A reliable tool for evaluating brain health

Abstract

Systems and a computer implemented method for classifying a brain status of a subject, from a neural activity response of the subject to an induced TMS stimulation; the method comprising: constructing a machine learning classifier (MLC) configured to classify a subjects brain status; training the MLC using a training set, the training set comprising pairs of training output-classification vectors and their corresponding training input vectors, all extracted from a database of subjects with known brain status classifications; and applying the trained MLC on an input vector comprising features extracted from a tested-subjects brain neural activity response to the induced TMS stimulation, to obtain an output classification vector for the tested-subjects brain status.

1 . A computer implemented method for classifying a brain status of a subject, from a neural activity response of the subject to an induced TMS stimulation; the method comprising:
 constructing a machine learning classifier (MLC) configured to classify a subject's brain status;   training the MLC using a training set, the training set comprising pairs of training output-classification vectors and their corresponding training input vectors, all extracted from a database of subjects with known brain status classifications, wherein:
 each training output-classification vector is determined based on at least one of a database-subject's known: MRI readings, physician/s classification, cognitive test/s evaluation, and any combination thereof; 
 each training input vector comprises features extracted from a database-subject's brain neural activity response to the induced TMS stimulation; 
   applying the trained MLC on an input vector comprising features extracted from a tested-subject's brain neural activity response to the induced TMS stimulation, to obtain an output classification vector for the tested-subject's brain status;   wherein the TMS stimulation frequency is at least one selected from:
 below 0.5 Hz, for a neural response which does not depend upon stimulations history; and 
 above 0.5 Hz, for a neural response that is affected by previously provided stimulations pulses. 
   
     
     
         2 . The method of  claim 1 , wherein each output classification vector and accordingly each training output classification vector comprise features selected from:
 physical status selected from: healthy/not healthy, the brain's evaluated age, neurological disorders, neurodegenerative disorders, Alzheimer, Dementia, small vessels disease, Psychiatric disorders, depression, chronic pain, physical injury, pathophysiological abnormalities, structural damage of the grey matter, structural damage of the white matter, functional damage, internal bleeding, balance between excitation and inhibition in the regional cortical network, intra-cranial pressure, cerebro-vascular accident (CVA), Basal ganglia injury, brain stem injury, corticospinal track injury, frontal lobe injury, temporal lobe injury, diabetes, hypertension, any combination thereof;   brain MRI-T1—gray matter and white matter volume and/or surface of cortical and subcortical areas;   diffused tensor MRI imaging (MRI-DWI)—white matter measures of fractional anisotropy (FA) and mean diffusivity (MD);   and any combination thereof.   
     
     
         3 . The method of  claim 2 , further comprising a step of determining each training output-classification vector, based on at least one of a database-subject's known features. 
     
     
         4 . The method of  claim 1 , wherein the MLC comprises at least one module selected from:
 a multi layered MLC;   a classification module, configured for separation of the extracted features into discrete classification groups, selected from:
 support vector machine (SVM), 
 decision trees, and 
 K-nearest neighbors; 
   a registration module configured for continuous data prediction, selected from:
 linear and/or non-linear regression, 
 artificial neural network (NN), and 
 adaptive fuzzy logic learning; and 
   any combination thereof.   
     
     
         5 . The method of  claim 1 , wherein each of the input vectors and accordingly each of the training input vectors further comprises at least one feature selected from: age, gender, known medical status, drug treatment, blood pressure, and any combination thereof. 
     
     
         6 . The method of  claim 1 , wherein the TMS simulation frequencies are:
 at least one selected from below 0.5 Hz, for a neural response, which does not depend upon stimulations history; and   at least one selected from above 0.5 Hz, such that a neural response to pulses are affected by the previously provided pulses, thereby indicating short term plasticity.   
     
     
         7 . The method of  claim 1 , further comprising steps of:
 receiving, via an EEG device, a neural activity response of a subject's brain to the induced TMS stimulation to one or more predetermined brain regions of a subject; and   extracting response features from the subject's neural activity response, as elements for an input vectors or a training input vector.   
     
     
         8 . The method of  claim 7 , wherein the step of extracting is at least based on positive and negative peaks at the neural activity response, and wherein the response features comprise at least one of the response's:
 signal amplitudes;   amplitude latencies;   principle component analysis (PCA) and/or independent component analysis (ICA);   slopes between positive and negative peaks;   charge transfer;   lag of signal phase from healthy signal;   signal correlation to a healthy subject signal model;   ratio between segments in the signal of the same sensor, when TMS induced frequency is above 0.5 Hz;   coherence between the signal of different sensors;   brain connectivity; and   any combination thereof.   
     
     
         9 . The method of  claim 8 , wherein the step of extracting comprises determining the positive and negative peaks, at time steps selected from a group consisting of: about 60 mSec, about 100 mSec, about 180 mSec and any combination thereof. 
     
     
         10 . The method of  claim 8 , wherein the step of extracting comprises determining the positive and negative peaks, at time steps selected from a group consisting of: about 45 mSec, about 120 mSec, about 180 mSec, about 360 mSec and any combination thereof. 
     
     
         11 . The method of  claim 8 , wherein the extracted slopes are provided between determined positive and negative signal peaks, which are adjacent, thereby extracting peaks' relation. 
     
     
         12 . The method of  claim 9 , wherein the step of extracting features further comprises comparing the slope of the 60 mSec peak with the 100 mSec peak (60-100 slope), versus the slope of 100 mSec peak with 180 mSec peak (100-180 slope). 
     
     
         13 . The method of  claim 10 , wherein the step of extracting features further comprises comparing the slope of the 45 mSec peak with the 120 mSec peak (45-120 slope), versus the slope of 180 mSec peak with 300 mSec peak (180-300 slope). 
     
     
         14 . The method of  claim 1 , wherein the TMS is induced in several sequential stimulations, each at different intensity. 
     
     
         15 . The method of  claim 1 , wherein the TMS stimulated brain region is selected from a group consisting: frontal, parietal, temporal, occipital (right and left hemispheres) and any combination thereof. 
     
     
         16 . An apparatus configured to evaluate brain state of a subject, the apparatus comprising:
 a directed inspective/diagnostic stimulation unit, configured to induce TMS diagnostic stimulation to a predetermined brain region of the subject;   a brain activity EEG sensor, configured to measure a neural activity response to the diagnostic stimulation induced by the directed brain stimulation unit; and   a processing circuitry and at least one memory unit, in wired or wireless communication with the brain activity sensor, the processing circuitry is configured to execute at least the steps of a computer implemented method for classifying a brain status of a subject, from a neural activity response of the subject to an induced TMS stimulation; wherein the method steps are configured to:
 construct a machine learning classifier (MLC) configured to classify a subject's brain status; 
 train the MLC using a training set, the training set comprising pairs of training output-classification vectors and their corresponding training input vectors, all extracted from a database of subjects with known brain status classifications, wherein:
 each training output-classification vector is determined based on at least one of a database-subject's known: MRI readings, physician/s classification, cognitive test/s evaluation, and any combination thereof; 
 each training input vector comprises features extracted from a database-subject's brain neural activity response to the induced TMS stimulation; 
 
 apply the trained MLC on an input vector comprising features extracted from a tested-subject's brain neural activity response to the induced TMS stimulation, to obtain an output classification vector for the tested-subject's brain status; 
   and wherein the TMS stimulation frequency is at least one selected from:
 below 0.5 Hz, for a neural response which does not depend upon stimulations history; and 
 above 0.5 Hz, for a neural response that is affected by previously provided stimulations pulses. 
   
     
     
         17 . A non-transitory computer readable medium (CRM) that, when loaded into a memory of a computing device and executed by at least one processor of the computing device, configured to execute at least the steps of a computer implemented method for classifying a brain status of a subject, from a neural activity response of the subject to an induced TMS stimulation; wherein the method steps are configured to:
 construct a machine learning classifier (MLC) configured to classify a subject's brain status;   train the MLC using a training set, the training set comprising pairs of training output-classification vectors and their corresponding training input vectors, all extracted from a database of subjects with known brain status classifications, wherein:
 each training output-classification vector is determined based on at least one of a database-subject's known: MRI readings, physician/s classification, cognitive test/s evaluation, and any combination thereof; 
 each training input vector comprises features extracted from a database-subject's brain neural activity response to the induced TMS stimulation; 
   apply the trained MLC on an input vector comprising features extracted from a tested-subject's brain neural activity response to the induced TMS stimulation, to obtain an output classification vector for the tested-subject's brain status;   and wherein the TMS stimulation frequency is at least one selected from:
 below 0.5 Hz, for a neural response which does not depend upon stimulations history; and 
 above 0.5 Hz, for a neural response that is affected by previously provided stimulations pulses.