WO2023069057A1

WO2023069057A1 - A computer-implemented method and system for assisting in the diagnosis of hyperactivity disorder

Info

Publication number: WO2023069057A1
Application number: PCT/TR2022/051161
Authority: WO
Inventors: Fatma LATIFOGLU; Esra DEMIRCI; Mustafa Yasin ESAS
Original assignee: Erciyes Universitesi Strateji Gelistirme Daire Baskanligi
Priority date: 2021-10-21
Filing date: 2022-10-19
Publication date: 2023-04-27

Abstract

The present invention relates to a computer-implemented method for creating a hyperactivity disorder diagnostic tool, comprising the steps of obtaining simultaneously measured galvanic skin response, pulse, and angular velocity and acceleration data of body movement of multiple individuals, who are diagnosed as having and not having hyperactivity disorder and constitute a control group; training the obtained data with artificial intelligence algorithms and classifying individuals as having/not having hyperactivity disorder; determining the most accurate classification by testing the classifications with the control group; if the determined classification is above a selected threshold value, creating a hyperactivity disorder diagnostic tool that includes the artificial intelligence algorithm in which the determined classification is made; and to a system comprising a processing unit for classifying the individual as having/not having hyperactivity disorder, by processing the simultaneously measured galvanic skin response, pulse, angular velocity and acceleration data of body movement of an individual to be diagnosed, with a hyperactivity disorder diagnostic tool.

Description

A COMPUTER-IMPLEMENTED METHOD AND SYSTEM FOR ASSISTING IN THE DIAGNOSIS OF HYPERACTIVITY DISORDER

Field of the Invention

The present invention relates to a computer- implemented method for assisting in the diagnosis of hyperactivity disorder and a system comprising a hyperactivity disorder diagnostic tool created with the said method.

Background of the Invention

Attention-deficit hyperactivity disorder (ADHD) is considered to be one of the most common neuropsychiatric disorders of childhood. ADHD is particularly prevalent in children between the ages of 8-12 and various studies indicate that the worldwide prevalence in childhood is around 15%. Hyperactivity disorder is a subtype of ADHD. The diagnosis of hyperactivity disorder is performed with psychiatric tests. For this purpose, the psychiatric analyses and tests specified in the book titled DSM-V (The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition) published by the American Psychiatric Association are applied to the patient with the help of a clinician. Due to the nature of the recent application, the subjective contributions of the evaluating clinician who performs the test also have a direct impact on the diagnosis. The statements of the child, parent, and teacher regarding the psychiatric condition and their subjective answers to the questions can affect the diagnosis. For example, the teacher or parent of the child may describe him/her as too restless, when in fact the child may be exhibiting normal behavior for his/her age. Therefore, subjective influences may lead to misdiagnosis. As a result of misdiagnosis due to the lack of an objective evaluation tool, patients may be directed to unnecessary psychiatric drug use. In order to reduce the said subjective contributions and thus increase the accuracy rate of diagnoses, various studies have been conducted in the prior art to assist in the diagnosis. In some of these studies, the physiological data of the patient were evaluated, while in others the physical activity of the patient was evaluated. The studies, in which the said discrete evaluations are made, are not at a level to assist in the diagnosis of hyperactivity disorder at the desired level.

Therefore, there is a need in the art for a method and system that can assist the clinician/physician by providing a diagnosis of hyperactivity disorder with high accuracy and with reduced subjective contributions.

Detailed Description of the Invention

The objective of the invention is to provide a system that assists in performing an objective diagnosis of hyperactivity, in which the relationship between the physiological, especially galvanic skin response and pulse, responses measured during the physical activity of an individual and the said physical activity is elicited in such a way as to indicate the hyperactivity condition of the individual and the resulting relationship is used.

An embodiment of the invention relates to a computer-implemented method for creating a hyperactivity disorder diagnostic tool, comprising the following steps:

- obtaining simultaneously measured galvanic skin response, pulse, and angular velocity and acceleration data of physical body movement of multiple individuals, who are diagnosed as having and not having hyperactivity disorder, and constitute a control group;

- training the obtained data with artificial intelligence algorithms and classifying the individuals matching with the data as having/not having hyperactivity disorder;

- determining the most accurate classification by testing the classifications with the control group; - if the determined classification is above a selected threshold value, creating a hyperactivity disorder diagnostic tool that includes an artificial intelligence algorithm in which the determined classification is made.

The control group includes multiple individuals/humans. Each individual was clinically diagnosed as having hyperactivity disorder or not having hyperactivity disorder. The clinical diagnosis of each individual is preferably made by multiple clinicians/physicians with a consensus diagnosis. For example, if all clinicians/physicians diagnosed an individual as having or not having hyperactivity disorder, the individual was assigned to the control group. Individuals for whom all clinicians/physicians did not make the same diagnosis were not assigned to the control group. Preferably, in the control group, the number of individuals diagnosed as having hyperactivity disorder is equal to the number of individuals diagnosed as not having hyperactivity disorder. Preferably, DSM-V psychiatric analyses and tests were used for the diagnosis of hyperactivity disorder.

In the computer-implemented method, the simultaneously measured galvanic skin response and pulse data [physiological data] as well as angular velocity and acceleration data of body movement [physical activity data] of each individual in the control group are obtained. It should be noted that angular velocity and acceleration data are vector quantities. The physiological and physical activity data of the individual can be measured with any sensor. Preferably, the galvanic skin response is measured with a sensor in contact with the body skin, and the angular velocity and acceleration data of the body movement are measured with an accelerometer and gyroscope positioned on the waist area of the individual. The physical activity data can also be generated by recording a video image of the individual and processing this video. However, since recording a video image of the individual would require the individual to stay in the environment of the imaging unit that will make the recording, it will be a secondary preference to generate physical activity data by recording a video image of the individual. The critical situation here is that physiological data and physical activity data are data measured simultaneously/synchronously. Thus, the relationship regarding the hyperactivity condition between the galvanic skin response and pulse responses and the physical activity can be elicited. When physiological data and physical activity data are not measured simultaneously/ are asynchronous, it will not be possible to discuss a relationship between the data. In this case, each discrete data, e.g. acceleration of body movement, is evaluated separately on its own, and pulse data are evaluated separately on their own. Preferably, physiological and physical activity data of each individual that were measured simultaneously for at least 24 hours are used.

A database may include clinical diagnostic data of each individual in the control group as having or not having hyperactivity disorder, as well as simultaneously measured physiological and physical activity data of each individual.

The physiological and physical activity data obtained for each individual in the control group are trained with artificial intelligence algorithms and the individuals from whom the data are obtained are classified as having/not having hyperactivity disorder. The said classifications are tested with the control group (diagnosis/diagnostic data of each individual in the control group as having and not having hyperactivity disorder) and the classification with the highest accuracy is determined. If the determined classification is above a selected threshold value, a hyperactivity disorder diagnostic tool is created that includes an artificial intelligence algorithm in which the determined classification is made. In an exemplary embodiment of the invention, the selected threshold value is a value in the range 90-100%, preferably 95%. The hyperactivity disorder diagnostic tool is an algorithm that can diagnose hyperactivity disorder with a high accuracy in line with the selected threshold value, by using the said simultaneously measured physiological and physical activity data detected by artificial intelligence. In one embodiment of the invention, physiological and physical activity data are subjected to a noise filtering and normalization process before being trained with artificial intelligence algorithms. In the normalization process, the amplitude value mean and standard deviation of the data are calculated. Then, if a data is outside a selected tolerance range of the calculated standard deviation value and/or the mean of the amplitude value, it is assumed that the said data has the effect of corrupting the dataset and this data is defined as erroneous data. If a data is determined as erroneous data after the normalization process, all physiological and physical activity data of the concerned individual whose data is erroneous are discarded and not used for artificial intelligence algorithms. The said individual is excluded from the control group. IIR (infinite impulse response), FIR (finite impulse response) or adaptive filters can be used for noise filtering. Thus, an improved initial data set (physiological and physical activity data, control group (diagnosis/diagnostic data for each individual in the control group as having and not having hyperactivity disorder)) is presented to the artificial intelligence algorithms.

In one embodiment of the invention, the artificial intelligence algorithms comprise machine learning algorithms. Machine learning algorithms may include at least one of the following algorithms: decision tree, discriminant analysis, logistic regression, Naive Bayes, support vector machine, nearest neighbor classifier, and artificial neural network.

In one embodiment of the invention, features for the simultaneously measured galvanic skin response, pulse and angular velocity and acceleration data of body movement are realized to create a training data for machine learning algorithms. Feature coefficient weights may change during algorithm learning.

The features created for angular velocity and acceleration data of body movement may include at least one of the following parameters: mean, root mean square (RMS), standard deviation, kurtosis, skewness, difference between maximum and minimum, variance, signal-to-noise ratio (SNR), signal-to-noise and distortion ratio (SINAD), root sum of squares level, average frequency, peak to RMS, power parameter. Feature coefficient weights may change during algorithm learning.

The features created for galvanic skin response data may include at least one of the following parameters: mean, standard deviation, maximum amplitude, minimum amplitude, kurtosis, skewness, delay and rise time. Feature coefficient weights may change during algorithm learning.

The features created for the pulse data may include at least one of the following parameters: pulse rate, peak, power, width, length, curvature, standard deviation, mean, signal-to-noise ratio. Feature coefficient weights may change during algorithm learning.

In one embodiment of the invention, the artificial intelligence algorithms comprise deep learning algorithms. In a variation of this embodiment, the deep learning algorithms comprise a convolutional neural network algorithm. This embodiment and the variation thereof may include Convolution-Relu, Convolution-Relu-Fully Connected, Convolution-Relu-Fully Connected-Softmax, Convolution-Relu-Fully Connected-Softmax-Classification layers. Layer coefficient weights may change during algorithm learning.

In one embodiment of the invention, a system is realized in which an algorithm is used that can diagnose hyperactivity disorder with a high accuracy in line with a selected threshold value, by using the said simultaneously measured physiological and physical activity data detected by the artificial intelligence in the said computer-implemented methods.

The system comprises a processing unit for classifying the individual as having/not having hyperactivity disorder, by processing the simultaneously measured galvanic skin response, pulse, angular velocity and acceleration data of body movement of an individual to be diagnosed, with a hyperactivity disorder diagnostic tool.

The system in one embodiment of the invention comprises at least one wearable apparatus having sensors for generating data of galvanic skin response, pulse and angular velocity and acceleration of body movement of the individual to be diagnosed.

In a variation of this embodiment, the wearable apparatus including the galvanic skin response and pulse sensors is in the form of a wristband. The wristband may comprise a galvanic skin response sensor with multiple non-invasive electrodes. The said multiple non-invasive electrodes can also act as a pulse sensor. In this embodiment, it may be preferable for the wristband to have a hook-and-loop fastener so that the wearable apparatus in the form of a wristband fits snugly around the individual's wrist. Thus, it is ensured that the sensors in the wristband are more snugly and continuously in contact with the skin of the individual.

A variation of this embodiment comprises a wearable apparatus having sensors in the form of an accelerometer for generating acceleration data of the body movement and in the form of a gyroscope for generating angular velocity data of the body movement. In this embodiment, the wearable apparatus may be in a form that can be worn on the individual's waist. Herein, the wearable apparatus has a fastener which allows the sensors that enable to generate angular velocity and acceleration data of the body movement to be positioned on the human waist area. An example of a fastener may be a belt or a band with a hook-and-loop fastener. The reliability and accuracy of the physical activity data are significantly enhanced by positioning the wearable apparatus, which will enable to generate physical activity data, on the waist area of the individual. However, the said apparatus does not need to come into contact with the body of the individual for obtaining data. Thus, it provides ease of use especially for juvenile individuals. Another important improvement achieved with the invention is that freedom of movement is provided to the individual and data accuracy is increased. In the methods of hyperactivity diagnosis in the prior art, the physical activity data are individually evaluated. Therefore, when obtaining physical activity data of the individual, the individual is asked to perform predetermined physical activities. Thus, studies were carried out based on matching physical activity data with the said determined physical activities. Due to the fact that the individuals are mostly juvenile and also due to the very definition of the potential disorder, it is often difficult for the individual to perform the determined physical activities. Therefore, the accuracy of the obtained physical activity data is reduced. In the invention of the present application, a study based on the relationship between physiological data and physical activity data is carried out. Based on this study, a diagnosis of a hyperactivity disorder disease is performed. Therefore, when obtaining physical activity data, it is not necessary for the individual to perform predetermined physical activities. Physical activity data can be obtained when the individuals move as they wish, without being required to perform predetermined physical activity. Hence, the individual is given freedom of movement and a dataset with high accuracy is obtained.

In the system of one embodiment of the invention, the wearable apparatus comprises a recording unit adapted to record data generated by the sensors in a non-volatile storage unit. In a variation of this embodiment, the non-volatile recording unit is in portable form. Examples of portable recording units are memory cards and USB drives.

The system in one embodiment of the invention comprises a wireless data transmission unit for wireless transmission of the data generated by sensors in wearable apparatus. The examples of wireless data transfer units may be Wifi, NFC, Bluetooth, Zigbee units. The system in one embodiment of the invention comprises an interface for displaying the classification as having/not having hyperactivity disorder performed for the individual. The examples of interfaces may be monitors and screens.

In the system of one embodiment of the invention, the wearable apparatus may inherently comprise a power supply and a timestamp generator module/time module for performing simultaneous measurement.

The system in an exemplary embodiment of the invention operates as follows. The wearable apparatus (physiological data obtaining apparatus) in the form of a wristband including galvanic skin response and pulse sensors is worn by the individual to be diagnosed. Similarly, the wearable apparatus (physical activity data obtaining apparatus) including accelerometer and gyroscope is worn on the waist area of the individual. Then, both obtaining apparatuses are enabled to record for a selected period of time, preferably 24 hours, and store the data on a memory card therein. Subsequently, the simultaneously recorded physiological data and physical activity data of the individual on the memory cards are read on a device containing a processing unit, (e.g. a computer, phone or a customized device) and obtained data is processed. As a result of the process, a diagnosis of hyperactivity disorder is realized for the individual. The realized diagnosis is displayed on an interface, e.g. on the monitor (e.g. to the clinician/physician).

Claims

CLAIMS A computer-implemented method for creating a hyperactivity disorder diagnostic tool, comprising the following steps:

- obtaining simultaneously measured galvanic skin response, pulse, and angular velocity and acceleration data of body movement of multiple individuals, who are diagnosed as having and not having hyperactivity disorder and constitute a control group;

- determining the most accurate classification by testing the classifications with the control group;

- if the determined classification is above a selected threshold value, creating a hyperactivity disorder diagnostic tool that includes the artificial intelligence algorithm in which the determined classification is made. A computer-implemented method for creating a hyperactivity disorder diagnostic tool according to claim 1, wherein the artificial intelligence algorithms comprise machine learning algorithms. A computer-implemented method for creating a hyperactivity disorder diagnostic tool according to claim 2, comprising machine learning algorithms including at least one of the following algorithms: decision tree, discriminant analysis, logistic regression, Naive Bayes, support vector machine, nearest neighbor classifier, and artificial neural network. A computer-implemented method for creating a hyperactivity disorder diagnostic tool according to any one of claim 2 or 3, comprising features realized for simultaneously measured galvanic skin response, pulse and angular velocity and acceleration data of body movement to create a training data for machine learning algorithms; wherein the features created for angular velocity and acceleration data of body movement include at least one of the following parameters: mean, root mean square, standard deviation, kurtosis, skewness, difference between maximum and minimum, variance, signal-to-noise ratio, signal-to-noise and distortion ratio, root sum of squares level, average frequency, peak to RMS, and power parameter; wherein the features created for galvanic skin response data include at least one of the following parameters: mean, standard deviation, maximum amplitude, minimum amplitude, kurtosis, skewness, delay and rise time; wherein the features created for the pulse data include at least one of the following parameters: pulse rate, peak, power, width, length, curvature, standard deviation, mean, and signal-to-noise ratio. A computer-implemented method for creating a hyperactivity disorder diagnostic tool according to any one of the preceding claims, wherein the artificial intelligence algorithms comprise machine learning algorithms. A computer-implemented method for creating a hyperactivity disorder diagnostic tool according to claim 5, comprising deep learning algorithms including a convolutional neural network algorithm. A computer-implemented method for creating a hyperactivity disorder diagnostic tool according to claim 6, comprising a convolutional neural network algorithm having Convolution-Relu, Convolution-Relu-Fully Connected, Convolution-Relu-Fully Connected- Softmax, Convolution-Relu- Fully Connected-Softmax-Classification layers. A system comprising a processing unit for classifying the individual as having/not having hyperactivity disorder, by processing the simultaneously measured galvanic skin response, pulse, and angular velocity and acceleration data of body movement of an individual to be diagnosed, with a hyperactivity disorder diagnostic tool created according to any one of the preceding claims. The system according to claim 8, comprising at least one wearable apparatus having sensors for generating data of galvanic skin response, pulse and angular velocity and acceleration of body movement of the individual to be diagnosed. The system according to claim 9, comprising a wearable apparatus in the form of a wristband including galvanic skin response and pulse sensors. The system according to claim 9 or 10, comprising a galvanic skin response sensor having multiple non-invasive electrodes. The system according to any one of claim 9 or 11, comprising a wearable apparatus having sensors in the form of an accelerometer for generating acceleration data of the body movement and in the form of a gyroscope for generating angular velocity data of the body movement. The system according to claim 12, comprising a wearable apparatus having a fastener which allows the sensors that enable to generate angular velocity and acceleration data of the body movement to be positioned on the human waist area. The system according to any one of claims 9 to 13, comprising wearable apparatus having a recording unit adapted to record data generated by the sensors in a non-volatile storage unit.

15. The system according to any one of claims 8 to 14, comprising an interface for displaying the classification as having/not having hyperactivity disorder performed for the individual.