a model of the posture and the posture of the person to be tested, obtaining reference position and posture data of the person to be tested, reference position and posture data of the person to be tested and reference position and posture data of the person to be tested, and inputting the model of the posture and the posture of the person to be tested into the model of the posture and the posture of the person to be tested according to the first, second, third, fourth and fifth discrimination vectors to obtain a loss function of the posture and the posture of the person to be tested, training the posture and the posture of the person to be tested according to the loss function of the posture and the posture of the person to be tested, and obtaining gait similarity of the reference position and posture data of various brain diseases.
According to one embodiment of the invention, the pose similarity recognition model is a BP neural network model and is used for determining feature similarity between pose data of a person to be detected and reference pose data of an Alzheimer disease patient, reference pose data of a Parkinson disease patient and reference pose data of a hydrocephalus patient and obtaining gait discrimination information. Because the disease features of the patient are not obvious in the early stage of disease, the similarity of the pose data of the personnel to be detected and various reference pose data is judged manually, and the difficulty of determining which disease is most likely to suffer from is high, so that the judgment can be carried out through a pose similarity recognition model, and the similarity of the features of different dimensions of the pose data of the personnel to be detected and various reference pose data can be analyzed.
According to one embodiment of the invention, the pose similarity recognition model may be trained prior to processing the pose data using the pose similarity recognition model. Pose data in a historical database, pose data of control personnel without brain diseases and pose data of experimental patients without brain diseases, which do not belong to the historical database, can be used as training samples, so that the brain diseases can be distinguished by the pose similarity recognition model, the capability of non-disease factors is eliminated, and the accuracy of the pose similarity recognition model is improved.
According to the embodiment of the invention, the pose data of the Alzheimer's disease diagnosis patient in the historical database and the plurality of reference pose data are input into the pose similarity recognition model to obtain the first discrimination vector, so that whether the pose similarity recognition model can judge whether the feature similarity of the pose data of the Alzheimer's disease diagnosis patient and the reference pose data of the Alzheimer's disease patient is highest or not is determined, and the feature similarity of the pose data and the other two reference pose data is lower.
According to the embodiment of the invention, the pose data of the parkinsonism-diagnosed patient in the historical database and the plurality of reference pose data can be input into the pose similarity recognition model to obtain the second discrimination vector, so that whether the pose similarity recognition model can judge that the feature similarity of the pose data of the parkinsonism-diagnosed patient and the reference pose data of the parkinsonism patient is highest or not and the feature similarity of the other two reference pose data is lower is determined.
According to one embodiment of the invention, the pose data of the hydrocephalus diagnosis patient in the history database and the plurality of reference pose data are input into the pose similarity recognition model to obtain the third discrimination vector, so that whether the pose similarity recognition model can judge that the feature similarity of the pose data of the hydrocephalus diagnosis patient and the reference pose data of the hydrocephalus patient is highest or not, and the feature similarity of the pose data and the other two reference pose data is lower.
According to one embodiment of the invention, the pose of a patient suffering from a brain disease is not completely different from the pose of a person without a brain disease when walking, and some similarities may exist, and the similarities may be used as non-disease factors to disturb the resolution of a pose similarity recognition model on the brain disease patient, so that the pose similarity recognition model can be trained to exclude the interference of the non-disease factors, and thus focus attention on the recognition of the characteristics of different diseases. The pose data of the contrast person without brain diseases and the plurality of reference pose data can be input into a pose similarity recognition model to obtain a fourth discrimination vector, so that whether the pose similarity recognition model can determine that the feature similarity of the pose data of the contrast person without brain diseases and the plurality of reference pose data is lower or not and whether non-disease factors can be eliminated is determined.
According to the embodiment of the invention, the pose data of the experimental patient which does not belong to the historical database but has any one of the three brain diseases including Alzheimer disease, parkinson disease and hydrocephalus and the plurality of reference pose data are input into the pose similarity recognition model to obtain the fifth discrimination vector so as to determine whether the pose similarity recognition model can be suitable for discriminating the pose data outside the historical database, and the robustness of the pose similarity recognition model is improved.
According to one embodiment of the invention, obtaining a loss function of a pose similarity recognition model according to the first discrimination vector, the second discrimination vector, the third discrimination vector, the fourth discrimination vector and the fifth discrimination vector comprises obtaining a loss function of a pose similarity recognition model according to formula (1),
(1)
Wherein,A first discrimination vector corresponding to pose data of the patient is determined for the ith Alzheimer's disease,For the first set of annotation vectors to be preset,For the second annotation vector to be preset,For the third set of annotation vectors to be preset,The number of pose data for a patient diagnosed for alzheimer's disease in a training batch,A second discrimination vector corresponding to the pose data of the patient is determined for the jth parkinsonism,The number of pose data for a patient diagnosed with parkinson's disease in a training batch,A third discrimination vector corresponding to the pose data of the kth hydrocephalus diagnosis patient is determined,The number of pose data for a hydrocephalus diagnosis patient in a training batch, and,Is a two-norm number of the two-norm,As a function of the conditions,A fourth discrimination vector corresponding to the pose data of the t comparison personnel,For the number of pose data of the control person in a training batch,Is thatIs used to determine the transposed vector of (c),A fifth discrimination vector corresponding to the pose data of the s-th experimental patient,For the number of pose data of the experimental patient in one training batch,Is the labeling vector of the s-th experimental patient,、、、AndIs a preset weight, i is less than or equal to,j≤,k≤,t≤,s≤And i,、j、、k、、t、Sum of sAre all positive integers.
According to an embodiment of the present invention, the first labeling vector is a theoretical value of gait discrimination information obtained by processing pose data of an alzheimer's disease patient by the pose similarity recognition model, for example, in the first labeling vector, the similarity between the pose data and reference pose data of the alzheimer's disease patient is highest (for example, the similarity is 100%), and the similarity between the pose data and the reference pose data of the other two reference pose data is lower (for example, the similarity is 0). The second labeling vector pose similarity recognition model processes pose data of the parkinson's disease patient to obtain a theoretical value of gait discrimination information, for example, in the second labeling vector, the similarity between the pose data and reference pose data of the parkinson's disease patient is highest (for example, the similarity is 100%), and the similarity between the pose data and reference pose data of the other two types of reference pose data is lower (for example, the similarity is 0). The third labeling vector pose similarity recognition model processes pose data of the hydrocephalus patient to obtain a theoretical value of gait discrimination information, for example, in the third labeling vector, the pose data has the highest similarity (for example, the similarity is 100%) with reference pose data of the hydrocephalus patient, and has lower similarity (for example, the similarity is 0) with other two reference pose data.
According to one embodiment of the present invention, in equation (1),The module representing the difference between the first discrimination vector and the first labeling vector is smaller than the module representing the difference between the first discrimination vector and the second labeling vector and smaller than the module representing the difference between the first discrimination vector and the third labeling vector, that is, the similarity between the first discrimination vector and the first labeling vector is the highest, and when the condition is satisfied, the judgment representing the pose similarity recognition model is correct, and only the specific feature similarity value has an error with the first labeling vector, so that the condition function value can be the module representing the difference between the first discrimination vector and the first labeling vectorOtherwise, the pose similarity recognition model is judged to be wrong, namely, the first judgment vector is the wrong judgment vector, the first judgment vector is the judgment error, and the value of the model can be taken as a conditional function value. Therefore, the condition function value is an error under two conditions, and the average value of errors corresponding to the pose data of the patient diagnosed by the Alzheimer's disease in one training batch can be obtained and used as one of the loss functions, so that the loss function can be reduced in the training process, the errors are reduced, and the judgment accuracy of the pose data of the patient diagnosed by the Alzheimer's disease is improved.
In accordance with one embodiment of the present invention,The module representing the difference between the second discrimination vector and the second labeling vector is smaller than the module representing the difference between the second discrimination vector and the first labeling vector and smaller than the module representing the difference between the second discrimination vector and the third labeling vector, that is, the similarity between the second discrimination vector and the second labeling vector is the highest, and when the condition is satisfied, the judgment representing the pose similarity recognition model is correct, and only the specific feature similarity value has an error with the second labeling vector, so that the condition function value can be the module representing the difference between the second discrimination vector and the second labeling vectorOtherwise, the second discrimination vector is the erroneous discrimination vector, and the second discrimination vector is the erroneous discrimination error, and the value of the model can be used as the conditional function value. Therefore, the condition function value is an error under two conditions, the average value of the errors corresponding to the pose data of the patient diagnosed with the Parkinson disease in one training batch can be obtained, and the average value is used as one of the loss functions, so that the loss function can be reduced in the training process, the errors are reduced, and the accuracy of judging the pose data of the patient diagnosed with the Parkinson disease is improved.
In accordance with one embodiment of the present invention,The module representing the difference between the third discrimination vector and the third labeling vector is smaller than the module representing the difference between the third discrimination vector and the first labeling vector and smaller than the module representing the difference between the third discrimination vector and the second labeling vector, that is, the similarity between the third discrimination vector and the third labeling vector is the highest, and when the condition is satisfied, the judgment representing the pose similarity recognition model is correct, and only the specific feature similarity value has an error with the third labeling vector, so that the condition function value can be the module representing the difference between the third discrimination vector and the third labeling vectorOtherwise, the pose similarity recognition model is judged to be wrong, namely, the third judgment vector is the wrong judgment vector, and the third judgment vector is the judgment error, and the value of the model can be taken as a conditional function value. Therefore, the condition function value is an error under two conditions, the average value of errors corresponding to the pose data of the hydrocephalus diagnosis patient in one training batch can be obtained, and the average value is used as one of the loss functions, so that the loss function can be reduced in the training process, the errors are reduced, and the accuracy of judging the pose data of the hydrocephalus diagnosis patient is improved.
According to one embodiment of the present invention, the first labeling vector, the second labeling vector and the third labeling vector are vectors composed of similarity to three brain diseases, and among the three similarity values, the similarity to the reference pose data of the patient with one of the diseases is higher, and the similarity to the rest two diseases is lower, so that the first labeling vector, the second labeling vector and the third labeling vector are added to obtain a vector composed of three 100% similarity values, for example, a vector composed of three 100% similarity values.The cosine similarity of the fourth discrimination vector and the sum of the three labeling vectors is obtained by processing pose data of a person without brain diseases, so that the similarity of the fourth discrimination vector and the three brain diseases is lower in theory, and the similarity of the fourth discrimination vector and the sum of the three labeling vectors is lower, and therefore, the average value of cosine similarity corresponding to the fourth discrimination vector of each person without brain diseases can be used as one item of a loss function, and therefore, the cosine similarity can be reduced in the training process, the difference between the fourth discrimination vector and the sum of the three labeling vectors can be improved, interference of non-disease factors can be eliminated, and the judgment accuracy of the pose similarity recognition model can be improved.
In accordance with one embodiment of the present invention,The fifth discrimination vector corresponding to the pose data of the experimental patient is a model of a difference between the corresponding labeling vector, the labeling vector is related to the type of brain diseases suffered by the experimental patient, and the numerical value of the component is the same as the first labeling vector, the second labeling vector or the third labeling vector, which are not described herein. The model value is the error value of the fifth discrimination vector, and the average value of the error values of the fifth discrimination vector corresponding to each experimental patient can be used as one item of the loss function, so that the error value is reduced in the training process, the accuracy of the pose similarity recognition model is improved, the pose similarity recognition model is applicable to discrimination of pose data outside a historical database, and the robustness of the pose similarity recognition model is improved.
According to one embodiment of the invention, the five items can be weighted and summed to obtain the loss function of the pose similarity recognition model, the loss function can be counter-propagated, and the parameters of the pose similarity recognition model can be adjusted through a gradient descent method, so that the pose similarity recognition model is trained, and the judgment accuracy and the robustness are improved. After multiple times of training, a trained pose similarity recognition model can be obtained. The pose similarity recognition model can accurately recognize the pose characteristics of the gait of the person to be detected in early disease stage, and judge the disease type of the person to be detected, so that gait discrimination information of the person to be detected is obtained.
In this way, the accuracy of the judgment of the pose similarity recognition model can be trained through pose data in the historical database, and when the judgment of the model is wrong, the wrong judgment vector is used as one item of the loss function, so that the training strength is improved, and the probability of wrong judgment is reduced. The pose similarity recognition model can be trained through pose data of a control person without brain diseases to eliminate the interference of non-disease factors, and the anti-interference capability and judgment accuracy of the pose similarity recognition model are improved. Furthermore, the robustness of the pose similarity recognition model can be trained through pose data of experimental patients with brain diseases, which do not belong to a historical database, so that the pose similarity recognition model can be suitable for distinguishing pose data outside the historical database, and the performance of the pose similarity recognition model is improved.
According to one embodiment of the invention, after the pose similarity recognition model is trained, the gait similarity of the pose data and the reference pose data of various brain diseases, namely, the gait similarity of the gait of the person to be tested and the gait of various brain disease patients can be determined through the trained pose similarity recognition model, so that gait discrimination information is formed.
According to one embodiment of the invention, the image detection model is a deep learning neural network model, for example, a convolutional neural network model, and can monitor a first location area where brain tissue is located in a brain CT image to determine the edge of the first location area.
According to one embodiment of the invention, the brain CT image may be a three-dimensional CT image, may comprise a plurality of two-dimensional CT image slices, and the region of brain tissue may be detected in each two-dimensional CT image slice, i.e. a first location region in the plurality of brain CT images.
According to one embodiment of the invention, the brain CT distinguishing information of the person to be detected is determined according to the first position area, and the method comprises the steps of obtaining a first circumcircle of the first position area in a plurality of brain CT images, obtaining a middle brain CT image in the plurality of brain CT images, obtaining the circle center of the first circumcircle in the middle brain CT image, obtaining an upper brain CT image in the plurality of brain CT images, obtaining the radius of the first circumcircle in the upper brain CT image, identifying the area where the ventricle is located in the upper brain CT image, determining the second circumcircle of the area where the ventricle is located, obtaining the radius of the second circumcircle, and obtaining the brain CT distinguishing information of the person to be detected according to the circle center of the first circumcircle in the middle brain CT image, the radius of the first circumcircle in the upper brain CT image and the radius of the second circumcircle.
According to one embodiment of the present invention, a first circumscribed circle of a first location area in each two-dimensional CT image slice may be determined. Further, according to the circle center of the first circumscribing circle, a first circular test line and a second circular test line are determined, wherein the circle center of the first circumscribing circle is used as the circle center of the first circumscribing circle, the radiuses of the first circular test line and the second circular test line are smaller than the radius of the first circumscribing circle, the length ratio of the radius of the first circular test line to the radius of the first circumscribing circle is equal to a first preset proportion, the length ratio of the radius of the second circular test line to the radius of the first circumscribing circle is equal to a second preset proportion, and the first preset proportion is larger than the second preset proportion.
According to one embodiment of the present invention, the scanning direction of the brain CT image may be from top to bottom, and a two-dimensional CT image slice having a specific sequence number may be selected as a midbrain CT image in which midbrain substantia nigra is included at a central position of a first location area where brain tissue is located in a middle area of the top-bottom direction. Determining the center of a first circumscribing circle in the midbrain CT image, determining a first circular test line and a second circular test line by taking the center of the first circumscribing circle as the center of the circle, wherein the radius of the first circular test line is larger than that of the second circular test line, and the radii of the first circular test line and the second circular test line are smaller than that of the first circumscribing circle. The first circular test line was used to test whether atrophy occurred in the area of the brain tissue margin and whether lateral fissures between different sections of brain tissue were widened. The second circular test line was used to test whether the midbrain substantia nigra was degenerated.
According to one embodiment of the present invention, the upper brain CT image is a two-dimensional CT image slice having a specific sequence number in an upper region in the up-down direction. In the upper brain CT image, including the area of the ventricle, it can be identified, for example, by an image identification model, which is a convolutional neural network model. The second circumscribed circle of the area where the ventricle is located can be determined, the radius of the second circumscribed circle can be determined, the radius of the first circumscribed circle in the upper brain CT image can be determined, and the radius of the second circumscribed circle and the radius of the first circumscribed circle can be used for judging whether the area where the ventricle is located is enlarged or not.
According to one embodiment of the invention, brain CT distinguishing information of a person to be detected is obtained according to the circle center of a first circumcircle in a middle brain CT image, the radius of the first circumcircle in an upper brain CT image and the radius of a second circumcircle, wherein the method comprises the steps of determining the image similarity of the brain CT image of the person to be detected and brain CT images of various diseases according to the circle center of the first circumcircle in the middle brain CT image, the radius of the first circumcircle in the upper brain CT image and the radius of the second circumcircle, and determining the brain CT distinguishing information of the person to be detected according to the image similarity.
According to one embodiment of the invention, the image similarity of the brain CT image of the person to be detected and the brain CT image of various diseases is determined according to the circle center of a first circumcircle in the middle brain CT image, the radius of the first circumcircle in the upper brain CT image and the radius of a second circumcircle, and the image similarity of the brain CT image of the person to be detected and the brain CT image of hydrocephalus is determined according to a first circular test line in the middle brain CT image, the image similarity of the brain CT image of the person to be detected and the brain CT image of Parkinson disease is determined according to a second circular test line in the middle brain CT image, and the image similarity of the brain CT image of the person to be detected and the brain CT image of hydrocephalus is determined according to the radius of the first circumcircle in the upper brain CT image and the radius of the second circumcircle.
According to one embodiment of the invention, the method for determining the image similarity of the brain CT image of the person to be tested and the Alzheimer's disease brain CT image according to the first circular test line in the middle brain CT image comprises the steps of obtaining a first critical point selection condition on the first circular test line according to a formula (2),
(2)
Wherein,Is the coordinates of the a-th pixel point on the first circular test line,Is the pixel value of the a-th pixel point on the first circular test line,Is the direction vector at the a-th pixel point on the first circular test line,The pixel value of the a pixel point on the first circular test line is along the direction vectorIs used for the gradient of (a),A first preset proportion is obtained by conforming to the critical point selection condition on a first round test lineThe pixel points of the brain CT image and the Alzheimer's disease brain CT image of the person to be tested are obtained according to the formula (3),
(3)
Wherein,Is an average value of pixel values of a plurality of pixel points between the b-th first critical point and the b+1th first critical point on the first circular test line,For a first predetermined pixel value,For the serial number of the b first critical point in the plurality of pixel points on the first circular test line,For the number of the (b+1) th first critical point in the plurality of pixel points on the first circular test line,For a first predetermined width of the sheet material,,For the number of first critical points in the first set of critical points,B is less than or equal to,b、AndAre all positive integers.
According to one embodiment of the present invention, in equation (2),The pixel value of the a pixel point on the first circular test line is along the direction vectorAnd (b) representing the rate of change of the pixel values along the direction vector, and if the rate of change is greater than or equal to a first preset ratio, representing that the first circular test line crosses the boundary of the brain tissue and the background area along the direction of the direction vector, so that the a-th pixel point is a first critical point on the boundary of the brain tissue and the background area, determining all the first critical points on the first circular test line, and obtaining a first critical point set.
According to one embodiment of the invention, brain CT for alzheimer's disease is typically characterized by atrophy of areas of the brain tissue edges, widening of the side-fissures between different sections of brain tissue. If the position of the first circular test line between the two first critical points is a background area, the first circular test line between the two first critical points passes through side cracks between brain tissues of different partitions, so that the distance between the two first critical points can be used as a precursor of Alzheimer's disease if the distance between the two first critical points is increased compared with a normal level.
According to one embodiment of the present invention, in equation (3),Is shown inIf true, the outer layer condition function value is 1, otherwise。
In accordance with one embodiment of the present invention,Is shown inIf true, the inner layer condition function value isOtherwise, 0.The average value of the pixel values representing the b-th first critical point and the b+1th first critical point on the first circular test line is higher than or equal to a first preset pixel value, the pixel value representing the pixel point between the two first critical points is closer to the background color, therefore, the first circular test line between the two first critical points passes through the side cracks between brain tissues of different partitions, in this case, the inner layer condition function value is thatThat is, the serial number difference between the two first critical points may represent the distance between the two first critical points on the first circular test line. If the above condition is not satisfied, the inner layer condition function value is 0.The sum of the distances of the first critical points at the two ends of the side crack can also be expressed as the sum of the widths of the side cracks.Wherein, the method comprises the steps of, wherein,Indicating that the first circular test line between the two first critical points crosses the side crack, the condition function value is 1, otherwise 0, and therefore,I.e. the number of side cracks. Thus, the first and second substrates are bonded together,The average width of the side cracks can be represented, if the average width is greater than or equal to the first preset width, the outer layer condition function value is 1, the probability that the person to be tested suffers from Alzheimer's disease is 1, and the image similarity between the brain CT image of the person to be tested and the brain CT image of Alzheimer's disease is also 1.
According to one embodiment of the present invention, if the average width is smaller than the first preset width, the outer layer condition function value isI.e. the ratio of the average width of the side cracks to the first preset width, in which case the larger the average width of the side cracks, the greater the possibility that the person to be tested suffers from alzheimer's disease, the value can be taken as the image similarity of the brain CT image of the person to be tested and the brain CT image of the alzheimer's disease.
According to an embodiment of the present invention, the value of the outer layer condition function may be used as a brain CT alzheimer's disease discrimination probability to describe the image similarity between the brain CT image of the person to be tested and the brain CT image of the alzheimer's disease.
In this way, the first critical point can be screened through the gradient change of the pixel value of the pixel point on the first circular test line, the average width of the side crack is determined through the condition function, and then whether the brain tissue of the person to be tested has atrophy or not is determined through the average width of the side crack, so that the width of the side crack is increased, the brain CT Alzheimer disease discrimination probability is obtained, and the image similarity of the brain CT image of the person to be tested and the brain CT image of Alzheimer disease is objectively and accurately described.
According to one embodiment of the invention, the image similarity of the brain CT image of the person to be tested and the brain CT image of the Parkinson's disease is determined according to a second circular test line in the middle brain CT image, and the method comprises the steps of obtaining a second critical point selection condition on the second circular test line according to a formula (4),
(4)
Wherein,The coordinates of the c-th pixel point on the second circular test line,The pixel value of the c-th pixel point on the second circular test line,For the direction vector at the c-th pixel point on the second circular test line,The pixel value of the c-th pixel point on the second circular test line is along the direction vectorIs used for the gradient of (a),A second preset proportion is obtained by a second round test line meeting the selection condition of a second critical pointA second critical point set on a second round test line is formed, and the image similarity between the brain CT image of the person to be tested and the brain CT image of the Parkinson's disease is determined according to the formula (5),
(5)
Wherein,Is an average value of pixel values of a plurality of pixel points between the d-th second critical point and the d+1-th second critical point on the second circular test line,For a second predetermined pixel value,For the serial number of the d second critical point in the plurality of pixel points on the second circular test line,Is the serial number of the (d+1) th second critical point in the plurality of pixel points on the second circular test line,For a second predetermined width, the first width,,For the number of second critical points in the second set of critical points,As a conditional function, d,AndAre all positive integers.
According to one embodiment of the present invention, the substantia nigra is significantly different from the pixel values of other brain tissues, and in case of parkinsonism, since dopamine neurons are damaged, the substantia nigra is atrophic and less than normal in width, it is possible to determine whether aura of parkinsonism exists based on the width of substantia nigra.
According to an embodiment of the present invention, similar to the screening method of the first critical point, the second critical point between the midbrain substantia nigra and other brain tissues on the second circular test line can be screened by selecting conditions based on the second critical point of the gradient, and a second critical point set is formed.
According to one embodiment of the present invention, in equation (5),Namely the total width of a plurality of midbrain substantia nigra,The number of midbrain substantia nigra, and therefore,Is the average width of the jejunum. The outer layer condition function indicates that if the average width of the midbrain substantia nigra is smaller than or equal to the second preset width, the average width of the midbrain substantia nigra is too low, the person to be tested has aura of parkinsonism, the outer layer condition function value is 1, otherwise the outer layer condition function value isThat is, the ratio between the second preset width and the average width of the midbrain substantia nigra, the larger the average width of the midbrain substantia nigra, the lower the ratio, the lower the probability of suffering from parkinson's disease. The outer layer condition function value in the formula (5) can be used as the image similarity of the brain CT image of the person to be detected and the brain CT image of the Parkinson's disease.
In this way, the second critical point can be screened through the gradient change of the pixel value of the pixel point on the second circular test line, the average width of the midbrain substantia nigra is determined through the condition function, and then whether the midbrain substantia nigra of the person to be tested is atrophic is determined through the average width of the midbrain substantia nigra, so that the image similarity of the brain CT image of the person to be tested and the brain CT image of the Parkinson's disease is obtained, and the possibility of Alzheimer's disease is accurately and objectively described.
According to one embodiment of the invention, the image similarity of the brain CT image and the hydrocephalus brain CT image of the person to be detected is determined according to the radius of the first circumcircle and the radius of the second circumcircle in the upper brain CT image, which comprises the steps of determining the image similarity of the brain CT image and the hydrocephalus brain CT image of the person to be detected according to a formula (6),
(6)
Wherein,Is the radius of the second outer circle,Is the radius of the first circumscribing circle,In the third preset proportion, the first preset proportion and the second preset proportion are respectively set,As a conditional function.
According to one embodiment of the present invention, in the formula (6), the condition function may represent that in the case where the ratio of the radius of the second circumscribed circle to the radius of the first circumscribed circle is greater than or equal to the third preset ratio, the condition function value is 1, otherwise. If the ratio is greater than or equal to the third preset ratio, the increase of the ventricle area is obvious, the aura of hydrocephalus exists, and the condition function value is 1. If the ratio is less than the third predetermined ratio, the ratio of the ratio to the third predetermined ratio may be used as a conditional function value, indicating that the greater the ratio, the greater the likelihood of hydrocephalus being present. The condition function value can be used as the image similarity of the brain CT image of the person to be tested and the hydrocephalus brain CT image to describe the possibility that the person to be tested suffers from hydrocephalus.
In this way, whether the ventricle is significantly enlarged can be described by the ratio of the radius of the second circumscribed circle to the radius of the first circumscribed circle, so that the image similarity of the brain CT image of the person to be tested and the brain CT image of hydrocephalus can be obtained, and the possibility that the person to be tested suffers from hydrocephalus can be objectively and accurately described.
According to one embodiment of the invention, the image similarity of the brain CT image of the person to be detected and the brain CT image of Alzheimer's disease, the image similarity of the brain CT image of the person to be detected and the brain CT image of Parkinson's disease, and the image similarity of the brain CT image of the person to be detected and the brain CT image of hydrocephalus are respectively used as components of the brain CT distinguishing information in a vector form, so that the brain CT distinguishing information is formed.
According to one embodiment of the invention, the brain electrical data of the person to be detected can be obtained, and the brain electrical distinguishing information of the person to be detected can be obtained according to the brain electrical data. Determining the brain electricity distinguishing information of the person to be detected according to the brain electricity data, wherein the brain electricity distinguishing information comprises determining the waveform characteristic similarity of the brain electricity data of the person to be detected and the brain electricity data of various brain diseases, and determining the brain electricity distinguishing information of the person to be detected according to the waveform characteristic similarity.
According to one embodiment of the invention, the characteristics of the brain electrical data of the Alzheimer's disease patient include a slow alpha rhythm in the occipital region, a significantly increased diffuse slow wave activity, and a reduced electroencephalogram fast wave activity. The intensity of low-frequency brain waves on the electroencephalogram of the patient with the Parkinson's disease is weak, and the response time of the brain wave potential reaching the peak value is long. The brain waves of hydrocephalus patients may have characteristics such as slowing down the overall rhythm. Therefore, the brain data of different patients in the historical database can be obtained, the frequency domain data of the brain data can be obtained through fast Fourier transformation, and the extreme points in the frequency domain data can be used as characteristic values to form the characteristic vectors of different types of brain diseases. Then, the electroencephalogram data of the person to be detected can be subjected to fast Fourier transform to obtain frequency domain data, and then the characteristic vector of the electroencephalogram data of the person to be detected can be obtained by taking the extreme point in the frequency domain data as the characteristic value. Further, the waveform characteristic similarity of the brain electrical data of the person to be detected and the brain electrical data of various brain diseases can be determined by comparing the similarity between the characteristic vector of the brain electrical data of the person to be detected and the characteristic vector of various brain diseases. In another example, the difference between the frequency domain data of the brain electrical data of different patients and the frequency domain data of the brain electrical data of the person to be tested in the database under different frequencies can be solved, and then the difference under different frequencies is integrated to obtain the difference between the brain electrical data of the person to be tested and the brain electrical data of the patients with various brain diseases, and further the relative difference between the brain electrical data of the person to be tested and the brain electrical data of the patients with various brain diseases can be solved, and the waveform characteristic similarity between the brain electrical data of the person to be tested and the brain electrical data of the patients with various brain diseases can be obtained by subtracting the relative difference from 1. Further, the brain electricity distinguishing information can be formed by the similarity of the waveform characteristics of the brain electricity data of the person to be detected and the brain electricity data of various brain diseases.
According to one embodiment of the present invention, gait discrimination information, brain CT discrimination information and brain electrical discrimination information may be weighted and summed, and in the resulting sum vector, each component may represent the similarity of the person under test to early symptoms of various brain diseases. Brain disease prevention information in which the similarity of the person to be tested to the early symptoms of various brain diseases can be described, respectively, can be generated based on the sum vector. Further, the brain disease prevention information can be pushed to an information receiving terminal, for example, a mobile phone of a person to be detected, a computer of a doctor and the like, so that the person to be detected is prompted to prevent the brain disease from further deepening.
According to the brain disease prevention information pushing system provided by the embodiment of the invention, the pose data of the personnel to be detected can be monitored in real time through wearing the motion detection equipment, a data base is provided for timely finding gait abnormalities, brain disease prevention information can be generated together with information such as CT images and brain electricity data, early brain diseases can be timely found and pushed to the information receiving terminal, and the personnel to be detected can be reminded of paying attention to the brain diseases. When the pose similarity recognition model is trained, the judgment accuracy of the pose similarity recognition model can be trained through pose data in the historical database, and when errors occur in judgment, the wrong judgment vector is used as one item of a loss function, so that the training strength is improved, and the probability of occurrence of judgment errors is reduced. The pose similarity recognition model can be trained through pose data of a control person without brain diseases to eliminate the interference of non-disease factors, and the anti-interference capability and judgment accuracy of the pose similarity recognition model are improved. Furthermore, the robustness of the pose similarity recognition model can be trained through pose data of experimental patients with brain diseases, which do not belong to a historical database, so that the pose similarity recognition model can be suitable for distinguishing pose data outside the historical database, and the performance of the pose similarity recognition model is improved. When the image similarity of the brain CT image of the person to be detected and the brain CT image of the Alzheimer's disease is determined, a first critical point can be screened through the gradient change of the pixel value of the pixel point on the first circular test line, the average width of the side crack is determined through a condition function, and then whether the brain tissue of the person to be detected has atrophy or not is determined through the average width of the side crack, so that the width of the side crack is increased, the brain CT Alzheimer's disease discrimination probability is obtained, and the image similarity of the brain CT image of the person to be detected and the brain CT image of the Alzheimer's disease is objectively and accurately described. When the image similarity of the brain CT image of the person to be tested and the brain CT image of the Parkinson's disease is determined, the second critical point can be screened through the gradient change of the pixel value of the pixel point on the second circular test line, the average width of the midbrain substantia nigra is determined through a condition function, and then whether the midbrain substantia nigra of the person to be tested is atrophic is determined through the average width of the midbrain substantia nigra, so that the image similarity of the brain CT image of the person to be tested and the brain CT image of the Parkinson's disease is obtained, and the possibility of suffering from Alzheimer's disease is accurately and objectively described. When the image similarity of the brain CT image and the hydrocephalus brain CT image of the person to be detected is determined, whether the ventricle is obviously enlarged or not can be described through the ratio of the radius of the second circumscribed circle to the radius of the first circumscribed circle, so that the image similarity of the brain CT image and the hydrocephalus brain CT image of the person to be detected is obtained, and the possibility that the person to be detected suffers from hydrocephalus is objectively and accurately described.
It will be appreciated by persons skilled in the art that the embodiments of the invention described above and shown in the drawings are by way of example only and are not limiting. The objects of the present invention have been fully and effectively achieved. The functional and structural principles of the present invention have been shown and described in the examples and embodiments of the invention may be modified or practiced without departing from the principles described.
It should be noted that the above embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the technical solution described in the above embodiments may be modified or some or all of the technical features may be equivalently replaced, and these modifications or substitutions do not make the essence of the corresponding technical solution deviate from the scope of the technical solution of the embodiments of the present invention.