WO2018067761A1

WO2018067761A1 - Data processing for simultaneous neuromodulation and neuroimaging

Info

Publication number: WO2018067761A1
Application number: PCT/US2017/055229
Authority: WO
Inventors: Vladimir MISKOVIC
Original assignee: The Research Foundation For The State University Of New York
Priority date: 2016-10-07
Filing date: 2017-10-05
Publication date: 2018-04-12

Abstract

A method of processing EEG signals, comprising: receiving a set of EEG signals representing brain activity and artifacts from a plurality of electrodes, subject to an external stimulation which generates electrical artifacts and stimulates brain activity reflected in EEG signals; filtering a direct effect of the external stimulation on the EEG signals; performing a dimensionality reduction on the EEG signals; performing a separation operation on the reduced dimensionality EEG signals; and outputting a modified set of EEG signals from which the artifacts generated by the external stimulation, but not the response of the brain to the external stimulation, are suppressed. The separation operation may be a blind source separation operation.

Description

DATA PROCESSING FOR SIMULTANEOUS NEUROMODULATION AND

NEUROIMAGING

FIELD OF THE INVENTION

[1] The present invention relates to the removal of stimulation-dependent electrical artifacts from ongoing recordings of human brain electrical activity with simultaneous brain simulation.

BACKGROUND

[2] Non-invasive brain stimulation devices that transcranially alter human brain activity are becoming increasingly popular, both in basic and applied science. The capacity to combine, in real time, technologies for the stimulation of human brain activity alongside devices that record ongoing brain activity is a major goal for researchers. Such co-registration paves the way for so- called closed-loop neuromodulation, i.e. , the ability to tailor brain stimulation protocols to dynamic and ongoing changes in neuronal function. A major impediment to realizing this goal is that stimulation technologies create strong electrical artifacts that massively distort recordings of neuronal function. There is thus a need for technologies that selectively remove stimulation- induced electrical artifacts from signals, such as EEG signals, that represent the brain electrical activity of human subjects.

[3] The simultaneous integration of neurostimulation devices with technology for monitoring largescale neuronal dynamics creates an exciting opportunity for exploring stimulation effects in real time. This integration permits closed-loop neuromodulation protocols tailored to match constantly fluctuating patterns of brain activity.

[4] Several challenges have stymied efforts to combine subthreshold transcranial neuromodulation with real time measurement of electroencephalography (EEG) or

magnetoencephalography (MEG). First, and foremost, the stimulation-dependent induction of electrical artifact saturates and corrupts neural signals. Second, ongoing physiological processes non- linearly modulate stimulation artifacts, as in the case of transcranial electric stimulation (tES) [1]. Implementation constraints add additional challenges - traditional tES electrodes occupy a sizeable surface on the scalp (5 x 7 cm²), reducing the spatial sampling of

electrocortical activity. [5] The optimal combination of stimulation and recording methods requires using modalities that minimize spatiotemporal barriers between each other. Here, we briefly describe encouraging preliminary results demonstrating the integration of high-density EEG with a novel subthreshold magnetic neuromodulation device. [6] Subthreshold Magnetic Stimulation (StMS) is a form of subthreshold brain stimulation that utilizes static or alternating magnetic fields to induces small, subthreshold changes to the polarity of the underlying brain tissue. Low field magnetic stimulation (LFMS) is a form of StMS that utilizes a large (15" diameter) X-gradient MRI coil that delivers repeated 500 Hz trains of low field strength (< 50 Gauss) magnetic stimulation broadly to the cortex. Each 500 ms train is repeated with a 2 s period for 20 minutes (Fig. 1A). LFMS was discovered serendipitously in a spectroscopic MRI study of bipolar depression. Following the scan, patients self-reported rapid improvements in mood [2]. Subsequently, a head coil device that applied the same low field strength and temporal parameters of the relevant MR sequence was developed and applied to patients with unipolar or bipolar depression [3]. This treatment improved depressed mood over the sham condition (audio, no magnetic field) in only 20 minutes.

[7] The fields generated by LFMS are weak compared to related technologies such as transcranial magnetic stimulation (TMS), and the induced electric field strengths are

subthreshold for neurons. (The possible effects on calcium signaling in glial networks remain largely unexplored [4]). In spite of this, the effects of low- strength exogenous fields (< 1 V/m) can be amplified within recurrently connected networks to change neuronal spiking patterns, alter membrane potentials, and influence synaptic integration [5-7]. A detailed characterization of how subthreshold neuromodulation technologies, such as LFMS, acutely alter the

spatiotemporal structure of cortical activity is of critical importance, and represents a novel research avenue with considerable basic and translational interest. See, Agostino, R., Iezzi, E., Dinapoli, L., Gilio, F., Conte, A., Mari, F., & Berardelli, A. (2007). Effects of 5 Hz subthreshold magnetic stimulation of primary motor cortex on fast finger movements in normal subjects. Experimental brain research, 180(1), 105-111; Petersen, Nicolas T., Jane E. Butler, Veronique Marchand - Pauvert, Rebecca Fisher, Annick Ledebt, Henrik S. Pyndt, Naja L. Hansen, and Jens B. Nielsen. "Suppression of EMG activity by transcranial magnetic stimulation in human subjects during walking." The Journal of Physiology 537, no. 2 (2001): 651-656; Mazzocchio, R., Rothwell, J. C, Day, B. L., & Thompson, P. D. (1994). Effect of tonic voluntary activity on the excitability of human motor cortex. The Journal of Physiology, 474(2), 261-267; Dayan, E., Censor, N., Buch, E. R., Sandrini, M., & Cohen, L. G. (2013). Noninvasive brain stimulation: from physiology to network dynamics and back. Nature neuroscience, 16(7), 838-844; Fang, J. H., Chen, J. J., Hwang, I. S., & Huang, Y. Z. (2010). Repetitive transcranial magnetic stimulation over the human primary motor cortex for modulating motor control and motor learning. J. Med. Biol. Eng, 30, 193-201.

[8] Each of the following references, and any others cited herein, are expressly incorporated herein by reference in their entirety.

[9] [1] Noury N, Hipp JF, Siegel M. Physiological processes non-linearly affect

electrophysiological recordings during transcranial electric stimulation. Neuroimage 2016; in press.

[10] [2] Rohan M, Parow A, Stoll AL, Demopulos C, Friedman S, Dager S. et al. Low-field magnetic stimulation in bipolar depression using an MRI-based stimulator. Am J Psychiatry 2004;161; 93-98. [11] [3] Rohan ML, Yamamoto RT, Ravichandran CT, Cayetano KR, Morales OG, Olson DP et al. Rapid mood-elevating effects of low field magnetic stimulation in depression. Biol Psychiatry 2014; 76; 186-193.

[12] [4] Monai H, Ohkura M, Tanaka M, Oe Y, Konno A, Hirai H, Mikoshiba K et al.

Calcium imaging reveals glial involvement in transcranial direct current stimulation-induced plasticity in mouse brain. Nat. Commun. 2016;7;11100.

[13] [5] Frohlich F, McCormick DA. Endogenous electric fields may guide neocortical network activity. Neuron 2010; 67; 129-143.

[14] [6] Anastassiou CA, Montgomery SM, Barahona M, Buzsaki G, Koch C. The effect of spatially inhomogenous extracellular electric fields on neurons. J Neurosci 2010; 30; 1925-1936. [15] [7] Francis JT, Gluckman BJ, Schiff SJ. Sensitivity of neurons to weak electric fields. J Neurosci 2003;23;7255-7261. [16] [8] Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods 2004;134; 9-21.

[17] Pascual-Marqui, Roberto D., Christoph M. Michel, and Dietrich Lehmann. "Low resolution electromagnetic tomography: a new method for localizing electrical activity in the brain." International Journal of psychophysiology 18.1 (1994): 49-65.

[18] Ilmoniemi, Risto J., et al. "Neuronal responses to magnetic stimulation reveal cortical reactivity and connectivity." Neuroreport 8.16 (1997): 3537-3540.

[19] Herwig, Uwe, Peyman Satrapi, and Carlos Schonfeldt-Lecuona. "Using the international 10-20 EEG system for positioning of transcranial magnetic stimulation." Brain topography 16.2 (2003): 95-99.

[20] Roth, Yiftach, Abraham Zangen, and Mark Hallett. "A coil design for transcranial magnetic stimulation of deep brain regions." Journal of Clinical Neurophysiology 19.4 (2002): 361-370.

[21] Bonato, C, C. Miniussi, and P. M. Rossini. "Transcranial magnetic stimulation and cortical evoked potentials: a TMS/EEG co-registration study." Clinical neurophysiology 117.8 (2006): 1699-1707.

[22] Bonato, C, C. Miniussi, and P. M. Rossini. "Transcranial magnetic stimulation and cortical evoked potentials: a TMS/EEG co-registration study." Clinical neurophysiology 117.8 (2006): 1699-1707. [23] Leuchter, Andrew F., et al. "Synchronized Transcranial Magnetic Stimulation (sTMS): Efficacy and safety of low-field synchronized transcranial magnetic stimulation (sTMS) for treatment of major depression." Brain Stimul (2015).

[24] Rossini, P. M., et al. "Non-invasive electrical and magnetic stimulation of the brain, spinal cord, roots and peripheral nerves: basic principles and procedures for routine clinical and research application. An updated report from an IFCN Committee." Clinical Neurophysiology 126.6 (2015): 1071-1107. [25] Gonzalez-Rosa, Javier J., et al. "Static magnetic field stimulation over the visual cortex increases alpha oscillations and slows visual search in humans." The Journal of Neuroscience 35.24 (2015): 9182-9193.

[26] Opitz, Alexander, et al. "Determinants of the electric field during transcranial direct current stimulation." Neuroimage 109 (2015): 140-150.

[27] Saturnino, Guilherme B., Andre Antunes, and Axel Thielscher. "On the importance of electrode parameters for shaping electric field patterns generated by tDCS." Neuroimage 120 (2015): 25-35.

[28] Farzan, Faranak, et al. "Characterizing and Modulating Brain Circuitry through

Transcranial Magnetic Stimulation Combined with Electroencephalography." Frontiers in Neural Circuits 10 (2016): 73.

[29] Capotosto, Paolo, et al. "Dynamics of EEG rhythms support distinct visual selection mechanisms in parietal cortex: A simultaneous transcranial magnetic stimulation and EEG study." The Journal of Neuroscience 35.2 (2015): 721-730. [30] Garcia-Cossio, Eliana, et al. "Simultaneous transcranial direct current stimulation (tDCS) and whole -head magnetoencephalography (MEG): assessing the impact of tDCS on slow cortical magnetic fields." Neuroimage (2015).

[31] Espy, Michelle A., et al. "System and method for magnetic current density imaging at ultra low magnetic fields." U.S. Patent No. 9,254,097. 9 Feb. 2016. [32] Dowsett, James, and Christoph S. Herrmann. "Transcranial Alternating Current

Stimulation with Sawtooth Waves: Simultaneous Stimulation and EEG Recording." Frontiers in human neuroscience 10 (2016).

[33] Kitajo, Keiichi, et al. "A contemporary research topic: manipulative approaches to human brain dynamics." Frontiers in human neuroscience 9 (2015). [34] Sato, Sumire, Til Ole Bergmann, and Michael R. Borich. "Opportunities for concurrent transcranial magnetic stimulation and electroencephalography to characterize cortical activity in stroke." Frontiers in human neuroscience 9 (2015). [35] Neuling, Toralf, et al. "Friends, not foes: magnetoencephalography as a tool to uncover brain dynamics during transcranial alternating current stimulation." Neuroimage 118 (2015): 406-413.

[36] Chung, Sung Wook, et al. "Measuring brain stimulation induced changes in cortical properties using TMS-EEG." Brain stimulation 8.6 (2015): 1010-1020.

[37] US Patent and Pat. Pub. Nos. 6117066; 6488617; 6516246; 6572528; 6629935; 6978179;

20060264691; 20070203390; 20090287274; 20110015469; 20110034822; 20110130615 20110137104; 20120022315; 20120083647; 20120271377; 20120283193; 20120289869 20120296569; 20130066138; 20130137918; 20130204122; 20140228621; 20140235929 20150045606; 20150051663; 20150066104; 20150112409; 20150119689; 20150126802 20150223721; 20150265830; 20150265836; 20150328454; 20150343222; 20150351655 20150367141; 20150379878; 20160001074; 20160001096; 20160008620; 20160008632 20160038049; 20160051818; 20160067496; 20160096032; 20160121114; 20160175589 20160199662; 20160220850; 20160235983; 20160250465; and 20160256105.

SUMMARY OF THE INVENTION

[38] It is therefore an object to provide a method of processing EEG signals, comprising: receiving a set of EEG signals representing brain activity and artifacts from a plurality of electrodes, subject to an external stimulation which generates electrical artifacts and stimulates brain activity reflected in EEG signals; filtering a direct effect of the external stimulation on the EEG signals; performing a dimensionality reduction on the EEG signals; performing a separation operation on the reduced dimensionality EEG signals; and outputting a modified set of EEG signals from which the artifacts generated by the external stimulation, but not the response of the brain to the external stimulation, are suppressed. [39] It is also an object to provide a system for processing EEG signals, comprising: an input configured to receive a set of EEG signals representing brain activity and artifacts from a plurality of electrodes, subject to an external stimulation which generates electrical artifacts and stimulates brain activity reflected in EEG signals; a filter configured to filter a direct effect of the external stimulation on the EEG signals; at least one automated processor configured to reduce a dimensionality of the EEG signals; perform a separation operation on the reduced dimensionality EEG signals; and output a modified set of EEG signals from which the artifacts generated by the external stimulation, but not the response of the brain to the external stimulation, are suppressed.

[40] It is another object to provide a non-transitory computer-readable medium storing instructions for controlling a programmable processor to implement a method of processing EEG signals, comprising: instructions for receiving a set of EEG signals representing brain activity and artifacts from a plurality of electrodes, subject to an external stimulation which generates electrical artifacts and stimulates brain activity reflected in EEG signals; instructions for filtering a direct effect of the external stimulation on the EEG signals; instructions for performing a dimensionality reduction on the EEG signals; instructions for performing a separation operation on the reduced dimensionality EEG signals; and instructions for outputting a modified set of

EEG signals from which the artifacts generated by the external stimulation, but not the response of the brain to the external stimulation, are suppressed. .

[41] It is a further object to provide a method of processing EEG signals acquired in an electrically noisy environment comprising an external stimulation of a brain which induces or generates artifacts and stimulates or excites brain activity reflected in EEG signals, comprising: receiving EEG signals representing activity of the brain and the induced or generated artifacts from a plurality of electrodes; filtering a direct effect of the external stimulation on the EEG signals; performing a dimensionality reduction on the EEG signals; separating a first component of the reduced dimensionality EEG signals reflecting brain activity from a second component of the reduced dimensionality EEG signals reflecting residual artifacts and an indirect effect induced or generated by the external stimulation; and selectively outputting the first component of the reduced dimensionality EEG signals reflecting the brain activity, wherein the artifacts induced or generated by the external stimulation, but not the brain activity stimulated or excited by the external stimulation, are suppressed.

[42] The separation operation may be a blind source separation operation. Residual non- filterable artifacts may be identified, and replaced with statistically normal signals. The external stimulation may comprise an LFMS procedure. At least one of the plurality of electrodes may have an impedance of up to 60 kOhm. The EEG signals may be sampled at a frequency between 0.5 and 5kHz. The LFMS procedure may be conducted using a device coil housing positioned over the head of a user. The LFMS procedure may be modified based upon the modified set of EEG signals.

BRIEF DESCRIPTION OF THE DRAWINGS

[43] Figs. 1A-1F show combined high-density EEG recordings with:

[44] Fig. 1A, simultaneous LFMS stimulation using the standard protocol employed in previous reports. [45] Fig. IB shows the LFMS trains induced stimulation artifacts in ongoing EEG (raw), and after digital filtering with a combined PCA-ICA, which allowed elimination of stimulation artifacts in the EEG (cleaned).

[46] Fig. 1C shows stimulation onset and offset transients were effectively captured by a single ICA component whose time activation is depicted. [47] Fig. ID shows Time x frequency decompositions before removal of the ICA-LFMS component are depicted for electrode Fz in a representative subject (S2).

[48] Fig. IE shows Time x frequency decompositions after removal of the ICA-LFMS component are depicted for electrode Fz in a representative subject (S2).

[49] Fig. IF shows topographic distributions of projection weights for the ICALFMS component are shown for single subjects, providing a rough measure of LFMS field distribution across the scalp. (Note: the relative polarity of ICA projection weights is functionally neutral - both red and blue spots indicate regions that contribute equally to the component in question, while green regions do not).

[50] Fig. 2 shows a flow diagram of one embodiment of the Invention. [51] Fig. 3 shows a block diagram prior art computer system suitable for implementing the technology. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[52] The present technology combines transcranial low field magnetic stimulation (LFMS), a form of Subthreshold Magnetic Stimulation (StMS), with real-time monitoring of brain activity through electroencephalography (EEG) in human subjects and have developed an efficient data processing pipeline that is able to selectively remove stimulation-induced electrical artifacts from the combined signal. By using the present technology , it is possible to recover clean recordings of human brain activity, even in the presence of active neurostimulation. The present technology relies statistical procedures (i.e. , digital filtering, spatial principal components analysis and independent components analysis) used in a specific application (i.e. , brain electrical recordings in the context of simultaneous magnetic brain stimulation) to solve a specific problem (i.e. , the removal of stimulation-dependent recording artifacts and the recovery of clean brain signals from collected data).

[53] Several challenges have hampered efforts to combine transcranial neuromodulation with real-time measurement of EEG or magnetoencephalography (MEG). For example, the stimulation-dependent induction of electrical artifact saturates and corrupts the neural signals. Also, ongoing physiological processes (e.g. , respiration) non- linearly modulate stimulation- dependent artifacts, as in the case of transcranial electrical stimulation (tES). Implementation constraints add additional challenges; for example, traditional tES electrodes occupy a sizeable portion of the scalp surface (~5 x 7 cm²), reducing the spatial sampling of electrocortical activity. [54] Spatiotemporal barriers between stimulation and recording are minimized by combining high-density EEG with subthreshold magnetic neuromodulation device manufactured by Tal Medical, Inc. (Boston, MA). See,

[55] Volkow ND, Tomasi D, Wang G-J, et al. Effects of Low-Field Magnetic Stimulation on Brain Glucose Metabolism. Neurolmage. 2010;51(2):623-628.

doi: 10.1016/j.neuroimage.2010.02.015.,

www.ncbi.nlm.nih.gov/pmc/articles/PMC2862488/pdf/nihmsl89576.pdf Rapid Mood- Elevating Effects of Low Field Magnetic Stimulation in Depression, Rohan, Michael L. et al., Biological Psychiatry , Volume 76 , Issue 3 , 186 - 193.

www.biologicalpsychiatryjournal.com article/S0006-3223(13)00981-5/fulltext. [56] As previously described, the simultaneous magnetic stimulation with EEG recording induces prominent electrical artifacts in ongoing recordings of electrical brain activity. The present technology provides a suite of computational algorithms that minimize or remove such electrical artifacts from signals representing human brain activity. Such algorithms consist of a series of linked operations that include the following:

[57] (1) a digital filtering operation to remove the high frequency artifacts introduced by oscillating magnetic pulses; and

[58] (2) the removal of stimulation onset and offset electrical transients by a combined procedure that first reduces the dimensionality of the EEG data and subsequently uses a mathematical procedure known as blind source separation (or independent components analysis).

[59] See MIT HST-582J/6.555J/16.456J, Course Materials, Biomedical Signal and Image

Processing, Spring 2008, Chapter 15 - BLIND SOURCE SEPARATION:, Principal &

Independent Component Analysis © G.D. Clifford 2005-2008,

www.mit.edu/~gari/teaching/6.555/LECTURE_NOTES/chl5_bss.pdf;

www.mit.edu/~gari/teaching/6.555/SLIDES/BSShandouts.pdf;

[60] Matthieu Puigt, "A Very Short Introduction to Blind Source Separation a.k.a. How You

Will Definitively Enjoy Differently a Cocktail Party", Foundation for Research and Technology -

Hellas Institute of Computer Science www.ics.forth.gr/~mpuigt (April 12 / May 3, 2011), www- lisic.univ-littoral.fr/~puigt/LectureNotes/Intro_BSS/Lecture_BSS.pdf;

[61] K. H. Knuth, "Difficulties Applying Recent Blind Source Separation Techniques To EEG and MEG", Maximum Entropy and Bayesian Methods, Boise, Idaho, 1997, G.J. Erickson, J.T.

Rychert and C.R . Smith (eds.), Kluwer Academic, Dordrecht, pp. 209-222,

arxiv.org/ftp/arxiv/papers/1501/1501.05068.pdf;

[62] Lucas Parra, "Tutorial on Blind Source Separation and Independent Component

Analysis" Adaptive Image & Signal Processing Group, Sarnoff Corporation, February 09, 2002, bme.ccny.cuny.edu/faculty/parra/publish/tutorial-bss.pdf;

[63] Pedersen, Michael S., et al. "A survey of convolutive blind source separation methods." Multichannel Speech Processing Handbook (2007): 1065-1084.,

www2.imm.dtu.dk/pubdb/views/edoc_download.php/4924/pdf/imm4924.pdf;

[64] Cichocki, Andrzej S. "Blind source separation: new tools for extraction of source signals and denoising." Defense and Security. International Society for Optics and Photonics, 2005, www.bsp.brain.riken.jp/publications/2005/Cichocki_SPIE_05.pdf; [65] Cichocki, Andrzej, et al. "EEG filtering based on blind source separation (BSS) for early detection of Alzheimer's disease." Clinical Neurophysiology 116.3 (2005): 729-737., www.bsp.brain.riken.jp/publications/2005/Ci_Shis_AD_Clin_NeurPhys.pdf;

[66] Parra, Lucas, and Paul Sajda. "Blind source separation via generalized eigenvalue decomposition." Journal of Machine Learning Research 4.Dec (2003): 1261-1269.,

www.jmlr.org/papers/volume4/parra03a/parra03a.pdf;

[67] Li, Yuanqing, et al. "Sparse representation and its applications in blind source separation." Advances in neural information processing systems 16 (2004): 241-248., papers.nips.cc/paper/2379-sparse-representation-and-its-applications-in-blind-source- separation.pdf;

[68] Belouchrani, Adel, et al. "A blind source separation technique using second-order statistics." IEEE Transactions on signal processing 45.2 (1997): 434-444.,

www.bsp.brain.riken.jp/ICApub/A97-434.pdf;

[69] Jung, Tzyy-Ping, et al. "Removing electroencephalographic artifacts by blind source separation." Psychophysiology 37.02 (2000): 163-178., sccn.ucsd.edu/~jung/pdf/PSYOO.pdf;

[70] Tzyy-Ping Jung & Scott Makeig, "EEG artifact removal with Blind Source Separation", (2011) sccn.ucsd.edu/~jung/Site/EEG_artifact_removal.html;

[71] Jung TP, Makeig S, Humphries C, Lee TW, McKeown MJ, Iragui V, Sejnowski TJ. Removing electroencephalographic artifacts by blind source separation. Psychophysiology, 2000; 37: 163-78., sccn.ucsd.edu/~jung/pdf/PSY00.pdf

[72] Schmidt, Mikkel. "Linearly constrained bayesian matrix factorization for blind source separation." Advances in neural information processing systems. 2009.,

mlg.eng.cam.ac.uk/pub/pdf/Sch09b.pdf;

[73] L. Parra; C. Spence; B. De Vries, "Convolutive blind source separation based on multiple decorrelation", Neural Networks for Signal Processing VIII, 1998. Proceedings of the 1998 IEEE Signal Processing Society Workshop ( 1998) Pages: 23 - 32, DOI: 10.1109/NNSP.1998.710626, ieeexplore.ieee.org/document/841214/;

[74] Bobin, Jerome, et al. "5 Blind Source Separation: The Sparsity Revolution." Advances in Imaging and Electron Physics 152.1 (2008): 221-302., hal-univ-diderot.archives-ouvertes.fr/hal- 00252075/document; [75] Bronstein, Alexander M., Michael M. Bronstein, and Michael Zibulevsky. "Blind source separation using block-coordinate relative Newton method." Signal processing 84.8 (2004): 1447-1459., vista.eng.tau.ac.il/publications/BroBroZibSIGPRO04.pdf;

[76] Harmeling, Stefan, et al. "Kernel-based nonlinear blind source separation." Neural Computation 15.5 (2003): 1089-1124.,

www.mitpressjournals.org/doi/pdf/10.1162/089976603765202677;

[77] James, C. J., & Wang, S. (2006, August). Blind source separation in single-channel EEG analysis: An application to BCI. In Proc. 28th Annual Int. Conf. (2005) IEEE Engineering in Medicine and Biology Society (pp. 6544-6547). www.ncbi.nlm.nih.gov/pubmed/17959448;

[78] Stone, James V. "Blind source separation using temporal predictability." Neural computation 13.7 (2001): 1559-1574., eprints.whiterose.ac.uk/1430/l/stone.j.wl.pdf;

[79] Hyvarinen, Aapo. "Blind source separation by nonstationarity of variance: a cumulant- based approach." IEEE Transactions on Neural Networks 12.6 (2001): 1471-1474.,

www.cs.helsinki.fi/u/ahyvarin/papers/TNNO 1.ps, ;

[80] Cedric Fevotte, Remi Gribonval, Emmanuel Vincent. BSS EVAL Toolbox User Guide { Revision 2.0. [Technical Report] 2005, pp.19. <inria-00564760>, hal.inria.fr/inria- 00564760/document;

[81] "BSS Eval, A toolbox for performance measurement in (blind) source separation", (2013) bass-db.gforge.inria.fr/bss_eval/;

[82] Zhao, Xiaochen, et al. "Blind Source Separation Based on Dictionary Learning: A

Singularity-Aware Approach." Blind Source Separation. Springer Berlin Heidelberg, 2014. 39- 59. citeseerx.ist.psu.edu/viewdoc/download?doi= 10.1.1.719.1902&rep=rep 1 &type=pdf

[83] Peterson, David A., et al. "Feature selection and blind source separation in an EEG-based brain-computer interface." EURASIP Journal on Applied Signal Processing 2005 (2005): 3128- 3140., www.cs.colostate.edu/eeg/publications/jasp05.pdf;

[84] Fitzgibbon, S. P., et al. "Removal of EEG noise and artifact using blind source separation." Journal of Clinical Neurophysiology 24.3 (2007): 232-243.,

www.ncbi.nlm.nih.gov/pubmed/17545826;

[85] Torse, D. A., R. R. Maggavi, and S. A. Pujari. "Nonlinear blind source separation for EEG signal pre-processing in brain-computer interface system for epilepsy." International Journal of Computer Applications 50.14 (2012).,

research.ijcaonline.org/volume50/numberl4/pxc3880911.pdf; [86] Lan Ma, Thierry Blu, and William S-Y. Wan, "An EEG Blind Source Separation Algorithm Based On A Weak Exclusion Principle",

www.ee.cuhk.edu.hk/~tblu/monsite/pdfs/mal601.pdf;

[87] Vazquez, R. Romo, et al. "Blind source separation, wavelet denoising and discriminant analysis for EEG artefacts and noise cancelling." Biomedical Signal Processing and Control 7.4 (2012): 389-400., hal.archives-ouvertes.fr/hal-00600103/document;

[88] Ball, Kenneth, et al. "PWC-ICA: A Method for Stationary Ordered Blind Source Separation with Application to EEG." Computational Intelligence and Neuroscience 2016 (2016)., www.hindawi.com/j ournals/cin/2016/9754813/ [89] The blind source separation is used to precisely isolate these electrical artifacts from the background brain activity. Following isolation of electrical artifacts, their contribution is selectively removed from EEG signals, and thus clean brain signals are restored. A flowchart setting forth the processing pipeline of the present invention is set forth in Fig. 2.

[90] The technology may be implemented using MATLAB (Mathworks, Natick MA) routines and functions, including, for example, the MATLAB Statistics Toolbox, such as the 'princomp' function, which conducts a principal components analysis, along with a collection of routines from the EEGLAB Toolbox, see sccn.ucsd.edu/eeglab/ U.C. San Diego), which is an open source, publicly available toolbox for the analysis of EEG data that runs on the MATLAB computational environment. EXAMPLE

[91] The feasibility of simultaneous EEG-LFMS was assessed in a pilot study conducted at SUNY Binghamton. High-density EEG was measured using a 129-sensor HydroCel Geodesic Sensor Net (Electrical Geodesies, Inc.) during a 20 minute resting (eyes closed) session with concurrent LFMS. Target EEG electrode impedances were 60 kOhm at maximum. The EEG was sampled at 1 KHz and referenced on-line to electrode Cz. Participants were supine on an exam table with their head positioned on an integrated neck rest. The device coil housing was aligned over the participant's head, up to the brow ridge.

[92] The post-processing pipeline was implemented in the EEGLAB toolbox (e.g., version 12.0.2.5b) for MATLAB [8] as follows: [93] First, a 1 Hz high-pass filter was applied using a Hann window with an order of 3100 in conjunction with a 55 Hz low-pass filter (310 order, Hann window).

[94] Second, channel means were subtracted and bad channels (identified by visual inspection) were subjected to a spherical spline interpolation. Next, the data dimensionality was reduced by a principal-components-analysis (PCA) to identify the number of orthogonal components accounting for 98% of variance in spatial topography.

[95] Finally, a reduced-rank independent components analysis (ICA; Infomax algorithm) was performed to extract as many components as were identified by the spatial PCA.

[96] By preceding ICA with a dimensionality reduction step, the residual stereotyped onset and offset artifacts were extracted with minimal likelihood of over- fitting the data and fragmenting the LFMS transients over multiple components (Fig. 1C, N = 4).

[97] To characterize the spectral profile of LFMS artifacts (see Fig. IB), the EEG from 1.4 sees prior to LFMS train onset to 1.8 sees following train offset was time-locked. A

time x frequency analysis was then conducted before and after removal of the ICA-LFMS component.

[98] Fig. ID illustrates the broadband artifacts associated with the onset and offset of the stimulation trains. Removal of the ICA component, followed by back projection to the sensor level, retained the ongoing cortical alpha band power while eliminating the transient broadband spikes (see Fig. IE). Note that low-pass filtering was sufficient to remove the contribution of the 500 Hz magnetic waveform fields; applying a notch filter (and no low-pass filtering) was similarly successful.

[99] These observations have several important implications for furthering research and therapeutic development. First, removal of stimulation artifacts allows the measurement of direct effects of LFMS on human cortical activity, as well as any potential secondary effects due to plasticity induced by LFMS. Furthermore, analysis of the artifacts provides insights into the spatial distribution of LFMS. Lastly, and most importantly, these observations help to pave the way for closed-loop neuromodulation based on the simultaneous or short-latency analysis of ongoing neural activity to tune the spatiotemporal parameters of brain stimulation. [100] According to one embodiment, the technology may be characterized as follows: (a) apply a filter to remove a direct contribution from the stimulation, i.e., a low pass or notch filter, though other filters such as an optimal or synchronous filter may be employed; (b) (optional) identify large amplitude distortions, such as channel saturation or likely non-linear modes, and replace with statistically normal representation; (c) reduce data dimensionality, e.g., using a statistical technique such as principal component analysis; and (d) perform an independent component analysis on the reduced dimensionality data set, with a corresponding dimensionality, to extract residual onset and offset artifacts. With the artifacts suppressed, the modified EEG data may then be further analyzed, or the artifact-suppressed reduced dimensionality dataset may be further analyzed.

[101] This technology is applicable to the fields of neuroscience, psychiatry and psychology, and can be used to clean recording artifacts introduced into EEG recordings by regular, rhythmic and stereotyped electrical or magnetic artifacts, for example as introduced by Magnetic

Resonance Imaging (MRI) and functional Magnetic Resonance Imaging (fMRI) modalities. When combined with neuromodulation technology such as the coil LFMS device from Tal

Medical, Inc., the present technology may be applicable to purposes of cognitive enhancement (e.g. , bolstering memory or attention) as well as in the treatment of somatic disorders (e.g. , sleep disorders).

[102] The present technology offers several advantages. First, the ability to remove stimulation artifacts allows for the measurement of direct effects of LFMS on human brain activity, as well as any potential secondary effects due to brain plasticity induced by LFMS. Furthermore, analysis of the artifacts provides insights into the spatial distribution of LFMS effects. Lastly, these observations help to pave the way for closed-loop neuromodulation based on the simultaneous or short-latency analysis of ongoing neural activity to tune the strength, spatial parameters and temporal parameters of brain stimulation.

[103] The device may be designed to be used in conjunction with a standard hospital bed, gurney or exam table. Once the subject lies down on the bed with their neck comfortably resting on the integrated neck rest, the device coil housing slides over the subject's head up to the supraorbital ridge. A measurement of the magnetic field or induced electric field may be performed at the skull vertex, either using a B field monitor, the EEG cap, or a small inductive wire coil, so as to verify field strength and control for relative changes in strength due to differing head sizes.

[104] The EEG may be recorded using an EGI (Electrical Geodesies, Inc., Eugene, OR) 128- channel system, with vertex (Cz) used as recording reference, e.g., digitized at a rate of 1 KHz, with impedances kept below, e.g., 50kQ, as recommended by EGL. In order to mitigate the impact of recording artifacts without introducing discontinuities in the signal (as would result from typical segment-rejection approaches), an artifact detection approach is used in preference to an artifact rejection approach.

[105] In a first step, a spatial principal components analysis (PCA) is performed, treating sensors as variables and time points as observations to identify a reduced number of orthogonal components capable of accounting for, e.g., 98% of variance in spatial topography. This is followed by a reduced-rank independent components analysis (ICA), e.g., using the Infomax algorithm implemented in the EEGLAB toolbox (version 12.0.2.5b) for MATLAB to extract as many ICA components as there are spatial principal components. [106] Independent components may then be removed before back-projecting in cases where they are deemed near-exclusively artifactual based on manual inspection of the temporal, spatial, and spectral properties or in cases where they possess a rhythmic nature that is expected to seriously distort subsequent analyses (i.e., heartbeats).

[107] The artifact correction strategy for EEG data collected simultaneous with LFMS treatment will differ in order to remove the periodic magnetic induced gradients. The preferred approach is to use Brain Vision Analyzer software for correction of gradient artifacts.

[108] Since the temporal characteristics of the LFMS field modulation are controlled and known by us ahead of time (the LFMS device provides an output to precisely track the stimulus delivery and waveform), the same general strategy may be used for artifact removal. [109] In order to mitigate the impact of volume conduction on estimates of functional connectivity, all subsequent analyses may be taken out of the raw sensor space by transforming the raw data using a surface Laplacian calculation using the spherical spline method. The spherical spline order (m) is set to 3, the smoothing (λ) parameter is set to 10^"5 and the number of iterations used in computing the Legendre polynomial is typically around 40 (this parameter is also sometimes referred to as the order of the polynomial).

[110] For robustness, several distinct quantitative measures of EEG functional connectivity may be employed. These may include (i) the debiased weighted phase-lag index (wPLI) and (ii) orthogonalized power envelope correlation. Both of these measures have the benefit that they ignore instances of functional connectivity with a phase difference of π: or 2π, thereby excluding spurious connectivity arising due to volume conduction from a common source. However, they are also complementary measures of the large-scale cortical correlation structure. Currently, there are two recognized neuronal intrinsic coupling modes - one that is mediated by phase synchronization of frequency band-limited signals and the other by the aperiodic fluctuations of signal amplitude envelopes. Phase ICMs are uniquely quantifiable only on the basis of

EEG/MEG data, while amplitude ICMs can be measured in both electrophysiological and hemodynamic measures. Both of these coupling modes may be captured. Basic graph theoretic indices may be used to provide quantitative summaries of network topology.

[I l l] To estimate brain signal complexity, the ICA-denoised time domain EEG data may be used for estimating multiscale entropy (MSE). Briefly, this procedure involves quantifying the sample entropy of EEG signals at multiple time scales. The first step involves temporal coarse graining where for a given time scale (x) the corresponding time series is calculated by averaging neighboring data points within non-overlapping windows of length x from the original time series signal. Subsequent to this coarse graining procedure, sample entropy may be quantified separately at each of the time scale factors (from 1 to 20 in this case, where 1 represents the original signal and 20 indicates a window size of 20 sample points).

[112] Sample entropy represents the conditional probability that any two consecutive data sequences of pattern length (m + 1) will match each other given a match for the first m points across the duration of a time series at a given scale factor. Subsequent patterns are considered to recur if the absolute amplitude difference falls - within a particular criterion or tolerance range, r. Sample entropy is calculated using the following equation:

S_E (m, r, N) = - ln where n™ is the number of matches and N is the length of the original time series. ¾ thus captures the regularity of a signal: low values indicate high self-similarity (low complexity) and high values denote irregularity (high complexity).

[113] Our main hypotheses may be tested using the mass univariate approach involving Monte Carlo-based sampling of t_max test distributions (> 1000 random shuffles). This mass univariate approach provides the most powerful way to combine exploratory analyses of the high dimensional spatio-temporal EEG data while protecting against the likelihood of committing Type II errors. For analyzing the specific spatial topography of functional connectivity networks and putative differences between the sham and active LFMS conditions, a similar nonparametric approach may be employed that corrects for coupling differences across multiple frequency bins and node pairs. Additional analyses of functional connectivity changes may be conducting using graph theoretic summaries of network topology as the outcome variables.

[114] The time series convolution may also be computed between the fluctuations in affect rating dial data and concurrent EEG data that has been bandpass filtered in multiple narrow frequency bands (α,θ,β,γ). This permits detection of potential similar time courses for the affective and neurophysiological shifts while identifying the particular oscillatory ranges that are most likely to be critical to mediating the mood enhancing effects of LFMS. Analyses of the LPP and viewing time data may also include low-level image characteristics (complexity as well as the degree of energy in low and high spatial frequency channels) as covariates in the main analyses.

HARDWARE OVERVIEW

[115] Fig. 3 (see US 7,702,660, issued to Chan, expressly incorporated herein by reference), shows a block diagram that illustrates a computer system 400 upon which an embodiment of the technology may be implemented. Computer system 400 includes a bus 402 or other

communication mechanism for communicating information, and a processor 404 coupled with bus 402 for processing information. Computer system 400 also includes a main memory 406, such as a random access memory (RAM) or other dynamic storage device, coupled to bus 402 for storing information and instructions to be executed by processor 404. Main memory 406 also may be used for storing temporary variables or other intermediate information during execution of instructions to be executed by processor 404. Computer system 400 further includes a read only memory (ROM) 408 or other static storage device coupled to bus 402 for storing static information and instructions for processor 404. A storage device 410, such as a magnetic disk or optical disk, is provided and coupled to bus 402 for storing information and instructions. [116] Computer system 400 may be coupled via bus 402 to a display 412, such as a liquid crystal display (LCD) or organic light emitting diode (oLED) display monitor, for displaying information to a computer user. An input device 414, including alphanumeric and other keys, is coupled to bus 402 for communicating information and command selections to processor 404. Another type of user input device is cursor control 416, such as a mouse, a trackball, or cursor direction keys for communicating direction information and command selections to processor 404 and for controlling cursor movement on display 412. This input device typically has two degrees of freedom in two axes, a first axis (e.g., x) and a second axis (e.g., y), that allows the device to specify positions in a plane.

[117] The technology is related to the use of computer system 400 for implementing the techniques described herein. According to one embodiment , those techniques are performed by computer system 400 in response to processor 404 executing one or more sequences of one or more instructions contained in main memory 406. Such instructions may be read into main memory 406 from another machine-readable medium, such as storage device 410. Execution of the sequences of instructions contained in main memory 406 causes processor 404 to perform the process steps described herein. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions to implement the technology. Thus, embodiments are not limited to any specific combination of hardware circuitry and software.

[118] The term "machine-readable medium" as used herein refers to any medium that participates in providing data that causes a machine to operation in a specific fashion. In an embodiment implemented using computer system 400, various machine-readable media are involved, for example, in providing instructions to processor 404 for execution. Such a medium may take many forms, including but not limited to, non-volatile media, volatile media. Nonvolatile media includes, for example, optical or magnetic disks, such as storage device 410. Volatile media includes dynamic memory, such as main memory 406. All such media are tangible to enable the instructions carried by the media to be detected by a physical mechanism that reads the instructions into a machine. The instructions may be stored in a non-transitory manner.

[119] Common forms of machine-readable media include, for example, hard disk, flash memory, optical disk, RAM, ROM, any other memory chip or cartridge, or any other medium from which a computer can read.

[120] Various forms of machine-readable media may be involved in carrying one or more sequences of one or more instructions to processor 404 for execution. For example, the instructions may initially be carried on a magnetic disk of a remote computer. The remote computer can load the instructions into its dynamic memory (e.g., RAM) and send the instructions over a communication medium, such as Ethernet or a wireless network such as LTE or WiFi. A local to computer system 400 can receive the data which may be stored in main memory 406, from which processor 404 retrieves and executes the instructions. The instructions received by main memory 406 may optionally be stored on storage device 410 either before or after execution by processor 404. [121] Computer system 400 also includes a communication interface 418 coupled to bus 402. Communication interface 418 provides a two-way data communication coupling to a network link 420 that is connected to a local network 422. For example, communication interface 418 may be a local area network (LAN) card to provide a data communication connection to a compatible LAN. Wireless links may also be implemented. In any such implementation, communication interface 418 sends and receives electrical, electromagnetic or optical signals that carry digital data streams representing various types of information.

[122] Network link 420 typically provides data communication through one or more networks to other data devices. For example, network link 420 may provide a connection through local network 422 to a host computer 424 or to data equipment operated by an Internet Service Provider (ISP) 426. ISP 426 in turn provides data communication services through the world wide packet data communication network now commonly referred to as the "Internet" 428. Local network 422 and Internet 428 both use electrical, electromagnetic or optical signals that carry digital data streams. The signals through the various networks and the signals on network link 420 and through communication interface 418, which carry the digital data to and from computer system 400, are exemplary forms of carrier waves transporting the information. [123] Computer system 400 can send messages and receive data, including program code, through the network(s), network link 420 and communication interface 418. In the Internet example, a server 430 might transmit a requested code for an application program through Internet 428, ISP 426, local network 422 and communication interface 418. [124] The received code may be executed by processor 404 as it is received, and/or stored in storage device 410, or other non- volatile storage for later execution.

[125] U.S. 2012/0173732, expressly incorporated herein by reference, discloses various embodiments of computer systems, the elements of which may be combined or subcombined according to the various permutations. [126] It is understood that this broad invention is not limited to the embodiments discussed herein, but rather is composed of the various combinations, subcombinations and permutations thereof of the elements disclosed herein, including aspects disclosed within the incorporated references. The invention is limited only by the following claims

[127] What is claimed is:

Claims

1. A method of processing EEG signals acquired in an electrically noisy

environment comprising an external stimulation of a brain which generates artifacts and stimulates brain activity reflected in EEG signals, comprising:

receiving EEG signals representing activity of the brain and the generated artifacts from a plurality of electrodes;

filtering a direct effect of the external stimulation on the EEG signals;

performing a dimensionality reduction on the EEG signals;

separating a first component of the reduced dimensionality EEG signals reflecting brain activity from a second component of the reduced dimensionality EEG signals reflecting residual artifacts and an indirect effect of the external stimulation; and

selectively outputting the first component of the reduced dimensionality EEG signals reflecting brain activity, wherein the artifacts generated by the external stimulation, but not the response of the brain to the external stimulation, are suppressed.

2. The method of claim 1, wherein said external stimulation comprises an LFMS procedure.

3. The method of claim 2, wherein at least one of said plurality of electrodes has an impedance of up to 60 kOhm.

4. The method of claim 2, wherein the EEG signals are sampled at a frequency between 0.5 and 5kHz.

5. The method of claim 2, wherein the LFMS procedure is conducted using a device coil housing positioned over the head of a user.

6. The method of claim 2, further comprising modifying the LFMS procedure based upon the modified set of EEG signals.

7. The method of claim 1, wherein said separation operation comprises a blind source separation operation.

8. The method of to claim 1, further comprising identifying residual non-filterable artifacts, and replacing the residual non-filterable artifacts with statistically normal signals.

9. A system for processing EEG signals, comprising:

an input configured to receive a set of EEG signals representing brain activity and artifacts from a plurality of electrodes, subject to an external stimulation which generates electrical artifacts and stimulates brain activity reflected in EEG signals;

a filter configured to filter a direct effect of the external stimulation on the EEG signals; and

at least one automated processor configured to:

reduce a dimensionality of the EEG signals;

perform a separation operation on the reduced dimensionality EEG signals; and output a modified set of EEG signals from which the artifacts generated by the external stimulation, but not the response of the brain to the external stimulation, are suppressed.

10. The system of claim 9, wherein said external stimulation comprises an LFMS procedure.

11. The system of claim 10, wherein at least one of said plurality of electrodes has an impedence of up to 60 kOhm.

12. The system of claim 10, wherein the EEG signals are sampled at a frequency between 0.5 and 5 kHz.

13. The system of claim 10, wherein the LFMS procedure is conducted using a device coil housing positioned over the head of a user.

14. The system of claim 10, further comprising modifying the LFMS procedure based upon the modified set of EEG signals.

15. The system of claim 9, wherein said separation operation comprises a blind source separation operation.

16. The system of to claim 9, wherein the at least one automate processor is further configured to also identify residual non-filterable artifacts, and replace the residual non-filterable artifacts with statistically normal signals.

17. A non-transitory computer-readable medium storing instructions for controlling a programmable processor to implement a method of processing EEG signals, comprising:

instructions for receiving a set of EEG signals representing brain activity and artifacts from a plurality of electrodes, subject to an external stimulation which generates electrical artifacts and stimulates brain activity reflected in EEG signals;

instructions for filtering a direct effect of the external stimulation on the EEG signals; instructions for performing a dimensionality reduction on the EEG signals;

instructions for performing a separation operation on the reduced dimensionality EEG signals; and

instructions for outputting a modified set of EEG signals from which the artifacts generated by the external stimulation, but not the response of the brain to the external stimulation, are suppressed.

18. The computer-readable medium of claim 17, wherein said external stimulation comprises an LFMS procedure.

19. The computer-readable medium of claim 18, wherein at least one of said plurality of electrodes has an impedence of up to 60 kOhm.

20. The computer-readable medium of claim 18, wherein the EEG signals are sampled at a frequency between 0.5 and 5 kHz.

21. The computer-readable medium of claim 18, wherein the LFMS procedure is conducted using a device coil housing positioned over the head of a user.

22. The computer-readable medium of claim 18, further comprising modifying the LFMS procedure based upon the modified set of EEG signals.

23. The computer-readable medium of claim 17, wherein said separation operation comprises a blind source separation operation.

24. The computer readable medium of to claim 17, further comprising instructions for identifying residual non- filterable artifacts, and replacing the residual non- filterable artifacts with statistically normal signals.