CN115054250A

CN115054250A - Image overflow motion detection analysis method and system

Info

Publication number: CN115054250A
Application number: CN202210658421.8A
Authority: CN
Inventors: 罗煜
Original assignee: Beihang University
Current assignee: Beihang University
Priority date: 2022-06-10
Filing date: 2022-06-10
Publication date: 2022-09-16

Abstract

The invention relates to a method and a system for detecting and analyzing image overflow motion, which comprises the following steps: the electronic goniometer is arranged on at least one finger of the left hand and the right hand of a user respectively and is used for measuring the angular offset movement of the corresponding finger movement, including the angular offset movement times, the angular offset and the angular offset cumulant; the electronic video acquisition equipment is used for acquiring a video image of finger movement of a hand at a certain side, and the video image is used for identifying the angular offset movement times of the finger movement; the electroencephalogram acquisition equipment is used for acquiring electroencephalogram information of a user while the electronic goniometer and the electronic video acquisition equipment detect and acquire finger motion; and the data analysis equipment is used for synchronously analyzing the data collected by the electronic goniometer, the electronic video collection equipment and the electroencephalogram collection equipment so as to identify and quantify the mirror image overflow motion of the user. It is suitable for detecting, identifying, quantifying or evaluating attention deficit hyperactivity disorder.

Description

Mirror image overflow motion detection analysis method and system

Technical Field

The invention relates to a biological/physiological detection technology, belongs to the crossing field of bioinformatics and medical neurology, in particular to a technology for comprehensively analyzing by utilizing electroencephalogram, an electronic goniometer and computer vision, and is suitable for detecting, identifying, quantifying and evaluating mirror image overflow dyskinesia of attention deficit hyperactivity disorder.

Background

Attention-deficit hyperactivity disorder (ADHD) is one of the most common and persistent neurodevelopmental disorders in childhood, and ADHD patients are characterized by age-inappropriate hyperactivity and/or impulsivity, as well as Attention deficit.

Currently, ADHD is clinically detected mainly according to imagination data, namely observation and inspection of behavior symptoms, behavior rating scales, neuropsychological tests and the like, which can provide reference for detection. In addition, attention deficit and hyperactivity/impulsivity symptoms are non-specific symptoms, and are seen in various diseases such as schizophrenia, anxiety, mood disorders, and the like. Thus, despite the extensive research on ADHD, ADHD symptoms lack specificity, objective signs and laboratory data that can aid in detection are small, and today, they are one of the more difficult to detect mental disorders in children. There is a need in the art to improve the accuracy of ADHD detection by more objective means.

Disclosure of Invention

The applicant found through research that: ADHD patients often have nervous system soft body disease and can be used for auxiliary detection of ADHD. ADHD neurological soft-bodied can manifest as excessive image-overflow dyskinesia. Hyperkinetic movement disorders, which are defined as involuntary and involuntary movements of the contralateral muscles accompanied by unilateral body voluntary movements. For example, when the fingers are tapped in a left-handed autonomous sequence, the unintended motion produced by the right hand is a mirror image spill. Compared with the neurobiological characteristics behind the ADHD more complex behavior phenomenon, the ADHD motor system features relatively concise and significant, so the neural basis of image overflow can deepen the overall understanding of the neural mechanism of ADHD, can improve the understanding of ADHD dyskinesia, and help to design biomarkers for classified recognition to assist the clinical detection of ADHD. As dyskinesias usually occur earlier than cognitive impairment, ADHD image overflow may also be used as a biomarker for early identification of ADHD.

Aiming at the problems that the ADHD is difficult to carry out accurate objective detection and the neural mechanism is unclear at present, the invention comprehensively uses an electronic goniometer and an electronic video to quantify the mirror image overflow movement, and simultaneously applies machine learning methods such as behavioral detection, tracing imaging, neural oscillation, a support vector machine, random forest and the like to carry out quantitative and objective screening and identification on ADHD patients and research the neural mechanism of ADHD mirror image overflow.

The invention provides a system for detecting and analyzing image overflow motion, which comprises: the electronic goniometer is arranged on at least one finger of the left hand and the right hand of a user respectively and is used for measuring the angular offset movement of the corresponding finger movement, including the angular offset movement times, the angular offset and the angular offset cumulant; the electronic video acquisition equipment is used for acquiring a video image of finger movement of a hand at a certain side, and the video image is used for identifying the angular offset movement times of the finger movement; and the brain wave acquisition equipment is used for acquiring brain wave information of the user while the electronic goniometer and the electronic video acquisition equipment detect and acquire finger motion.

In the technical scheme, the system further comprises data analysis equipment, wherein the data analysis equipment is used for synchronously analyzing the data collected by the electronic goniometer, the electronic video collection equipment and the electroencephalogram collection equipment so as to identify and quantify the mirror image overflow movement of the user.

In the above technical solution, the system is used for detecting, identifying, quantifying or evaluating the image overflow dyskinesia of attention deficit hyperactivity disorder.

The invention also provides a mirror image overflow motion detection method, which is suitable for quantifying the mirror image overflow motion of attention deficit hyperactivity disorder and comprises the following steps: when a tested crowd carries out a task of sequentially knocking fingers on one side, an electronic goniometer and electronic video acquisition equipment are applied to measure and quantify the mirror image overflow motion of the fingers on the other side; meanwhile, electroencephalogram signals of a tested person group when a single-side sequential finger knocking task is executed are collected by electroencephalogram collecting equipment; preprocessing the acquired electroencephalogram signals to obtain electroencephalogram data; the preprocessing comprises one or more of down-sampling, re-referencing, high-low pass filtering, segmentation, baseline correction, artifact identification and removal, bad track replacement, bad trial times identification and deletion and time window selection of interest; performing source imaging, ERD analysis of a source layer and brain network analysis on the preprocessed electroencephalogram data so as to research brain activation, a neuron group desynchronization state and interaction among neuron groups; and carrying out independent sample t test, mixed repeated measurement analysis of variance, adjustment analysis and Pearson correlation analysis on the electroencephalogram data after the behaviours and the processing.

The invention achieves the following technical effects:

(1) the ADHD image overflow detection model based on the electronic goniometer and the electronic video can measure and quantify the ADHD image overflow movement disorder, so that the ADHD detection precision is improved;

(2) the joint biomarker based on image overflow, EEG source imaging, ERD and brain network, which is designed by the invention, can accurately detect the image overflow dyskinesia, thereby being capable of better distinguishing ADHD patients and healthy control groups, being more objective and reliable than the currently used scales, not being influenced by subjective judgment, being used for the early screening of ADHD, and improving the accuracy, sensitivity and specificity of the ADHD classification and identification.

(3) Compared with the conventional ADHD detection method, the method for detecting the mirror image overflow movement disorder is relatively more economical and convenient. The 47-lead electroencephalogram acquisition equipment used in the present invention is less expensive than nuclear and magnetoencephalograms.

(4) The neural oscillation, brain network, regulation analysis and mixed repeated measurement variance model based on the source layer further determines whether the ADHD group and the healthy control group have significant difference and reasons for difference generation, and improves the understanding of the neural mechanism of ADHD mirror image overflow movement disorder.

Drawings

Fig. 1 is a flow chart of electroencephalogram data processing of the present invention.

Fig. 2 is a graph of the results of a t-test on independent samples of a mirror overflow of the Attention Deficit Hyperactivity Disorder (ADHD) group and a healthy control group (TD control).

FIG. 3 is a graph of the source layer surface α ERD nerve oscillations for the ADHD group and the TD group.

Fig. 4 is a graph of the quantized difference in the sensorimotor region α ERD for the ADHD group and the TD group.

Fig. 5 is a pearson correlation of lateral sensorimotor region α ERD with motor score for the ADHD group and the TD group.

FIG. 6 is a plot of the correlation of the lateral sensorimotor region α ERD with ADHD hyperactivity/impulsive symptoms Pearson for the ADHD group and the TD group.

FIG. 7 is a plot of the lateral sensorimotor region α ERD versus the behavioral image spill Pearson for the ADHD group and the TD group.

FIG. 8 is a graph of the correlation of ADHD group and TD group inter-hemispheric brain network connectivity to ADHD attention deficit symptom Pearson.

Detailed Description

The invention is described in further detail below with reference to the accompanying drawings and detailed description, in order to facilitate the understanding and implementation of the invention by those skilled in the art.

Mirror movements (minor movements) refer to involuntary movements in which one limb is voluntarily moved while the other limb is almost simultaneously duplicated.

When moving, the alpha oscillations in the brain waves are suppressed in the sensorimotor cortex. The corresponding amplitude drop represents a potential change in the synchronization status of the neuron population, i.e., Event-Related Desynchronization (ERD). Alpha rhythm is generally thought to reflect inhibition, and alpha ERD indicates activation and is associated with metabolic activation of brain regions. In conventional functional magnetic resonance imaging (fMRI) and other studies on image overflow, it is currently impossible to quantify behavioral image overflow while acquiring image data due to magnetic force limitations in an imaging technique such as fMRI, and electroencephalogram (EEG) can overcome this limitation. In addition, the electroencephalogram equipment is relatively cheap, convenient to operate, fast and widely available, is one of the preferable methods for researching children brain cortex activity and brain inter-regional information interaction, can reduce the limitation that the tested body needs to be kept motionless in the researches such as fMRI and magnetoencephalogram, and is suitable for pediatric people. The relationship between the brain and the behaviours can be more accurately explored through the detection of ADHD mirror image overflow quantification and the research based on electroencephalogram source imaging and neural oscillation ERD. The ERD analysis at the source level can characterize the dynamic activity change of the brain and the synchronous fluctuation in the neural activity caused by diseases with higher time resolution and spatial resolution, is helpful to improve the understanding of the neural mechanism of ADHD image overflow movement disorder and is also helpful to design biomarkers for ADHD screening.

The invention comprehensively uses an electronic goniometer and an electronic video to quantify the mirror image overflow movement, and simultaneously applies machine learning methods such as behavioral detection, tracing imaging, neural oscillation and support vector machine, random forest and the like to quantitatively and objectively screen and identify ADHD patients and research the neural mechanism of ADHD mirror image overflow.

The invention provides a device and a method for detecting and analyzing mirror image overflow based on an electronic goniometer and electroencephalogram source imaging, which comprise the technical improvements of the aspects of behavioural mirror image overflow quantification, electroencephalogram acquisition and pretreatment, source imaging analysis, source level neural oscillation analysis, source level brain network analysis, statistical models, machine learning models and the like.

1. Quantification of behavioural image overflow

The present invention uses an electronic goniometer and electronic video to measure and quantify the condition of image overflow. Wherein sensors of the electronic goniometer are respectively fixed on the knuckles of the index finger and the ring finger of the left hand and the right hand of the user so as to respectively capture the flexion and extension of the index finger and the ring finger of the left hand and the right hand. Meanwhile, the electronic video acquisition equipment, such as a camera, a camera and the like, is aligned to the left hand and the right hand of the user, so that the video acquisition equipment can clearly and completely capture the motion conditions of all fingers and palms of the user.

In particular, the present invention may use electronic video to quantify phase spillover of finger movements. Where single phase overflow motion refers to any bi-directional motion of a non-tapping hand occurring in 1 second or less, primarily an extension-flexion or flexion-extension motion, but may also include side-to-side motion, i.e., left-to-right or right-to-left motion, or wobble. During the playing process of playing back the collected electronic video, the number of phase overflow actions of the inactive knocker during the LHFT and RHFT can be counted by the scorer; or detecting the number of phase overflow actions of the inactive knocker during the LHFT and RHFT in real time by adopting a machine vision mode. Specifically, a video camera with the resolution of 1.07 megapixels can be used for recording video data, the video camera is installed on a tripod to keep stable, and images are enlarged as much as possible, so that the hands of a shooting object are stable and clear; in a Player, such as a Windows Media Player, the video recording is viewed at 250 times magnification during playback, facilitating the evaluator to observe finger motion details, facilitating the evaluator to calculate the amount of phase-overflow motion of the non-task hand during LHFT and RHFT.

More specifically, the present invention may also simultaneously use an electronic goniometer to quantify the total spill over of the mirrored spill over motion. Total overflow is defined as the magnitude of the fingers' deviations from their resting baseline position, including persistent postural (rigidity) and phase overflow. The electronic goniometer measures the angular deviation of the forefinger and the metacarpophalangeal joints (MCP), and the goniometer calibrates 0 degrees and 45 degrees during the experiment of each subject, i.e., each hand of the subject is firstly placed on the plane of a wood block to calibrate 0 degrees, and then each hand is placed on the edge of the wood block cut to have an angle of 45 degrees, so that the fingers of the subject can form an angle of 45 degrees with the MCP joints. The measurement data of the electronic goniometer are then sampled at 100Hz during the acquisition, preferably using the Biopac MP100 system with the Biopac acquknowledge3.9.1v software. The total overflow can be calculated in an off-line manner, with the total angle of angular displacement of the finger for a non-tapping hand baseline reading as the total overflow. The data of each finger knocking block corresponds to four signal acquisition channels and respectively represents the angle positions of the left index finger, the right index finger and the ring finger relative to the MCP joint. Preferably, each signal acquisition channel correspondence is stored as a data file that is aligned in a time-synchronized manner with the corresponding video acquisition file such that the first finger tap (index finger-thumb) in the video corresponds to the peak of the first tap in the index finger signal channel record. The time of baseline data, the time of the first tap, and the time to complete all taps in the data and video files are recorded with the synchronized video as a reference. To remove high frequency artifacts, smoothing may be performed using a moving window of 10 data points; simultaneously, further analyzing and calculating the average value of the baseline data corresponding to the index finger and the ring finger of the non-knocking hand; the total overflow was determined by subtracting the baseline average from the total angular displacement values generated by the index and ring fingers of the non-tapping hand. In order to also take into account finger flexion (positive deflection) and extension (negative deflection), the absolute value of the angular displacement is used for the calculation.

Those skilled in the art will appreciate that the sensors of the electronic goniometer may be placed on more fingers of the user's left and right hand, or on other fingers. The invention only installs the sensors on the index finger and the ring finger of the left hand and the right hand of the user, which not only can accurately reflect the mirror image overflow condition of the whole hand, but also can improve the comfort and the flexibility of the tested user.

Before the experiment, the sensor is calibrated at 0 ° and 45 ° to accurately reflect the bending movement of the finger, preferably using ACQ Knowledge software. And simultaneously, the recorded finger movement condition of the user is checked and confirmed by observing images and videos obtained by the electronic video acquisition equipment through naked eyes.

The user performing the test sequentially performs the task of tapping fingers on one hand as required, and detects the motion of the non-tapped hand on the other hand to evaluate the image overflow. For example, a user performing a test in an experiment taps each finger in turn with a self-defined rhythm in a fixed order, such as index finger-middle finger-ring finger-little finger, and alternates left-hand (non-dominant) finger tapping (LHFT) and right-hand (dominant) finger tapping (RHFT). The start signal "Go" is sent out through the computer monitor. Preferably, the experiment is divided into 5 modules (blocks), each module containing 20 trials (dials), each trial lasting 6 seconds, with a rest/baseline time of 1 second before the start of each trial, the LHFT and RHFT experiments alternating between different modules. The first trial time adopts LHFT or RHFT random distribution, and the total number of the LHFT and the RHFT trial times is equal. The ethological image overflow calculation method recorded by the electronic goniometer is obtained by averaging the accumulated angular deflection of all tested non-knockers in the LHFT and RHFT processes.

Y _{totaloverflow} ＝Y _LHFToverflow +Y _RHFToverflow

In the above formula: y is _LHFToverflow And Y _RHFToverflow The average mirror image overflow number, Y, respectively obtained by quantification during the autonomous movement of the left hand and the right hand _{totaloverflow} Is the total mirror overflow average for left and right hand movements; i is the number of trials, N is the total number of trials;

and

is the image overflow number measured by the left and right hand ring finger of each trial,

and

is the number of mirror overflows measured by the index finger of the left and right hand of each trial.

In order to avoid influence of handedness habits on test results, the testers select right handedness.

2. Electroencephalogram acquisition and preprocessing

Acquisition of EEG data was performed using 47 channels, preferably using the WaveGuard electrode cap system and asa-lab amplifier (Advanced Neuro Technologies, Netherlands); all electrode placements are in accordance with the international 10-20 system electrode placement method. The sampling rate of the EEG data is 1024Hz, down-sampled to 512 Hz. All electrode impedances remain below 15k omega.

Raw electroencephalographic data is contaminated with artifacts from many non-physiological (power frequency, poor electrode contact, bad electrodes, etc.) and physiological (heart pulse, muscle activity, perspiration, movement, etc.) sources. These artifact components must be carefully identified and removed or excluded from further analysis, i.e., preprocessing of the EEG data is required.

The pretreatment of the invention comprises the following steps:

(1) the average of all electrode data collected was taken as the Re-reference (Re-reference).

(2) The re-reference values are filtered using a 1Hz high pass Finite Impulse Response filter and an 80Hz two-way least squares Finite Impulse Response (FIR) low pass filter.

(3) Segmentation and baseline correction. The successive EEG data is segmented, starting with the start time of the "Go" signal for each trial, i.e. the subject starts tapping the finger as soon as he sees the "Go" signal appearing on the screen. The continuous EEG data is segmented into segments (epochs) of length from-1 second to 5 seconds. Then, baseline correction (baseline correction) is performed to correct for the drift of the signal.

(4) And removing the artifacts. The use of a trap to remove the 60Hz power frequency interference followed by the use of a blind source separation algorithm to identify and remove artifacts (bad three removal) that typically reflect muscle activity, eye movement and motion artifacts associated with blinking.

The method can further comprise replacing channels of bad data by interpolation, visually checking and manually deleting bad test times.

Preferably, each segment of EEG data is selected for analysis with a time window of 1.5 seconds of data, for example from 1.5 to 3 seconds after the onset of tapping finger movement, since mirroring overflow movement may occur during this period. It is worth noting that the present invention collects EEG signals while quantifying image spillover, and what is recorded is task-state EEG data that is directly related to behavioral data, unlike MRI studies where behavioral spill-over motion quantification is done separately from imaging data collection, which allows for better data consistency.

3. Source imaging analysis

With the advancement of digital technology, electroencephalographic analysis has evolved from purely visual detection of amplitude and frequency changes over time to a comprehensive exploration of the spatiotemporal characteristics of the recorded signals. The currently existing electroencephalograms are transformed into a true neuroimaging modality, i.e. a high-density electroencephalographic system combines precise information of the head anatomy and complex source localization algorithms. According to poisson's equation, the potential difference between electrodes placed at different scalp locations is caused by current propagation due to synchronized post-synaptic potentials of the pyramidal neurons of the head. However, due to the high electrical conductivity of the skull, the current is strongly attenuated by the skull, and the propagation is not uniform. By building a suitable head model and building a lead field, the inverse problem is then solved, i.e. the cranial source that produces a given scalp electroencephalographic potential measurement is determined. Solving this inverse problem is a fundamental challenge, since a large number of different source distributions can produce the same potential field on the scalp. The development of brain source imaging technology has provided a more exciting option for localizing brain sources from the scalp electroencephalogram signals and has largely replaced dipole source localization methods. These so-called distributed source localization methods do not make a priori assumptions about the number of dipoles.

The present invention uses the ICBM152 template to add EEG spatial locations (ICBM, International Consortium for Brain imaging) in EEG source imaging. To describe the propagation of the electric field from the cortical surface to the scalp, a forward model was constructed using OpenMEEG as the boundary meta-model for each test. This symmetric boundary meta-model uses three real layers, including the scalp, the inner skull, and the outer skull. The electrode locations and boundaries are jointly recorded to match four anatomical reference points, namely the apex, the alar, the left and right anterior ear points. The noise covariance matrix is then calculated from the baselines of the individual trials. Next, an inverse model is calculated using the sLORETA algorithm.

The source imaging algorithm pseudo-code is as follows:

algorithm time-locking non-phase-locking EEG brain network construction algorithm

Description of the drawings: modeling the source, the conductivity and the sensor by utilizing an electrostatic law, a Biao law, a Saval law, a symmetric boundary element method in OpenMEEG and the like to calculate a forward model; the inverse problem in the traceable localization is then solved using standardized low-resolution brain electromagnetic tomography (sLORETA) by estimating the potential sources under the assumption that neighboring voxels should have maximally similar electrical activity. Some of the algorithm details can be found in OpenMEEG, open resources software for qualitative bioelectromagnetics; Pascal-Marqui R.D.Standard low-resolution bridge electromagnetic mapping (sLORETA): technical details [ J ]. Methods Find Exp Clin Pharmacol,2002.24(Suppl D):5-12.

Inputting: EEG sensor array X _raw ∈R ^N×T It is a T-second EEG time series with N channels and T sampling indices, with a sampling rate of 2 ^f Hz, wherein

And (3) outputting: a brain network G ═ { D, E, Q }, where D ═ M × 1, the vector representing nodes or brain regions in the brain network; e-mxm, which represents the relationship between the edges or brain regions of the brain network; q is M × P, and the matrix represents a region feature or a time series. In the above formula, M represents the number of nodes, and P represents a time series of any one node.

1 initializing parameters

2 for i (i ═ 1, …, s) do; wherein s represents the number of trials

EEG pretreatment

4 obtaining a preprocessed EEG segment X with N channels and L trial runs

5:for l(l＝1,…,L)do

6:

In the formula, σ represents the conductance field,

denotes the partial derivative operator of the vector, V denotes the potential, J ^p Is a distribution of dipole sources within the domain, which, when considering brain activation, represents the average post-synaptic current in pyramidal cortical neurons. Boundary conditions must be applied, typically normal currents at domain boundaries are controlled

7:

Boundary condition

8:

Wherein B represents a (synthetic) magnetic field, u ₀ Representing the magnetic constant, r the field position and r' the source position. The above equations are used to solve a positive problem, which includes simulating V and/or B when σ, Jp, and boundary current j are known

9:

10:

11:

In the above formula, G represents a local lead field matrix, which defines the contribution of a dipole source at a certain position to a measurement X, and C is a covariance matrix calculated from measurement data samples. S (t) represents the source, O (t) represents the signal of the baseline, W represents the spatial filter of the source, and the above equations are used to solve the inverse problem, respectively.

12:end

13:end

4. Source-level neural oscillation analysis

After source imaging analysis, the EEG signal is time-frequency decomposed using the Morlet wavelet algorithm. For the gaussian kernel of the Morlet wavelet, the parameters of the mother wavelet are set to 3 second half width with a center frequency of 1 Hz. And finally calculating the ERD. The calculation formula for ERD is as follows:

where x is the normalized data and μ is the mean of the baseline measurements. After power spectrum analysis, the alpha frequency (8-12Hz) component is of major concern, which is also relevant for motion control. The boundaries of the sensory motor region were selected by the Broadmann map, including the primary somatosensory cortex, primary motor cortex, auxiliary motor zone and dorsolateral motor anterior cortex. The calculation of alpha rhythm was consistent between the experimental and control groups and between left and right hand movements.

5. Brain network analysis at the source level

After neuro-oscillation analysis of the source layer surface, inter-hemispheric functional connections of the sensory-motor region were calculated from the source level data using amplitude envelope correlation. Functional connectivity is calculated from the mean amplitude envelope correlation of the region of interest (ROI), whose boundaries are defined by the Broadmann atlas. The ROI includes a somatosensory cortex, a primary motor cortex, an auxiliary motor region, and a dorsal anterior motor cortex. The ROI was chosen because previous studies showed that these regions were involved in the execution of motion. With regard to brain network connection calculation, the "task" functional connection is first calculated using data from 1.5 seconds to 3 seconds after the start of the movement, since image overflow may occur during this time window. The "baseline" functional connection was then calculated using data from-1 to 0 seconds (baseline period) in each trial. The "baseline" is finally subtracted from the "task" functional connectivity, thereby excluding the effects of other factors besides image overflow. Such connectivity analysis was performed for each test.

6. Clinical examination grouping

ADHD was detected using structured or semi-structured parental interviews including "schedules of Affective Disorders and Schizophrenia in Children" (Kiddie Schedule for effective Disorders and schizohrnia, the Kiddie Schedule for effective Disorders and schizohrnia, K-SADS) and "interviews of detection of Children and Adolescents" (Diagnostic Interview for childhood and Adolescents-IV, DICA-IV). The ADHD Rating Scale (ADHD-RS) and the Conner Rating Scale (Conner's Rating Scale-Revised-Version 3, CPRS-R) were used to confirm the detection and to quantify the severity of ADHD symptoms. All tested were evaluated by the Edinburgh hands-free scale as Right handedness.

The inclusion criteria for the ADHD group were (1) tested for evaluation by K-SADS or DICA-IV, meeting the ADHD detection criteria, and (2) the CPRS-R evaluation item had a T score of not less than 60 points for attention deficit or hyperactivity/impulsivity symptoms; the ADHD-RS assessment entries had an attention deficit or hyperactivity/impulsivity symptom score of at least 6 (9 total) with a score of 2 or 3 (indicating frequent or very frequent occurrence of symptoms). The ADHD subtype was determined by a combination of interviews and symptoms recognized by the Conners and DuPaul scoring scale. In the ADHD group, 68% met the criteria for the combined subtype and 32% met the criteria for the attention-deficit subtype. ADHD children (52%) taking stimulant drugs required discontinuation of the drug the day before the experiment. In addition, children taking psychotropic drugs other than stimulant drugs may still be experimented with. Only two children with ADHD took SSRIs psychotropic drugs during the experiment. The exclusion criteria were: all children with a full blown Chinesian of less than 80 points were excluded from the study, excluding patients with other history of neurological or psychiatric disorders. The General Ability Index (GAI) test showed that ADHD children had a lower tendency for their IQ than the TD control (mean IQ for ADHD group 112.48 + -11.41, TD group 118.72 + -11.68, p value of 0.062, Cohen's d of 0.54). GAI is the comprehensive ability score of the Wechsler intellectual Scale for childrens-IV, WISC-IV, for checking Children's wisdom quotient while minimizing the impact on processing speed and work memory tasks.

7. Statistical model

Statistical models were constructed for behavioral and statistical analysis of EEG data. Independent sample t-tests were performed between ADHD and TD groups to test for potential differences in demographic, behavioral image overflow, and clinical data (Conners ADHD attention deficit and Conners ADHD hyperactivity/impulsivity symptoms). All t-tests were two-sided, α is 0.05. Subsequently, potential confusion for age, gender, and GAI were corrected when statistical analysis between groups was performed.

To explore whether the ERD difference was mainly due to tapping the left hand, all were tested as right handedness. The main hypothesis testing model constructed by the invention is a mixed repeated measurement analysis of variance model, the factors in the group are 'hemisphere' (left hemisphere and right hemisphere) and 'chirality' (LHFT and RHFT), the factors between the groups are 'group' (ADHD group and healthy control group), and age, sex and GAI are added as covariates. Furthermore, the interaction effect of chiral x groups is important in hypothesis testing. In addition to the interaction effect of chirality x group, the interaction effect of group x hemisphere and group x chirality x hemisphere were further analyzed. Then, (1) whether there is interaction of chirality (LHFT/RHFT) x group on the behavioural mirror overflow was explored; (2) using an accommodation analysis model, explore whether the brain (ERD) -behavior (mirror spill) relationship is similar or different in LHFT and RHFT in ADHD patients compared to TD controls; (3) using Pearson correlation analysis, it was investigated whether there was a group chiral interaction between ERD and ADHD clinical symptom severity. The results are illustrated in the accompanying figures.

Among these, fig. 2 is used to illustrate that there is a significant inter-group difference in mirror image spill during both LHFT and RHFT, with the difference being more significant in the non-dominant hand voluntary movements (LFHT) between ADHD children and TD controls. FIG. 3 is used to illustrate that the sensory motor cortex has significant activation α ERD when performing the tapping finger task. In the case of LHFT, the sensory-motor cortex is activated more strongly in the right hemisphere than in the left hemisphere. FIG. 4 is a graph illustrating the differences between the groups for the ADHD and the α ERD for the TD group, the differences between the α ERD for the contralateral and ipsilateral primary motor cortex (M1), primary somatosensory cortex (S1), and the motor-assisted (SMA) and premotor cortex at LHFT, and the differences between the α ERD for the contralateral and ipsilateral sensory motor regions at RHF. In the figure, p is less than 0.05, p is less than 0.01, and n.s. indicates no significant difference.

Fig. 5 is a graph illustrating the correlation of the alpha ERD of the contralateral primary motor cortex (M1), primary somatosensory cortex (S1), and motor-assisted zone (SMA) and anterior motor cortex of the ADHD and TD groups with nervous system Soft Signs scale (PANESS) scores during LHFT.

Fig. 6 is a graph illustrating the correlation of α ERD to Conners ADHD hyperactivity/impulsivity (H/I) symptoms during LHFT for the contralateral primary motor cortex (M1), primary somatosensory cortex (S1), and motor-assisted zone (SMA) and pre-motor cortex of ADHD and TD groups.

Fig. 7 is a graph illustrating the correlation between the alpha ERD and behavioural image spillover during LHFT of the contralateral primary motor cortex (M1), primary somatosensory cortex (S1) and motor supplementary zone (SMA) and anterior motor cortex of the ADHD and TD groups.

FIG. 8 is a graph illustrating the correlation between the functional connectivity between the sensory-motor regions of the left and right hemispheres of the ADHD group and TD group (vertical axis) and the Conners ADHD attention deficit symptoms (horizontal axis).

8. Machine learning model

In order to distinguish the ADHD group from the healthy control group, the present invention proposes a support vector machine, random forest (random forest) and long-short term memory machine learning model based on the source-level EEG features and behavioral combination features. The support vector machine is a highly popular supervised learning model and has unique advantages in the aspects of solving small samples, nonlinearity and high-dimensional pattern recognition. The support vector machine is suitable for the research of smaller sample capacity at present. Other machine learning methods, such as neural networks, tend to require large sample volumes, which may otherwise cause overfitting problems. Furthermore, support vector machines can efficiently process complex, non-linear and real-world data, while other algorithms (e.g., linear discriminant analysis) can only be applied to groups separated by linear combinations of features. The support vector machine is also a method for detecting complex and subtle differences among groups, and has been widely applied to the detection research of neurodevelopmental disorder. Therefore, the present invention uses a support vector machine for classification. In addition, a random forest algorithm is used for classification and comparison with a support vector machine. The random forest model is a robust, relatively simple and easy-to-interpret classifier, has discriminability, is suitable for small samples, and can capture the nonlinear relation between input features. In the aspect of feature extraction, the behavioral features come from demographic information and behavioral image overflow. In addition to behavioral features, the ERD and brain network connection parameters at the EEG source level together constitute a joint feature. In terms of classification, 80% of the data constitutes the training set, the remaining 20% is the test set, and then cross-validation is performed. Long-term memory is a special recurrent neural network structure used in the field of deep learning that can link previous information to the current task. Long-term and short-term memory is used for detection and prediction of diseases such as alzheimer's disease. The invention uses long-time memory to carry out 5 times of cross validation, the learning rate is 0.001, and the training times are 100. The average classification accuracy was calculated and compared to the accuracy calculated using only behavioral features and joint features. The result shows that the joint characteristics of the behavioral image overflow, the source imaging, the neural oscillation of the source level and the brain network connection can be used as biomarkers to objectively predict Attention Deficit Hyperactivity Disorder (ADHD).

The results of the experiments after classifying the ADHD group and the healthy control group using the machine learning model described above are shown in the following table.

Specifically, the invention further provides an Attention Deficit Hyperactivity Disorder (ADHD) image overflow quantification method, an ADHD objective detection model and an application method of the model based on the above technical solution. The method quantifies the image overflow movement by combining behavioural and neuroelectrophysiological experiments, applying an electronic goniometer and an electronic video, extracting combined characteristics by applying analysis methods such as source imaging, alpha ERD neural oscillation, brain network and machine learning, and the like, and classifying ADHD patients and healthy control groups, thereby improving the classification accuracy, sensitivity and specificity.

In order to achieve the purpose, the invention provides the following technical scheme:

the first step is as follows: when a tested crowd carries out a task of sequentially knocking fingers on one side, an electronic goniometer and an electronic video are applied to measure and quantify the mirror image overflow motion of the fingers on the other side; meanwhile, electroencephalogram signals of a tested person group are collected when a unilateral sequential finger-tapping task is executed by applying an electroencephalogram (EEG) technology;

the second step is that: preprocessing the acquired electroencephalogram signals to obtain electroencephalogram data; the preprocessing comprises preprocessing analysis such as down sampling, re-referencing, high-low pass filtering, segmentation, baseline correction, artifact identification and removal, bad track replacement, bad track identification and deletion, time window selection of interest and the like

The third step: performing source imaging, ERD analysis of a source layer and brain network analysis on the preprocessed electroencephalogram data so as to research brain activation, neuron group desynchronization state and interaction among neuron groups;

the fourth step: carrying out independent sample t test, mixed repeated measurement analysis of variance, adjustment analysis and Pearson correlation analysis on the behavioristic and processed electroencephalogram data;

by performing clinical tests on the test population, it can be grouped into Attention Deficit Hyperactivity Disorder (ADHD) groups and healthy control (TD) groups, so that differences between the ADHD and TD groups and relationships between the brain and the ethology can be further analyzed and studied.

There is further provided in accordance with yet another preferred embodiment of the present invention a system for detecting attention deficit hyperactivity disorder based on mirror-image overflow motion, including: the method comprises the following steps:

the electronic goniometer is arranged on at least one finger of the left hand and the right hand of a user respectively and is used for measuring the angular offset movement of the corresponding finger movement, including the angular offset movement times, the angular offset and the angular offset cumulant; the angular offset refers to the angular offset amplitude of the relative motion between a certain finger and a certain metacarpophalangeal joint, and includes but is not limited to: the angle offset motion of the metacarpophalangeal joints of the left index finger and the left ring finger, the angle offset motion of the metacarpophalangeal joints of the left ring finger and the left ring finger, the angle offset motion of the metacarpophalangeal joints of the right index finger and the right ring finger, and the angle offset motion of the metacarpophalangeal joints of the right ring finger and the right ring finger;

the electronic video acquisition equipment is used for acquiring a video image of finger movement of a hand at a certain side, and the video image is used for identifying the angular offset movement times of the finger movement;

the electroencephalogram acquisition equipment is used for acquiring electroencephalogram information of a user while the electronic goniometer and the electronic video acquisition equipment detect and acquire finger motion;

the data analysis unit is used for synchronously analyzing the data acquired by the electronic goniometer, the electronic video acquisition equipment and the electroencephalogram acquisition equipment so as to identify and quantify the mirror image overflow motion of the user as the behavioural characteristic;

and the attention deficit hyperactivity disorder recognition unit inputs the EEG characteristics and the ethological characteristics into a support vector machine, a random forest model or a long-time and short-time memory model for classification and outputs a recognition result.

To achieve the above object, according to another aspect of the present application, there is also provided a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in the storage of the decentralized personal health information described above when executing the computer program.

The processor may be a Central Processing Unit (CPU). The Processor may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or a combination thereof.

The memory, which is a non-transitory computer readable storage medium, may be used to store non-transitory software programs, non-transitory computer executable programs, and units, such as the corresponding program units in the above-described method embodiments of the present invention. The processor executes various functional applications of the processor and the processing of the work data by executing the non-transitory software programs, instructions and modules stored in the memory, that is, the method in the above method embodiment is realized.

The memory may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created by the processor, and the like. Further, the memory may include high speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory located remotely from the processor, and such remote memory may be coupled to the processor via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The one or more units are stored in the memory and when executed by the processor perform the method of the above embodiments.

The specific details of the computer device may be understood by referring to the corresponding related descriptions and effects in the above embodiments, and are not described herein again.

To achieve the above object, according to another aspect of the present application, there is also provided a computer-readable storage medium storing a computer program which, when executed in a computer processor, performs the steps in the storage of the decentralized personal health information described above. Those skilled in the art will appreciate that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can include the processes of the embodiments of the methods described above when executed. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk (Hard Disk Drive, abbreviated as HDD) or a Solid State Drive (SSD), etc.; the storage medium may also comprise a combination of memories of the kind described above.

It will be apparent to those skilled in the art that the modules or steps of the present invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of multiple computing devices, and they may alternatively be implemented by program code executable by a computing device, such that they may be stored in a storage device and executed by a computing device, or fabricated separately as individual integrated circuit modules, or fabricated as a single integrated circuit module from multiple modules or steps. Thus, the present invention is not limited to any specific combination of hardware and software.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the concept and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A system for detecting and analyzing image overflow motion, comprising: comprises that

The electronic goniometer is arranged on at least one finger of the left hand and the right hand of a user respectively and is used for measuring the angular offset movement of the corresponding finger movement, including the angular offset movement times, the angular offset and the angular offset cumulant; the angular offset refers to the angular offset amplitude of a relative motion between a certain finger and a certain metacarpophalangeal joint, and includes but is not limited to: the angle offset motion of the metacarpophalangeal joints of the left index finger and the left ring finger, the angle offset motion of the metacarpophalangeal joints of the left ring finger and the left ring finger, the angle offset motion of the metacarpophalangeal joints of the right index finger and the right ring finger, and the angle offset motion of the metacarpophalangeal joints of the right ring finger and the right ring finger;

and the brain wave acquisition equipment is used for acquiring brain wave information of the user while the electronic goniometer and the electronic video acquisition equipment detect and acquire finger motion.

2. The image overflow motion detection analysis system of claim 1, wherein: the system also comprises data analysis equipment which is used for synchronously analyzing the data collected by the electronic goniometer, the electronic video collection equipment and the electroencephalogram collection equipment so as to identify and quantify the mirror image overflow movement of the user.

3. The image overflow motion detection analysis system of claim 2, wherein: the mirror image overflowing movement refers to any bidirectional movement which occurs in 1 second or shorter time when the finger on one side of the user performs tapping movement according to the instruction and the non-tapping finger on the other side of the user performs tapping movement, and the movement comprises stretching-bending movement or bending-stretching movement, and can also comprise left-right movement or shaking movement.

4. The image overflow motion detection analysis system of claim 3, wherein: when a user sequentially executes a finger-knocking task on one hand according to requirements, the system detects and analyzes the motion condition of the non-knocking hand on the other side, wherein the motion condition comprises LHFT and RHFT; where LHFT is detected upon non-dominant left-hand finger tap and RHFT is detected upon dominant right-hand finger tap.

5. The image overflow motion detection analysis system of claim 4, wherein: the calculation method for the data collected by the electronic goniometer is to average the accumulated angular deflection of the non-knock hand in the LHFT and RHFT processes, in particular

Y _{totaloverflow} ＝Y _LHFToverflow +Y _RHFToverflow

In the above formula, Y _LHFToverflow And Y _RHFToverflow Respectively obtaining the average mirror image overflow number obtained by quantization when the left hand and the right hand move independently; y is _{totaloverflow} Is the total mirror overflow average for left and right hand movements; i is the trial count; n is the total number of trials;

and

the number of mirror image overflows measured by left and right hand ring fingers of each trial;

and

6. The image overflow motion detection analysis system of any of claims 1-5, wherein: the electroencephalogram acquisition equipment at least adopts 47 channels to acquire electroencephalogram data, the electrode placement conforms to an international 10-20 system electrode placement method, the electrode impedance is kept below 15k omega, and the sampling rate is more than or equal to 1024 Hz.

7. The image overflow motion detection analysis system of claim 6, wherein: the data analysis equipment is also used for preprocessing the electroencephalogram data collected by the electroencephalogram collection equipment, and the preprocessing comprises the following steps: taking the average value of all the collected electrode data as a re-reference value;

filtering the re-reference value;

carrying out segmentation and baseline correction on continuous electroencephalogram data;

the artifacts are identified and removed using a blind source separation algorithm.

8. The image overflow motion detection analysis system of claim 7, wherein: the data analysis equipment also carries out source imaging on the preprocessed electroencephalogram data, and carries out event-related desynchronization analysis and brain network analysis on a source layer so as to research brain activation, a neuron group desynchronization state and interaction among neuron groups.

9. The image overflow motion detection analysis system of any of claims 1-8, wherein: the system is used for detecting, identifying, quantifying or evaluating the image overflow dyskinesia of the attention deficit hyperactivity disorder.

10. A method for detecting a mirror overflow motion, which is suitable for quantifying a mirror overflow motion of attention deficit hyperactivity disorder, is characterized in that:

step S100: when a tested crowd carries out a task of sequentially knocking fingers on one side, an electronic goniometer and electronic video acquisition equipment are applied to measure and quantify the mirror image overflow motion of the fingers on the other side; meanwhile, electroencephalogram signals of a tested person group when a single-side sequential finger knocking task is executed are collected by electroencephalogram collecting equipment;

step S200: preprocessing the acquired electroencephalogram signals to obtain electroencephalogram data; the preprocessing comprises one or more items of down-sampling, re-referencing, high-low pass filtering, segmentation, baseline correction, artifact identification and removal, bad track replacement, bad trial times identification and deletion and time window selection of interest;

step S300: performing source imaging, ERD analysis of a source layer and brain network analysis on the preprocessed electroencephalogram data so as to research brain activation, a neuron group desynchronization state and interaction among neuron groups;

step S400: and carrying out independent sample t test, mixed repeated measurement analysis of variance, adjustment analysis and Pearson correlation analysis on the electroencephalogram data after the behaviours and the processing.