An efficient algorithm to improve Hardware -in -the -loop therapy for ADHD, based on evoked neural coherences.

Abstract

Various studies on ADHD patients come to the conclusion that in the spontaneous EEG the low frequency theta-band (4-7Hz) is increased, whereas the beta-(14-18 Hz), alpha-band and the sensorimotor rhythm (SMR) is decreased compared to healthy volunteers. [1;2;3]. Therefore the theta/beta-ratio is considered to be a good indicator for the diagnosis of ADHD also shown in standardized behavioral studies. [4].
It was therefore not too far fetched to utilize neurofeedback - training for ADHD treatments using the theta/beta-band ratio as a parameter at first.
But the application of neurofeedback methods for the ADHD therapy showed ambivalent results: whereas some studies concluded the evidence for the effectiveness of the method for children not responding to medical therapy, [5; 6; 7], other studies state no improvements of hardware – in – the – loop – technique compared to pharmaceutical treatment [8; 9]
Of a more serious concern regarding the neurofeedback method could possibly be that the method requires a degree of concentration and learning ability, which is actually lacking in this group.

Full text

An efficient algorithm to improve Hardware -in -the -loop therapy
for ADHD, based on evoked neural coherences
Matthias Schmidt, Dietmar Henrich & Tino Schmidt
University of Applied Sciences Zittau/Goerlitz, Institut of Applied Psycholoy, 02826 Goerlitz, Germany (corresponding author e-mail:
matthias.schmidt@hszg.de)
Technical University of Dresden, Institute of General Psychology, 01157 Dresden; Germany, e-mail: schmidtt@b-tu.de
Medical Engineering Department, Brandenburg Technical University 01968 Senftenberg, Germany and Technical University of Dresden,
NeuroImaging Center, 01157 Dresden, Germany (corresponding author e-mail: dietmar.henrich@b-tu.de).
I. INTRODUCTION
Various studies on ADHD patients come to the conclusion that in the spontaneous EEG the low frequency theta-
band (4-7Hz) is increased, whereas the beta-(14-18 Hz), alpha-band and the sensorimotor rhythm (SMR) is
decreased compared to healthy volunteers. [1;2;3]. Therefore the theta/beta-ratio is considered to be a good
indicator for the diagnosis of ADHD also shown in standardized behavioral studies. [4].
It was therefore not too far fetched to utilize neurofeedback - training for ADHD treatments using the theta/beta-
band ratio as a parameter at first.
But the application of neurofeedback methods for the ADHD therapy showed ambivalent results: whereas some
studies concluded the evidence for the effectiveness of the method for children not responding to medical
therapy, [5; 6; 7], other studies state no improvements of hardware – in – the – loop – technique compared to
pharmaceutical treatment [8; 9]
Of a more serious concern regarding the neurofeedback method could possibly be that the method requires a
degree of concentration and learning ability, which is actually lacking in this group.
II. EXTERNALIZING THE EFFECT OF NEUROFEEDBACK
A. Hardware – in – the – loop - stimulation
To overcome this drawback of the neurofeedback method an external stimulus of to the neural system could be
applied by using a hardware – in – the – loop - stimulation. The underlying thalamic to cortical mechanism is
considered proven. [10; 11; 12]. But the stimulation frequency is questionable. This might be the reason why this
method is not yet widely applied for the ADHD therapy, although some positive responses are reported
[13;14].
B. Patient adaption
The actual problem of hardware – in – the – loop - methods is, that they are not tuned to the individual ADHD
patient, which is disadvantageous compared to neurofeedback.
At a glance the two methods neurofeedback versus hardware – in – the – loop - stimulation might look very
similar: the first to try to transform a dysfunctional EEG into „normal“ EEG by active training of the patient and
the second utilizing an external stimulus in a more passive manner. The underlying neural mechanism for the
transformation through neurofeedback are not understood..
So far it seems clear that a more large scale transformation of the neural system is needed in order to achieve to
desired changes in the EEG, although from the outside it appears to be just an operant conditioning. [15].
Apparently with neurofeedback we change neuronal mechanism related to the ADHD symptomatics, but with
hardware – in – the – loop - stimulation using passive amplification of the low amplitude frequency bands we
enhance processes not really related to ADHD. At the end the „dysfunctional“ EEG-pattern represent but the
summation potentials of different origins.
The disadvantage in the existing therapies lacking patient adaption explains why hardware – in – the – loop -
stimulation methods show less efficiency compared to feedback method although they don’t rely on
concentration and learning abilities.
III. DISTURBANCE RELEVANT FREQUENCY COMPONENTS

A. Theoretical Background
The difficulty of every hardware – in – the – loop - stimulation is to influence the relevant frequency component.
Since ADHD patients show an increased theta-band (4-7 Hz) and a decreased beta-band (14-18 Hz) EEG activity
it seems obvious to solely stimulate the beta band and to thus co-activate more power in this spectral bands.
Astonishingly this does not lead to an improvement in the ADHD symptoms. Moreover it has to be considered
that the misaligned theta over beta ratio is by itself just a symptom of deeper biological correlates in the brain.
Some studies suggest the misaligned theta over beta ratio is due to a lacking coherence in the phase
synchronization of different cognitive feedback loops [16;17]. In other words the connectivity of more distance
brain areas is disturbed, because the signal pathway of the intermitting layers is getting out of rhythm. Therefore
the patients lack the ability of action control and correction tuning [18; 19].
An effective hardware – in – the – loop - stimulation method should primarily try to re-establish the
synchronization of the involved feedback loops and not only to recover the imbalances seen in the frequency
components purely in „mechanical“ fashion.
B. Evoked versus. Steady-state-coherences
The main problem is determining of the relevant disturbing frequencies causing a de-synchronization that may
result in an attention deficit.
Whereas frequency analysis requires time intervals in the order of several seconds to minutes and is therefore
suitable for steady-state coherences representing basic brain operations, the detection of the specific
incoherences a short term method is necessary in order to identify rapidly changing brain areas involved in the
stimulus processing
IV. EVOKED COHERENCES DETECTION METHOD
The EEG is continuously analyzed with time variant spectral decomposition showing both the amplitude and
frequency. This is in accordance tot he classical δ, ϑ, α andβ bands. Each frequency band is represented by a
sinusoidal oscillation.
The component related coherence values emerge from the cross-correlation of the corresponding oscillations of
various locations. Length and position oft he post-stimulus intervals are determined depending on the average
wavelength of the component under consideration, e.g. 0,5-2,5 average periods after the stimulus. To eliminate
random effects single stimulus evoked coherences are averaged in a series of equal stimuli.
Such a database can be obtained based on oddball paradigm, where patients should count sounds deviating from
a series of equal tones.
Based on of a time variant analysis a audiovisual stimulation should allow a real-time detection in the coherence
changes and to determine the changes in the cognitive system through hardware – in – the – loop - stimulation.
In such a case one would apply a stimulation but unlike in the normal hardware – in – the – loop - stimulation
method this time not with a fixed parameter set, but one adaptive through pre and post comparison thus
controlling how effective the stimulus is.. Moreover the time variant spectral analysis can be continued while
neurostimulating with dynamically adjusted parameters.
The procedure to detect significant changes in the coherence pattern of the EEG is described in the following
paragraph.
Assuming that at the beginning we are interested only at coherences at a single position p (p=1, .., P) we end up
with P-1 variables for each frequency band. An ICA component analysis would lead us to a reduction of M
principal components instead of P-1. Doing this for all possible positions p, we get P*M principal components
instead of P * (P-1)/2 primary variables for one electrode position. Due to the similarity of the electrical activity
in a close neighborhood, the principal components are therefore also very similar. A second order ICA is applied
where higher order components are determined such that their scores give different average values according to
different cognitive tasks, e.g. concentration versus relaxation. The principal components of second order 2Fk
(k=1,..,K) are derived from the first order components 1Fpm (m=1,..,M) via a multiplication of the corresponding
coefficients apm according to:
Based on this, it has to be clarified which electrode positions and which coherences gives the highest score
contributions. The interest is focused on the coherences instead of the principal components of the first degree.
Fpm in equation (1)
The principal components of the first order can be expressed in analogy to equation (1) as:

Where bpmq denotes the score factors from the first stage ICA and Cpq the coherence between electrode position p
and q (both being z-transformed)
Replacing (2) into (1) the result is:
The coefficients of the first order (b) and the second order (a) can be combined to a new coefficient g:
with
The principal components of the second order 2Fk results according to (4) with the partial contributions 2Fkp at the
electrode positions p:
The amplitude 2Fkpq of each single coherence between the electrode position p and q to 2Fkp is therefore:
(7) .
V. CONCLUSION
The detection of evoked coherences offers a technical feasible possibility assisting to alter the cognitive
processes activated autonomously during neurofeeedback.
The application of such an analytic method increases the efficiency of a neurofeedback training run in parallel.
Anyway in cases where neurofeedback is difficult due to lack in concentration and learning ability, like ADHD
or autism the analysis of evoked potentials for the control of hardware – in – the – loop - stimulation could be an
decisive therapeutical advantage.
REFERENCES
[1] J. R. Chabot, F. di Michele, & L.J.E.R. Prichep, „The clinical role of computerized EEG in the evaluation and treatment of learning and
attention disorders in children and adolescents,” in Journal of Neuropsychiatry and Clinical Neuroscience, 2001, Vol. 13(2), p. 171f.
[2] A. R. Clarke, R. J. Barry, R. McCarthy & M. Selikowitz, “Age and sex effects in the EEG: differences in two subtypes of attention-
deficit/hyperactivity disorder,” in: Clinical Neurophysiology, 2001 Vol. 112, pp. 815-826.
[3] M. Holtmann, C. Stadler, U. Leins, U. Strehl & F. Poustka, „Neurofeedback in der Behandlung der Aufmerksamkeitsdefizit-
Hyperaktivitätsstörung (ADHS),“ in: Zeitschrift für Kinder- und Jugendpsychiatrie und Psychotherapie, 2003, pp. 53-67.
[4] V.J. Monastra, J. F. Lubar, & M. Linden, „The development of a quantitative electroencephalographic scanning process for attention
deficit- hyperactivity disorder: Reliability and validity studies,” in Neuropsychology, 2001, Vol 15 (1), pp. 136 – 144.
[5] S. Othmer, S. F. Othmer, & S. M. Clifford, “EEG biofeedback training for attention deficit disorder, specific learning disabilities, and
associated conduct problems.” in EEG Spectrum, 1991.
[6] D. A. Kaiser, & S. Othmer, “Effect of neurofeedback on variables of attentio n in a large multi-center trial.” in Journal of Neurotherapy,
2000, vol. 4, pp. 5 – 15.
[7] R. DeBeus, J. D. Ball, M. E. deBeus & R. Herrington, “ Attention training with ADHD children: Preliminary findings in a double-blind
placebo-controlled study.” Presented at the Annual conference of the international society for neuronal regulation, 2003, Houston.
[8] D. E.Cobb & J. R. Evans, “The use of biofeedback techniques with school -aged children exhibiting behavioural and/ or learning
problems,” in: Journal of Abnormal Child Psychology, 1981, vol. 9 (2), pp. 251 – 281.
[9] S. W. Lee, “Biofeedback as a treatment for childhood hyperactivity: A critical review of the literature,” in: Psychological Report, 1991,
vol. 68 (1), pp. 163 – 192.
[10] S. Bartley, “Relation of intensity and duration of brief retinal stimulation by light to the electrical response to the optic cortex of the
rabbit,” in: American Journal of Physiology, 1934, vol. 108, pp. 397-408.
[11] J. McClintic, “Physiology of the human body,” 1978, John Whiley & Sons, New York, NY.
[12] D. Siever, “Handbook of Neurofeedback: Dynamics and Clinical Applications”, Chapter: Audiovisual entrainment: history, p hysiology,
and clinical studies,” 2007 Binghamton, NY: The Haworth Medical Press, pp. 155-183.

[13] J. G. Patrick, “Improved Neuronal Regulation in ADHD: A Application of Fifteen Sessions of Photic ‐ Driven EEG,” in: J. Neurother,
1996, vol. 1, pp. 27–36.
[14] M. Joyce & D.Siever, “Audio-visual entrainment program as a treatment for behavior disorders in a school setting,” Journal of
Neurotherapy, 2000, vol. 4(2): pp. 9-15.
[15] M. B. Sterman, “Physiological origins and functional correlates of EEG rhythmic activities: Implications for self-regulation” in:
Biofeedback and Self Regulation, 1996, vol. 21 (1), pp. 3 – 33.
[16] ,J.R. Chabot & G. Serfontein, “Quantitative Electroencephalographic profiles of children with attention deficit disorder. in: Biological
Psychiatry, 1996, vol. 40, pp. 951- 963.
[17] H. J. Volke, P. Dettmar, P. Richter, M. Rudolf, U. Buhss, “Evoced coherences of EEG in mental load: An investigation in chess
players,” in: Zeitschr. für Psychologie, 1999, vol. 207 (2), pp. 233-262.
[18] H. J. Volke, P. Dettmar, P. Richter, M. Rudolf, U. Buhss, “On-coupling and off-coupling of neocortical areas in chess experts and
novices as revealed by evoked EEG coherence measures and factor-based topological analysis—a pilot study” in: J. Psychophysiol., 2002,
vol. 16, pp. 23–36.
[19] R. A. Barry, A.R. Clarke, R. McCarthy & M. Selikowitz, “EEG coherence in attention-deficit/hyperactivity disorder: a comparative
study of two DSM-IV types,” in: Clinical Neurophysiology, 2002, vol. 113, pp. 579-585