- Email: [email protected]
Reliability of quantitative EEG (qEEG) measures and LORETA current source density at 30 days
Neuroscience Letters 518 (2012) 27–31
Contents lists available at SciVerse ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
Reliability of quantitative EEG (qEEG) measures and LORETA current source density at 30 days Rex L. Cannon a,b,∗ , Debora R. Baldwin a , Tiffany L. Shaw a,b , Dominic J. Diloreto a , Sherman M. Phillips a , Annie M. Scruggs a , Timothy C. Riehl a a b
Clinical Neuroscience, Self-regulation and Biological Psychology Laboratory, Department of Psychology, University of Tennessee, Knoxville, TN 37996, United States Cole Neuroscience Center, Bldg B., Suite 102, University of Tennessee Medical Center, Knoxville, TN 37920, United States
a r t i c l e
i n f o
Article history: Received 22 February 2012 Received in revised form 10 April 2012 Accepted 16 April 2012 Keywords: EEG reliability LORETA Current source density Neuroimaging reliability
a b s t r a c t There is a growing interest for using quantitative EEG and LORETA current source density in clinical and research settings. Importantly, if these indices are to be employed in clinical settings then the reliability of these measures is of great concern. Neuroguide (Applied Neurosciences) is sophisticated software developed for the analyses of power, and connectivity measures of the EEG as well as LORETA current source density. To date there are relatively few data evaluating topographical EEG reliability contrasts for all 19 channels and no studies have evaluated reliability for LORETA calculations. We obtained 4 min eyes-closed and eyes-opened EEG recordings at 30-day intervals. The EEG was analyzed in Neuroguide and FFT power, coherence and phase was computed for traditional frequency bands (delta, theta, alpha and beta) and LORETA current source density was calculated in 1 Hz increments and summed for total power in eight regions of interest (ROI). In order to obtain a robust measure of reliability we utilized a random effects model with an absolute agreement definition. The results show very good reproducibility for total absolute power and coherence. Phase shows lower reliability coefficients. LORETA current source density shows very good reliability with an average 0.81 for ECB and 0.82 for EOB. Similarly, the eight regions of interest show good to very good agreement across time. Implications for future directions and use of qEEG and LORETA in clinical populations are discussed. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Quantitative EEG (qEEG) comprises computerized imaging and statistical procedures to aid in the detection of abnormal patterns associated with specific pathological conditions and normative patterns found in different cognitive and affective conditions [2]. qEEG is a direct signature of neural activity and provides ideal temporal resolution in the millisecond time domain [9]. Additionally, qEEG and low-resolution electromagnetic brain tomography (LORETA) [17,18] provide a method to evaluate neural mechanisms associated with the experience of a stimulus [4]. Reliability of neuroimaging methods is extremely important to the potential development of biomarkers and change indexes in stable mechanisms that may signify treatment response or decline [27]. qEEG power and connectivity measures, with the addition of LORETA, are important methodologies for demonstrating direct associations between psychiatric conditions and symptoms
∗ Corresponding author at: Clinical Neuroscience, Self-regulation and Biological Psychology Laboratory, Department of Psychology, University of Tennessee, Knoxville, TN 37996, United States. Tel.: +1 865 300 4983. E-mail address: [email protected] (R.L. Cannon). 0304-3940/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.neulet.2012.04.035
associated with neurologic functions [5,7,14] as well as monitoring pharmacological effects [21] and treatment outcomes [20]. Furthermore, due to its noninvasive properties and resource effectiveness, qEEG affords the opportunity to examine the brain’s electrical activity during longer periods in variable experimental conditions [4]. The validity and reliability of qEEG has been an area of concentrated study. A large number of studies have demonstrated the reliability and validity of qEEG methods [10]. Therefore, many arguments opposing the computerized analyses of EEG may not carry much validity, especially considering no existing studies published to date show significant reliability of non-computerized analysis of the EEG signal outside of indentifying epileptiform activity [1]. qEEG, therefore, is suggested to be clinically useful for revealing additional signs of brain dysfunction in individual patients and a valuable research tool for revealing statistical differences between groups [11]. To date, there are relatively few data evaluating topographical EEG reliability contrasts for all 19 channels, coherence and phase, and no studies have evaluated reliability for LORETA calculations. LORETA is a collection of independent modules run in specific sequence to transform raw EEG signal into LORETA images. It is one of the most widely used algorithms to address the inverse solution for source localization of the EEG produced on the scalp [22]. The
28
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31
methods of qEEG and LORETA current source density (CSD) comparisons [24], with Neuroguide (Applied Neuroscience Laboratories), permit a normative database comparison for an individual with age-similar groupings of the estimated intracerebral CSD distribution using LORETA [25]. LORETA [18] has received considerable validation from studies that combined this method with more established localization methods, including fMRI [6], structural MRI [8], PET [16], and invasive implanted electrode recordings [28]. It has also been demonstrated that LORETA can correctly localize deep structures, such as the anterior cingulate cortex [19] and mesial temporal lobes [28], without localization bias [17]. The current study sought to determine the reliability of Fast Fourier Transform (FFT) absolute power, phase, coherence and LORETA calculations using Neuroguide at 30-day intervals. Since the use of qEEG has increased in clinical and research settings and a reliable measure is needed across disciplines, this topic proves important.
2. Methods This study was conducted with 19 normative (7 female), undergraduate students with a mean age of 20.74, SD = 1.24. One participant was left handed. All participants read, signed and agreed to informed consent approved by the University of Tennessee Institutional Review Board. Exclusion criteria included psychiatric or neurological syndromes, history of head trauma, epilepsy or neurovascular incidents, and use of alcohol or drugs within 2 weeks of the data collection procedures. All participants received extra course credit for participation in this study. After obtaining informed consent, participants were advised of all procedures and equipment. The ears and forehead were cleaned for recording with a mild abrasive gel (NuPrep) to remove any oil and dirt from the skin. The head was measured and marked prior to EEG recording using a measure of head circumference and the distance between the nasion and inion to determine the appropriate cap size for recording and placement of frontal electrodes. After fitting caps, each electrode site was injected with electrogel and prepared so that impedances between individual electrodes and each ear remain <10 k. Impedances were <2 k between any two electrodes. The EEG was recorded at 256 samples/s with linked-ear referenced using 9 mm tin cups. The participants were recorded in 4 min eyes-closed (ECB) and eyes-opened (EOB) baseline conditions using the Deymed Truscan EEG Acquisition System (Deymed Diagnostics) at exactly 30-day intervals; the time of day was constant across recordings. The EEG data were exported from storage and entered into Neuroguide for analysis. Neuroguide features an automatic selection method based on a contrast selection created by the user. For all participant records, at time 1 and time 2, a mean record length for automatic selection in Neuroguide of 6 s was obtained across individuals. These 6-s segments were utilized to minimize differences in selection bias between baseline conditions. Neuroguide also computes the reliability for the selected sections which was set to ≥0.90. Only these automatic segments were used for analyses – no extra epochs were added. Neurobatch analysis was conducted and data were extrapolated from the output and entered into SPSS 19 for reliability analysis. This study compared all electrode pairs (coherence and phase) and measures of absolute power of all electrodes for the reliability analysis. Neuroguide computes LORETA absolute power in 1 Hz increments, which were evaluated similarly. To test reliability, we used a random mixed model with an absolute agreement definition. The conceptual difference between absolute agreement and consistency definitions begins by noting their formal distinction, which is in the definition of the intraclass correlation coefficient (ICC) denominator. For consistency
Table 1 Paired t-test between length of recordings for times 1 and 2 in mean seconds of EEG. Although a noticeable difference exists between the amount of usable data between 30-day intervals, these differences did not reach significance with ECB t(18) = −1.69, p = 0.108 and EOB t(18) = −1.71, p = 0.105. Mean
N
SD
Pair 1
ECB time 1 ECB time 2
90.85 112.04
19 19
39.79 39.67
9.13 9.10
Pair 2
EOB time 1 EOB time 2
102.88 122.55
19 19
46.50 45.52
10.67 10.44
SE
measures, column variance is excluded from denominator variance, and for absolute agreement, it is not. Column variance is excluded from the denominators of consistency measures because it is deemed an irrelevant source of variance. In the case of absolute agreement when measurements disagree in absolute value, regardless of the reason, they are viewed as disagreements [15]. EEG frequency domains were contrasted for 1 Hz increments and for typical frequency bands (e.g., delta 1.0–4.0, theta 4.0–8.0, alpha 8.0–12.0 and beta 12.0–32.0). 3. Results The results for reliability contrasts are shown in Tables 1–6. Table 1 shows the contrast for the mean length of EEG records at time 1 and time 2 for ECB and EOB recordings. There is no significant difference between time 1 and 2. However, time 2 shows more applicable data overall. Table 2 shows results of reliability measures for all topographical measures. From left to right are the measure, alpha, ICC, F and degrees of freedom, and top to bottom are the results for absolute power, coherence for EOB/ECB and phase ECB/EOB. The results for absolute power and coherence show good to very good agreement for total power and frequency-wise contrasts. FFT phase shows low to moderate reproducibility for total power and frequency-wise contrasts. Table 3 shows the results for total absolute power at each individual electrode in the EOB condition. The results show good to very good agreement with the exception of C4, F8, T3 and T6. Table 4 shows the results for reliability contrasts for total absolute power at Table 2 Topographical reliability contrasts for absolute power and connectivity measures. In the table from left to right are the measure, Chronbach’s alpha, intraclass correlation coefficients, F for the contrast and the degrees of freedom. Measure (time 1 = time 2) ECB FFT absolute power EOB FFT absolute power ECB FFT coherence Delta Theta Alpha Beta EOB FFT coherence Delta Theta Alpha Beta ECB phase lag deg Delta Theta Alpha Beta EOB phase lag deg Delta Theta Alpha Beta **
Significance < 0.000.
Alpha
ICC
F
df **
0.91 0.78
0.9 0.77
10.51 4.53**
1.443 1.443
0.91 0.95 0.96 0.97
0.91 0.95 0.96 0.96
11.64 21.57** 25.88** 29.83**
3.248 3.248 3.248 3.248
0.85 0.88 0.9 0.92
0.85 0.87 0.89 0.91
6.67** 8.25** 9.71** 12.12**
3.248 3.248 3.248 3.248
0.064 0.29 0.74 0.58
0.064 0.29 0.74 0.58
1.06** 1.42** 3.84** 2.39**
3.248 3.248 3.248 3.248
0.12 0.4 0.4 0.39
0.12 0.4 0.4 0.39
1.38** 1.67** 1.66** 1.64**
3.248 3.248 3.248 3.248
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31 Table 3 Results for total absolute power at individual electrodes in the EOB condition (time 1 = time 2). In the table from left to right are the electrode, the intraclass correlation coefficient, Chronbach’s alpha and the F for the contrast. The average alpha for EOB is 0.77. These contrasts used 75 degrees of freedom (total power for delta, theta, alpha and beta or 4 frequency bands × 19 participants). Electrode
ICC
Alpha
F
C3 C4 CZ F3 F4 F7 F8 FP1 FP2 FZ O1 O2 P3 P4 PZ T3 T4 T5 T6
0.84 0.68 0.86 0.81 0.87 0.81 0.59 0.71 0.80 0.83 0.77 0.76 0.81 0.69 0.76 0.62 0.76 0.77 0.67
0.85 0.69 0.88 0.81 0.88 0.81 0.60 0.72 0.81 0.85 0.79 0.78 0.82 0.71 0.77 0.62 0.77 0.78 0.67
6.5** 3.17** 8.07** 5.24** 8.61** 5.19** 2.48** 3.53** 5.32** 6.65** 4.72** 4.48** 5.55** 3.42** 4.43** 2.6** 4.26** 4.57** 3.04**
**
Significance < 0.000.
individual electrodes in the ECB condition. The agreement for total power is good to very good with the exception of F8. Table 5 shows reliability coefficients at time 1 and time 2 for ECB (left) and EOB (right) for LORETA current source density in 1 Hz frequency bins. Considering the computational load and number of contrasts, the agreement for comparisons are good to very good for each 1 Hz frequency bin for all 2394 voxels. Table 6 shows the reliability results for 8 individual regions of interest within LORETA space. In the table from left to right are the region, Talairach coordinates, alpha, ICC and F for the contrast. Table 7 shows the reliability results for the same 8 regions used in Table 6 for ECB recordings. Individual ROIs show good to very good reliability with the voxel in left anterior cingulate (0.65) and left prefrontal cortex (0.67) showing the lowest reliability coefficients in ECB.
Table 4 Results for total absolute power at individual electrodes in the ECB condition (time 1 = time 2). In the table from left to right are the electrode, the intraclass correlation coefficient, Chronbach’s alpha and the F for the contrast. The average alpha for ECB is 0.85. These contrasts used 75 degrees of freedom (total power for delta, theta, alpha and beta or 4 frequency bands × 19 participants). Electrode
ICC
Alpha
F
C3 C4 CZ F3 F4 F7 F8 FP1 FP2 FZ O1 O2 P3 P4 PZ T3 T4 T5 T6
0.87 0.84 0.9 0.84 0.86 0.87 0.6 0.87 0.86 0.83 0.93 0.96 0.86 0.87 0.89 0.75 0.78 0.84 0.78
0.87 0.84 0.9 0.84 0.86 0.87 0.6 0.87 0.86 0.83 0.93 0.96 0.86 0.87 0.89 0.75 0.78 0.84 0.83
7.53** 6.22** 9.72** 6.23** 7.2** 7.54** 2.44** 7.6** 7.23** 5.77** 14.67** 25.4** 7.33** 7.76** 9.17** 4.02** 4.94** 6.19** 6.09**
LORETA absolute power. ** Significance < 0.000.
29
Table 5 In the table are the results for the reliability analyses for absolute power LORETA current source density for all 2394 7 mm3 voxels in 1 Hz increments for ECB and EOB. Shown in each table are the frequency domain, intraclass correlation coefficient, Chronbach’s alpha and the F for the ANOVA. The average reliability for ECB = 0.81 (left) and EOB = 0.82 (right) at 30 days. The specific contrasts between time 1 and time 2 were conducted with 45,486 − 1 (45,485) degrees of freedom (e.g., 2394 × 19). Frequency 1 Hz 2 Hz 3 Hz 4 Hz 5 Hz 6 Hz 7 hz 8 Hz 9 Hz 10 Hz 11 Hz 12 Hz 13 Hz 14 Hz 15 Hz 16 Hz 17 Hz 18 Hz 19 Hz 20 Hz 21 Hz 22 Hz 23 Hz 24 Hz 25 Hz 26 Hz 27 Hz 28 Hz 29 Hz 30 Hz **
ICC 0.76 0.78 0.77 0.78 0.77 0.76 0.81 0.84 0.9 0.9 0.84 0.81 0.76 0.78 0.79 0.81 0.81 0.85 0.85 0.86 0.86 0.87 0.86 0.85 0.85 0.84 0.84 0.82 0.81 0.82
Alpha 0.76 0.78 0.77 0.78 0.77 0.76 0.81 0.85 0.9 0.9 0.84 0.81 0.76 0.78 0.79 0.81 0.81 0.85 0.85 0.86 0.86 0.87 0.86 0.85 0.85 0.83 0.84 0.83 0.81 0.82
F **
4.23 4.5** 4.38** 4.46** 4.42** 4.11** 5.23** 6.5** 10.15** 9.55** 6.4** 5.16** 4.19** 4.54** 4.83** 5.23** 5.34** 6.63** 6.78** 6.95** 7.36** 7.45** 7.11** 6.75** 6.49** 6.05** 6.17** 5.7** 5.27** 5.66**
Frequency
ICC
Alpha
F
1 Hz 2 Hz 3 Hz 4 Hz 5 Hz 6 Hz 7 hz 8 Hz 9 Hz 10 Hz 11 Hz 12 Hz 13 Hz 14 Hz 15 Hz 16 Hz 17 Hz 18 Hz 19 Hz 20 Hz 21 Hz 22 Hz 23 Hz 24 Hz 25 Hz 26 Hz 27 Hz 28 Hz 29 Hz 30 Hz
0.84 0.83 0.82 0.81 0.8 0.78 0.76 0.78 0.81 0.82 0.87 0.87 0.82 0.8 0.77 0.77 0.75 0.78 0.77 0.75 0.77 0.78 0.79 0.8 0.8 0.8 0.82 0.81 0.81 0.8
0.85 0.84 0.83 0.82 0.81 0.79 0.79 0.8 0.83 0.83 0.88 0.88 0.83 0.81 0.78 0.76 0.76 0.77 0.76 0.76 0.77 0.78 0.79 0.8 0.8 0.8 0.81 0.81 0.81 0.8
6.67** 6.22** 5.97** 5.66** 5.34** 4.79** 4.82** 4.99** 5.83** 5.81** 8.01** 8.01** 5.72** 5.26** 4.48** 4.44** 4.15** 4.3** 4.27** 4.17** 4.41** 4.6** 4.84** 5.1** 5.05** 5.05** 5.47** 5.37** 5.38** 4.99**
Significance < 0.000.
4. Discussion In general, findings demonstrate the reliability of computerized analysis of EEG show good to very good reproducibility at 30 days. Similarly, LORETA calculations for 1 Hz increments across time show similar robust agreement, as do the individual regions of interest. The absolute power coefficients show good to very good reliability for ECB and EOB across the 30-day time interval for the eyes-closed baseline for all contrasts. These results apply to total power and frequency-wise contrasts. Hence, the data demonstrate good to very good reproducibility of qEEG measures and LORETA
Table 6 Reliability for 8 regions of interest for total absolute power for EOB 1 = EOB 2. In the table from left to right are the neuroanatomical regions of interest with Brodmann Area designation (BA) and abbreviation (e.g., left anterior cingulate = ACC), (voxel 7 mm3 ), x, y and z coordinates in Talairach space, Chronbach’s alpha, intraclass correlation coefficient and the F for the contrast. Voxel
Coordinates
Alpha
ICC
F
Left anterior cingulate gyrus (BA 32) ACC Left precuneus (BA 19) LPC Left prefrontal (BA 8) LPFC Right prefrontal (BA 8) RPFC Right postcentral gyrus (BA 40) RPCG Left supramarginal gyrus (BA 40) LSMG Right supramarginal gyrus (BA 40) RSMG Left cuneus (BA 7)
(−3, 31, 29)
0.98
0.98
43.42**
(−31, −81, 22) (−38, 31, 43) (39, 24, 43) (53, −18, 43)
0.94 0.95 0.97 0.95
0.94 0.95 0.97 0.95
17.05** 18.59** 31.57** 18.18**
(−59, −53, 29)
0.93
0.93
15.15**
(60, −53, 29)
0.92
0.92
13.12**
(−3, −67, 29)
0.91
0.91
10.84**
**
Significance < 0.000.
30
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31
Table 7 Reliability for 8 regions of interest for total absolute power for ECB 1 = ECB 2. In the table from left to right are the neuroanatomical regions of interest with Brodmann Area designation (BA) and abbreviation (e.g., left prefrontal cortex = LPFC) (voxel 7 mm3 ), x, y and z coordinates in Talairach space, Chronbach’s alpha, intraclass correlation coefficient and the F for the contrast. Voxel
Coordinates
Alpha
ICC
F
Left anterior cingulate gyrus (BA 32) ACC Left precuneus (BA 19) LPC Left prefrontal (BA 8) LPFC Right prefrontal (BA 8) RPFC Right postcentral gyrus (BA 40) RPCG Left supramarginal gyrus (BA 40) LSMG Right supramarginal gyrus (BA 40) RSMG Left cuneus (BA 7)
(−3, 31, 29)
0.65
0.65
2.89**
(−31, −81, 22) (−38, 31, 43) (39, 24, 43) (53, −18, 43)
0.83 0.67 0.89 0.84
0.83 0.65 0.88 0.84
5.89** 2.91** 8.85** 6.21**
(−59, −53, 29)
0.86
0.86
7.16**
(60, −53, 29)
0.85
0.85
6.49**
(−3, −67, 29)
0.86
0.86
7.05**
**
Significance < 0.000.
current source density over a 30-day interval for this study population of clinically normal individuals. The coherence results between all electrode pairs at 30 days shows very good reliability for ECB and good to very good results of EOB for each frequency domain. Since the degree of synchrony in EEG signals is commonly characterized by time-series measures, it is important for these measures to be reliable and consistent across studies since reference site may produce inaccurate results [13]. Coherence is a statistic of phase differences and yields a much finer measure of shared energy between mixtures of periodic signals than can be achieved using the Pearson product-moment correlation coefficient of amplitudes (or more simply functional integration or differentiation) [5,23,25]. Thus, coherence may be a potential target for the evaluation of traumatic brain injury, psychological syndromes or concussion [23,25,26]. Of all contrasts conducted in this work, FFT phase shows the lowest reliability coefficients. Reasons for this may be three fold. First, subtle artifacts may produce effects on the phase relationship between any two signals. Second, this may be a function of learning and experience. The prior experience with baseline procedures may have reduced the neural power necessary to reach the selfregulatory state of baseline. Several fMRI studies demonstrate that during conditioning phase synchrony develops between neuronal populations [12]. This is certainly a topic for future study. Finally, this may be the result of automatic selection procedures from which extra EEG segments could be selected and added by the examiner. The CSD for 1 Hz frequency bins across all 2394 voxels shows very good reliability with 0.81 for ECB and 0.82 for EOB. Similarly, eight regions (single voxels 7 mm3 ) show good to very good reliability except for the anterior cingulate (0.65) and left prefrontal (0.67) in ECB condition. Even though these are low compared to the remaining regional voxels, they are still in the good range. The current data show good to very good reliability for qEEG power and coherence measures as computed by Neuroguide software. LORETA current source density shows good to very good agreement for baselines across time for 1 Hz frequency bins as well as individual ROIs over time. LORETA, as such, is the only neuroimaging technique to produce these results for baseline measures to date. Other imaging techniques have attempted to evaluate reliability across time for specific tasks with variable results [3]. Additions may be added to the current data and will be the focus of subsequent works. Higher frequency bands (e.g., high-beta and gamma) will be examined in future works in both topographical and LORETA contrasts. The current population was from a convenience sample in a university population, which may not represent the general population. Similarly, reliability measures may be taken
throughout the day to determine potential differences in the EEG relative to time of day. Similarly, we did not include Neuroguide automatic eye-movement and electrocardiograph (EKG) artifact channels. It should also be noted that an extensive visual inspection of all automatic selection samples was not completed; since artifacting programs are not regarded as stand alone without additional inspection by qualified individuals, this may have influenced the results. Further, if longer samples were obtained (e.g., 10–15 s) with a higher reliability criteria (e.g., 0.94 or 0.95), then results may have differed respectively. These are topics of future studies. The stability of electrical brain activity is a topic of intense interest given its potential importance in developing diagnostic biomarkers and as a mechanism to evaluate treatment outcomes. Thus, many arguments against the use of qEEG and LORETA in clinical diagnostic procedures may not hold much weight. The challenge becomes how to find specific, characteristic patterns of neural activity that best fit any particular psychiatric syndrome, traumatic brain injury or other learning issues and their neurological correlates. At present, our lab is addressing some of these issues and much more research is certainly needed for future developments. References [1] S.R. Benbadis, W.C. LaFrance, G.D. Papandonatos Jr., K. Korabathina, K. Lin, H.C. Kraemer, Interrater reliability of EEG-video monitoring, Neurology 73 (2009) 843–846. [2] M.H. Bjork, T. Sand, G. Brathen, O.M. Linaker, G. Morken, B.M. Nilsen, A.E. Vaaler, Quantitative EEG findings in patients with acute, brief depression combined with other fluctuating psychiatric symptoms: a controlled study from an acute psychiatric department, BMC Psychiatry 8 (2008) 89. [3] G.G. Brown, D.H. Mathalon, H. Stern, J. Ford, B. Mueller, D.N. Greve, G. McCarthy, J. Voyvodic, G. Glover, M. Diaz, E. Yetter, I.B. Ozyurt, K.W. Jorgensen, C.G. Wible, J.A. Turner, W.K. Thompson, S.G. Potkin, Multisite reliability of cognitive BOLD data, Neuroimage 54 (2011) 2163–2175. [4] R. Cannon, D. Baldwin, J. Lubar, Self-perception and experiential schemata in the addicted brain, Applied Psychophysiology and Biofeedback 33 (2008) 223–238. [5] R.L. Cannon, M. Congedo, J.F. Lubar, T. Hutchens, Differentiating a network of executive attention: LORETA neurofeedback in anterior cingulate and dorsolateral prefrontal cortices, International Journal of Neuroscience 119 (2009) 404–441. [6] R. Cannon, C. Kerson, A. Hampshire, sLORETA and fMRI detection of medial prefrontal default network anomalies in adult ADHD, Journal of Neurotherapy 15 (2011) 358–373. [7] R. Cannon, J. Lubar, M. Congedo, K. Thornton, K. Towler, T. Hutchens, The effects of neurofeedback training in the cognitive division of the anterior cingulate gyrus, International Journal of Neuroscience 117 (2007) 337–357. [8] R.L. Cannon, M.K. Crane, P.D. Campbell, J.H. Dougherty, D.R. Baldwin, J.D. Effler, L.S. Phillips, F. Hare, M. Zachary, K.E. Cox, D.J. Di Loreto, A 9-year-old boy with multifocal encephalomalacia: EEG Loreta and lifespan database, magnetic resonance imaging and neuropsychological agreement, Journal of Neurotherapy 15 (2011) 3–17. [9] K.L. Coburn, E.C. Lauterbach, N.N. Boutros, K.J. Black, D.B. Arciniegas, C.E. Coffey, The value of quantitative electroencephalography in clinical psychiatry: a report by the Committee on Research of the American Neuropsychiatric Association, Journal of Neuropsychiatry and Clinical Neurosciences 18 (2006) 460–500. [10] M. Corsi-Cabrera, L. Galindo-Vilchis, Y. del-Rio-Portilla, C. Arce, J. Ramos-Loyo, Within-subject reliability and inter-session stability of EEG power and coherent activity in women evaluated monthly over nine months, Clinical Neurophysiology 118 (2007) 9–21. [11] P. Coutin-Churchman, Y. Anez, M. Uzcategui, L. Alvarez, F. Vergara, L. Mendez, R. Fleitas, Quantitative spectral analysis of EEG in psychiatry revisited: drawing signs out of numbers in a clinical setting, Clinical Neurophysiology 114 (2003) 2294–2306. [12] S. Grossberg, M. Versace, Spikes, synchrony, and attentive learning by laminar thalamocortical circuits, Brain Research 1218 (2008) 278–312. [13] S. Hu, M. Stead, Q. Dai, G.A. Worrell, On the recording reference contribution to EEG correlation, phase synchrony, and coherence, IEEE Transactions on Systems Man and Cybernetics Part B: Cybernetics 40 (2010) 1294–1304. [14] J.R. Hughes, E.R. John, Conventional and quantitative electroencephalography in psychiatry, Journal of Neuropsychiatry and Clinical Neurosciences 11 (1999) 190–208. [15] K.O. McGraw, S.P. Wong, Forming inferences about some intraclass correlation coefficients, Psychological Methods 1 (1996) 30–46. [16] T.R. Oakes, D.A. Pizzagalli, A.M. Hendrick, K.A. Horras, C.L. Larson, H.C. Abercrombie, S.M. Schaefer, J.V. Koger, R.J. Davidson, Functional coupling of simultaneous electrical and metabolic activity in the human brain, Human Brain Mapping 21 (2004) 257–270.
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31 [17] R.D. Pascual-Marqui, Standardized low-resolution brain electromagnetic tomography (sLORETA): technical details, Methods and Findings in Experimental and Clinical Pharmacology 24 (Suppl. D) (2002) 5–12. [18] R.D. Pascual-Marqui, C.M. Michel, D. Lehmann, Low resolution electromagnetic tomography: a new method for localizing electrical activity in the brain, International Journal of Psychophysiology 18 (1994) 49–65. [19] D.A. Pizzagalli, T.R. Oakes, R.J. Davidson, Coupling of theta activity and glucose metabolism in the human rostral anterior cingulate cortex: an EEG/PET study of normal and depressed subjects, Psychophysiology 40 (2003) 939–949. [20] L.S. Prichep, K.R. Alper, L. Sverdlov, S.C. Kowalik, E.R. John, H. Merkin, M.L. Tom, B. Howard, M.S. Rosenthal, Outcome related electrophysiological subtypes of cocaine dependence, Clinical Electroencephalography 33 (2002) 8–20. [21] B. Saletu, G. Saletu-Zyhlarz, P. Anderer, N. Brandstatter, R. Frey, G. Gruber, G. Klosch, M. Mandl, J. Grunberger, L. Linzmayer, Nonorganic insomnia in generalized anxiety disorder. 2. Comparative studies on sleep, awakening, daytime vigilance and anxiety under lorazepam plus diphenhydramine (Somnium) versus lorazepam alone, utilizing clinical, polysomnographic and EEG mapping methods, Neuropsychobiology 36 (1997) 130–152. [22] S.C. Sanei, J.A. EEG Signal Processing, John Wiley & Sons, Ltd., West Sussex, 2007.
31
[23] R.W. Thatcher, C. Biver, J.F. Gomez, D. North, R. Curtin, R.A. Walker, A. Salazar, Estimation of the EEG power spectrum using MRI T(2) relaxation time in traumatic brain injury, Clinical Neurophysiology 112 (2001) 1729–1745. [24] R.W. Thatcher, C.J. Biver, D.M. North, Quantitative EEG and the Frye and Daubert standards of admissibility, Clinical Electroencephalography 34 (2003) 39–53. [25] R.W. Thatcher, D. North, C. Biver, Evaluation and validity of a LORETA normative EEG database, Clinical EEG and Neuroscience 36 (2005) 116–122. [26] R.W. Thatcher, D.M. North, R.T. Curtin, R.A. Walker, C.J. Biver, J.F. Gomez, A.M. Salazar, An EEG severity index of traumatic brain injury, Journal of Neuropsychiatry and Clinical Neurosciences 13 (2001) 77–87. [27] S.U. Zuidema, A.L. Buursema, M.G. Gerritsen, K.C. Oosterwal, M.M. Smits, R.T. Koopmans, J.F. de Jonghe, Assessing neuropsychiatric symptoms in nursing home patients with dementia: reliability and Reliable Change Index of the Neuropsychiatric Inventory and the Cohen-Mansfield Agitation Inventory, International Journal of Geriatric Psychiatry 26 (2011) 127–134. [28] D. Zumsteg, D.M. Andrade, R.A. Wennberg, Source localization of small sharp spikes: low resolution electromagnetic tomography (LORETA) reveals two distinct cortical sources, Clinical Neurophysiology 117 (2006) 1380–1387.
Contents lists available at SciVerse ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
Reliability of quantitative EEG (qEEG) measures and LORETA current source density at 30 days Rex L. Cannon a,b,∗ , Debora R. Baldwin a , Tiffany L. Shaw a,b , Dominic J. Diloreto a , Sherman M. Phillips a , Annie M. Scruggs a , Timothy C. Riehl a a b
Clinical Neuroscience, Self-regulation and Biological Psychology Laboratory, Department of Psychology, University of Tennessee, Knoxville, TN 37996, United States Cole Neuroscience Center, Bldg B., Suite 102, University of Tennessee Medical Center, Knoxville, TN 37920, United States
a r t i c l e
i n f o
Article history: Received 22 February 2012 Received in revised form 10 April 2012 Accepted 16 April 2012 Keywords: EEG reliability LORETA Current source density Neuroimaging reliability
a b s t r a c t There is a growing interest for using quantitative EEG and LORETA current source density in clinical and research settings. Importantly, if these indices are to be employed in clinical settings then the reliability of these measures is of great concern. Neuroguide (Applied Neurosciences) is sophisticated software developed for the analyses of power, and connectivity measures of the EEG as well as LORETA current source density. To date there are relatively few data evaluating topographical EEG reliability contrasts for all 19 channels and no studies have evaluated reliability for LORETA calculations. We obtained 4 min eyes-closed and eyes-opened EEG recordings at 30-day intervals. The EEG was analyzed in Neuroguide and FFT power, coherence and phase was computed for traditional frequency bands (delta, theta, alpha and beta) and LORETA current source density was calculated in 1 Hz increments and summed for total power in eight regions of interest (ROI). In order to obtain a robust measure of reliability we utilized a random effects model with an absolute agreement definition. The results show very good reproducibility for total absolute power and coherence. Phase shows lower reliability coefficients. LORETA current source density shows very good reliability with an average 0.81 for ECB and 0.82 for EOB. Similarly, the eight regions of interest show good to very good agreement across time. Implications for future directions and use of qEEG and LORETA in clinical populations are discussed. © 2012 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Quantitative EEG (qEEG) comprises computerized imaging and statistical procedures to aid in the detection of abnormal patterns associated with specific pathological conditions and normative patterns found in different cognitive and affective conditions [2]. qEEG is a direct signature of neural activity and provides ideal temporal resolution in the millisecond time domain [9]. Additionally, qEEG and low-resolution electromagnetic brain tomography (LORETA) [17,18] provide a method to evaluate neural mechanisms associated with the experience of a stimulus [4]. Reliability of neuroimaging methods is extremely important to the potential development of biomarkers and change indexes in stable mechanisms that may signify treatment response or decline [27]. qEEG power and connectivity measures, with the addition of LORETA, are important methodologies for demonstrating direct associations between psychiatric conditions and symptoms
∗ Corresponding author at: Clinical Neuroscience, Self-regulation and Biological Psychology Laboratory, Department of Psychology, University of Tennessee, Knoxville, TN 37996, United States. Tel.: +1 865 300 4983. E-mail address: [email protected] (R.L. Cannon). 0304-3940/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.neulet.2012.04.035
associated with neurologic functions [5,7,14] as well as monitoring pharmacological effects [21] and treatment outcomes [20]. Furthermore, due to its noninvasive properties and resource effectiveness, qEEG affords the opportunity to examine the brain’s electrical activity during longer periods in variable experimental conditions [4]. The validity and reliability of qEEG has been an area of concentrated study. A large number of studies have demonstrated the reliability and validity of qEEG methods [10]. Therefore, many arguments opposing the computerized analyses of EEG may not carry much validity, especially considering no existing studies published to date show significant reliability of non-computerized analysis of the EEG signal outside of indentifying epileptiform activity [1]. qEEG, therefore, is suggested to be clinically useful for revealing additional signs of brain dysfunction in individual patients and a valuable research tool for revealing statistical differences between groups [11]. To date, there are relatively few data evaluating topographical EEG reliability contrasts for all 19 channels, coherence and phase, and no studies have evaluated reliability for LORETA calculations. LORETA is a collection of independent modules run in specific sequence to transform raw EEG signal into LORETA images. It is one of the most widely used algorithms to address the inverse solution for source localization of the EEG produced on the scalp [22]. The
28
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31
methods of qEEG and LORETA current source density (CSD) comparisons [24], with Neuroguide (Applied Neuroscience Laboratories), permit a normative database comparison for an individual with age-similar groupings of the estimated intracerebral CSD distribution using LORETA [25]. LORETA [18] has received considerable validation from studies that combined this method with more established localization methods, including fMRI [6], structural MRI [8], PET [16], and invasive implanted electrode recordings [28]. It has also been demonstrated that LORETA can correctly localize deep structures, such as the anterior cingulate cortex [19] and mesial temporal lobes [28], without localization bias [17]. The current study sought to determine the reliability of Fast Fourier Transform (FFT) absolute power, phase, coherence and LORETA calculations using Neuroguide at 30-day intervals. Since the use of qEEG has increased in clinical and research settings and a reliable measure is needed across disciplines, this topic proves important.
2. Methods This study was conducted with 19 normative (7 female), undergraduate students with a mean age of 20.74, SD = 1.24. One participant was left handed. All participants read, signed and agreed to informed consent approved by the University of Tennessee Institutional Review Board. Exclusion criteria included psychiatric or neurological syndromes, history of head trauma, epilepsy or neurovascular incidents, and use of alcohol or drugs within 2 weeks of the data collection procedures. All participants received extra course credit for participation in this study. After obtaining informed consent, participants were advised of all procedures and equipment. The ears and forehead were cleaned for recording with a mild abrasive gel (NuPrep) to remove any oil and dirt from the skin. The head was measured and marked prior to EEG recording using a measure of head circumference and the distance between the nasion and inion to determine the appropriate cap size for recording and placement of frontal electrodes. After fitting caps, each electrode site was injected with electrogel and prepared so that impedances between individual electrodes and each ear remain <10 k. Impedances were <2 k between any two electrodes. The EEG was recorded at 256 samples/s with linked-ear referenced using 9 mm tin cups. The participants were recorded in 4 min eyes-closed (ECB) and eyes-opened (EOB) baseline conditions using the Deymed Truscan EEG Acquisition System (Deymed Diagnostics) at exactly 30-day intervals; the time of day was constant across recordings. The EEG data were exported from storage and entered into Neuroguide for analysis. Neuroguide features an automatic selection method based on a contrast selection created by the user. For all participant records, at time 1 and time 2, a mean record length for automatic selection in Neuroguide of 6 s was obtained across individuals. These 6-s segments were utilized to minimize differences in selection bias between baseline conditions. Neuroguide also computes the reliability for the selected sections which was set to ≥0.90. Only these automatic segments were used for analyses – no extra epochs were added. Neurobatch analysis was conducted and data were extrapolated from the output and entered into SPSS 19 for reliability analysis. This study compared all electrode pairs (coherence and phase) and measures of absolute power of all electrodes for the reliability analysis. Neuroguide computes LORETA absolute power in 1 Hz increments, which were evaluated similarly. To test reliability, we used a random mixed model with an absolute agreement definition. The conceptual difference between absolute agreement and consistency definitions begins by noting their formal distinction, which is in the definition of the intraclass correlation coefficient (ICC) denominator. For consistency
Table 1 Paired t-test between length of recordings for times 1 and 2 in mean seconds of EEG. Although a noticeable difference exists between the amount of usable data between 30-day intervals, these differences did not reach significance with ECB t(18) = −1.69, p = 0.108 and EOB t(18) = −1.71, p = 0.105. Mean
N
SD
Pair 1
ECB time 1 ECB time 2
90.85 112.04
19 19
39.79 39.67
9.13 9.10
Pair 2
EOB time 1 EOB time 2
102.88 122.55
19 19
46.50 45.52
10.67 10.44
SE
measures, column variance is excluded from denominator variance, and for absolute agreement, it is not. Column variance is excluded from the denominators of consistency measures because it is deemed an irrelevant source of variance. In the case of absolute agreement when measurements disagree in absolute value, regardless of the reason, they are viewed as disagreements [15]. EEG frequency domains were contrasted for 1 Hz increments and for typical frequency bands (e.g., delta 1.0–4.0, theta 4.0–8.0, alpha 8.0–12.0 and beta 12.0–32.0). 3. Results The results for reliability contrasts are shown in Tables 1–6. Table 1 shows the contrast for the mean length of EEG records at time 1 and time 2 for ECB and EOB recordings. There is no significant difference between time 1 and 2. However, time 2 shows more applicable data overall. Table 2 shows results of reliability measures for all topographical measures. From left to right are the measure, alpha, ICC, F and degrees of freedom, and top to bottom are the results for absolute power, coherence for EOB/ECB and phase ECB/EOB. The results for absolute power and coherence show good to very good agreement for total power and frequency-wise contrasts. FFT phase shows low to moderate reproducibility for total power and frequency-wise contrasts. Table 3 shows the results for total absolute power at each individual electrode in the EOB condition. The results show good to very good agreement with the exception of C4, F8, T3 and T6. Table 4 shows the results for reliability contrasts for total absolute power at Table 2 Topographical reliability contrasts for absolute power and connectivity measures. In the table from left to right are the measure, Chronbach’s alpha, intraclass correlation coefficients, F for the contrast and the degrees of freedom. Measure (time 1 = time 2) ECB FFT absolute power EOB FFT absolute power ECB FFT coherence Delta Theta Alpha Beta EOB FFT coherence Delta Theta Alpha Beta ECB phase lag deg Delta Theta Alpha Beta EOB phase lag deg Delta Theta Alpha Beta **
Significance < 0.000.
Alpha
ICC
F
df **
0.91 0.78
0.9 0.77
10.51 4.53**
1.443 1.443
0.91 0.95 0.96 0.97
0.91 0.95 0.96 0.96
11.64 21.57** 25.88** 29.83**
3.248 3.248 3.248 3.248
0.85 0.88 0.9 0.92
0.85 0.87 0.89 0.91
6.67** 8.25** 9.71** 12.12**
3.248 3.248 3.248 3.248
0.064 0.29 0.74 0.58
0.064 0.29 0.74 0.58
1.06** 1.42** 3.84** 2.39**
3.248 3.248 3.248 3.248
0.12 0.4 0.4 0.39
0.12 0.4 0.4 0.39
1.38** 1.67** 1.66** 1.64**
3.248 3.248 3.248 3.248
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31 Table 3 Results for total absolute power at individual electrodes in the EOB condition (time 1 = time 2). In the table from left to right are the electrode, the intraclass correlation coefficient, Chronbach’s alpha and the F for the contrast. The average alpha for EOB is 0.77. These contrasts used 75 degrees of freedom (total power for delta, theta, alpha and beta or 4 frequency bands × 19 participants). Electrode
ICC
Alpha
F
C3 C4 CZ F3 F4 F7 F8 FP1 FP2 FZ O1 O2 P3 P4 PZ T3 T4 T5 T6
0.84 0.68 0.86 0.81 0.87 0.81 0.59 0.71 0.80 0.83 0.77 0.76 0.81 0.69 0.76 0.62 0.76 0.77 0.67
0.85 0.69 0.88 0.81 0.88 0.81 0.60 0.72 0.81 0.85 0.79 0.78 0.82 0.71 0.77 0.62 0.77 0.78 0.67
6.5** 3.17** 8.07** 5.24** 8.61** 5.19** 2.48** 3.53** 5.32** 6.65** 4.72** 4.48** 5.55** 3.42** 4.43** 2.6** 4.26** 4.57** 3.04**
**
Significance < 0.000.
individual electrodes in the ECB condition. The agreement for total power is good to very good with the exception of F8. Table 5 shows reliability coefficients at time 1 and time 2 for ECB (left) and EOB (right) for LORETA current source density in 1 Hz frequency bins. Considering the computational load and number of contrasts, the agreement for comparisons are good to very good for each 1 Hz frequency bin for all 2394 voxels. Table 6 shows the reliability results for 8 individual regions of interest within LORETA space. In the table from left to right are the region, Talairach coordinates, alpha, ICC and F for the contrast. Table 7 shows the reliability results for the same 8 regions used in Table 6 for ECB recordings. Individual ROIs show good to very good reliability with the voxel in left anterior cingulate (0.65) and left prefrontal cortex (0.67) showing the lowest reliability coefficients in ECB.
Table 4 Results for total absolute power at individual electrodes in the ECB condition (time 1 = time 2). In the table from left to right are the electrode, the intraclass correlation coefficient, Chronbach’s alpha and the F for the contrast. The average alpha for ECB is 0.85. These contrasts used 75 degrees of freedom (total power for delta, theta, alpha and beta or 4 frequency bands × 19 participants). Electrode
ICC
Alpha
F
C3 C4 CZ F3 F4 F7 F8 FP1 FP2 FZ O1 O2 P3 P4 PZ T3 T4 T5 T6
0.87 0.84 0.9 0.84 0.86 0.87 0.6 0.87 0.86 0.83 0.93 0.96 0.86 0.87 0.89 0.75 0.78 0.84 0.78
0.87 0.84 0.9 0.84 0.86 0.87 0.6 0.87 0.86 0.83 0.93 0.96 0.86 0.87 0.89 0.75 0.78 0.84 0.83
7.53** 6.22** 9.72** 6.23** 7.2** 7.54** 2.44** 7.6** 7.23** 5.77** 14.67** 25.4** 7.33** 7.76** 9.17** 4.02** 4.94** 6.19** 6.09**
LORETA absolute power. ** Significance < 0.000.
29
Table 5 In the table are the results for the reliability analyses for absolute power LORETA current source density for all 2394 7 mm3 voxels in 1 Hz increments for ECB and EOB. Shown in each table are the frequency domain, intraclass correlation coefficient, Chronbach’s alpha and the F for the ANOVA. The average reliability for ECB = 0.81 (left) and EOB = 0.82 (right) at 30 days. The specific contrasts between time 1 and time 2 were conducted with 45,486 − 1 (45,485) degrees of freedom (e.g., 2394 × 19). Frequency 1 Hz 2 Hz 3 Hz 4 Hz 5 Hz 6 Hz 7 hz 8 Hz 9 Hz 10 Hz 11 Hz 12 Hz 13 Hz 14 Hz 15 Hz 16 Hz 17 Hz 18 Hz 19 Hz 20 Hz 21 Hz 22 Hz 23 Hz 24 Hz 25 Hz 26 Hz 27 Hz 28 Hz 29 Hz 30 Hz **
ICC 0.76 0.78 0.77 0.78 0.77 0.76 0.81 0.84 0.9 0.9 0.84 0.81 0.76 0.78 0.79 0.81 0.81 0.85 0.85 0.86 0.86 0.87 0.86 0.85 0.85 0.84 0.84 0.82 0.81 0.82
Alpha 0.76 0.78 0.77 0.78 0.77 0.76 0.81 0.85 0.9 0.9 0.84 0.81 0.76 0.78 0.79 0.81 0.81 0.85 0.85 0.86 0.86 0.87 0.86 0.85 0.85 0.83 0.84 0.83 0.81 0.82
F **
4.23 4.5** 4.38** 4.46** 4.42** 4.11** 5.23** 6.5** 10.15** 9.55** 6.4** 5.16** 4.19** 4.54** 4.83** 5.23** 5.34** 6.63** 6.78** 6.95** 7.36** 7.45** 7.11** 6.75** 6.49** 6.05** 6.17** 5.7** 5.27** 5.66**
Frequency
ICC
Alpha
F
1 Hz 2 Hz 3 Hz 4 Hz 5 Hz 6 Hz 7 hz 8 Hz 9 Hz 10 Hz 11 Hz 12 Hz 13 Hz 14 Hz 15 Hz 16 Hz 17 Hz 18 Hz 19 Hz 20 Hz 21 Hz 22 Hz 23 Hz 24 Hz 25 Hz 26 Hz 27 Hz 28 Hz 29 Hz 30 Hz
0.84 0.83 0.82 0.81 0.8 0.78 0.76 0.78 0.81 0.82 0.87 0.87 0.82 0.8 0.77 0.77 0.75 0.78 0.77 0.75 0.77 0.78 0.79 0.8 0.8 0.8 0.82 0.81 0.81 0.8
0.85 0.84 0.83 0.82 0.81 0.79 0.79 0.8 0.83 0.83 0.88 0.88 0.83 0.81 0.78 0.76 0.76 0.77 0.76 0.76 0.77 0.78 0.79 0.8 0.8 0.8 0.81 0.81 0.81 0.8
6.67** 6.22** 5.97** 5.66** 5.34** 4.79** 4.82** 4.99** 5.83** 5.81** 8.01** 8.01** 5.72** 5.26** 4.48** 4.44** 4.15** 4.3** 4.27** 4.17** 4.41** 4.6** 4.84** 5.1** 5.05** 5.05** 5.47** 5.37** 5.38** 4.99**
Significance < 0.000.
4. Discussion In general, findings demonstrate the reliability of computerized analysis of EEG show good to very good reproducibility at 30 days. Similarly, LORETA calculations for 1 Hz increments across time show similar robust agreement, as do the individual regions of interest. The absolute power coefficients show good to very good reliability for ECB and EOB across the 30-day time interval for the eyes-closed baseline for all contrasts. These results apply to total power and frequency-wise contrasts. Hence, the data demonstrate good to very good reproducibility of qEEG measures and LORETA
Table 6 Reliability for 8 regions of interest for total absolute power for EOB 1 = EOB 2. In the table from left to right are the neuroanatomical regions of interest with Brodmann Area designation (BA) and abbreviation (e.g., left anterior cingulate = ACC), (voxel 7 mm3 ), x, y and z coordinates in Talairach space, Chronbach’s alpha, intraclass correlation coefficient and the F for the contrast. Voxel
Coordinates
Alpha
ICC
F
Left anterior cingulate gyrus (BA 32) ACC Left precuneus (BA 19) LPC Left prefrontal (BA 8) LPFC Right prefrontal (BA 8) RPFC Right postcentral gyrus (BA 40) RPCG Left supramarginal gyrus (BA 40) LSMG Right supramarginal gyrus (BA 40) RSMG Left cuneus (BA 7)
(−3, 31, 29)
0.98
0.98
43.42**
(−31, −81, 22) (−38, 31, 43) (39, 24, 43) (53, −18, 43)
0.94 0.95 0.97 0.95
0.94 0.95 0.97 0.95
17.05** 18.59** 31.57** 18.18**
(−59, −53, 29)
0.93
0.93
15.15**
(60, −53, 29)
0.92
0.92
13.12**
(−3, −67, 29)
0.91
0.91
10.84**
**
Significance < 0.000.
30
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31
Table 7 Reliability for 8 regions of interest for total absolute power for ECB 1 = ECB 2. In the table from left to right are the neuroanatomical regions of interest with Brodmann Area designation (BA) and abbreviation (e.g., left prefrontal cortex = LPFC) (voxel 7 mm3 ), x, y and z coordinates in Talairach space, Chronbach’s alpha, intraclass correlation coefficient and the F for the contrast. Voxel
Coordinates
Alpha
ICC
F
Left anterior cingulate gyrus (BA 32) ACC Left precuneus (BA 19) LPC Left prefrontal (BA 8) LPFC Right prefrontal (BA 8) RPFC Right postcentral gyrus (BA 40) RPCG Left supramarginal gyrus (BA 40) LSMG Right supramarginal gyrus (BA 40) RSMG Left cuneus (BA 7)
(−3, 31, 29)
0.65
0.65
2.89**
(−31, −81, 22) (−38, 31, 43) (39, 24, 43) (53, −18, 43)
0.83 0.67 0.89 0.84
0.83 0.65 0.88 0.84
5.89** 2.91** 8.85** 6.21**
(−59, −53, 29)
0.86
0.86
7.16**
(60, −53, 29)
0.85
0.85
6.49**
(−3, −67, 29)
0.86
0.86
7.05**
**
Significance < 0.000.
current source density over a 30-day interval for this study population of clinically normal individuals. The coherence results between all electrode pairs at 30 days shows very good reliability for ECB and good to very good results of EOB for each frequency domain. Since the degree of synchrony in EEG signals is commonly characterized by time-series measures, it is important for these measures to be reliable and consistent across studies since reference site may produce inaccurate results [13]. Coherence is a statistic of phase differences and yields a much finer measure of shared energy between mixtures of periodic signals than can be achieved using the Pearson product-moment correlation coefficient of amplitudes (or more simply functional integration or differentiation) [5,23,25]. Thus, coherence may be a potential target for the evaluation of traumatic brain injury, psychological syndromes or concussion [23,25,26]. Of all contrasts conducted in this work, FFT phase shows the lowest reliability coefficients. Reasons for this may be three fold. First, subtle artifacts may produce effects on the phase relationship between any two signals. Second, this may be a function of learning and experience. The prior experience with baseline procedures may have reduced the neural power necessary to reach the selfregulatory state of baseline. Several fMRI studies demonstrate that during conditioning phase synchrony develops between neuronal populations [12]. This is certainly a topic for future study. Finally, this may be the result of automatic selection procedures from which extra EEG segments could be selected and added by the examiner. The CSD for 1 Hz frequency bins across all 2394 voxels shows very good reliability with 0.81 for ECB and 0.82 for EOB. Similarly, eight regions (single voxels 7 mm3 ) show good to very good reliability except for the anterior cingulate (0.65) and left prefrontal (0.67) in ECB condition. Even though these are low compared to the remaining regional voxels, they are still in the good range. The current data show good to very good reliability for qEEG power and coherence measures as computed by Neuroguide software. LORETA current source density shows good to very good agreement for baselines across time for 1 Hz frequency bins as well as individual ROIs over time. LORETA, as such, is the only neuroimaging technique to produce these results for baseline measures to date. Other imaging techniques have attempted to evaluate reliability across time for specific tasks with variable results [3]. Additions may be added to the current data and will be the focus of subsequent works. Higher frequency bands (e.g., high-beta and gamma) will be examined in future works in both topographical and LORETA contrasts. The current population was from a convenience sample in a university population, which may not represent the general population. Similarly, reliability measures may be taken
throughout the day to determine potential differences in the EEG relative to time of day. Similarly, we did not include Neuroguide automatic eye-movement and electrocardiograph (EKG) artifact channels. It should also be noted that an extensive visual inspection of all automatic selection samples was not completed; since artifacting programs are not regarded as stand alone without additional inspection by qualified individuals, this may have influenced the results. Further, if longer samples were obtained (e.g., 10–15 s) with a higher reliability criteria (e.g., 0.94 or 0.95), then results may have differed respectively. These are topics of future studies. The stability of electrical brain activity is a topic of intense interest given its potential importance in developing diagnostic biomarkers and as a mechanism to evaluate treatment outcomes. Thus, many arguments against the use of qEEG and LORETA in clinical diagnostic procedures may not hold much weight. The challenge becomes how to find specific, characteristic patterns of neural activity that best fit any particular psychiatric syndrome, traumatic brain injury or other learning issues and their neurological correlates. At present, our lab is addressing some of these issues and much more research is certainly needed for future developments. References [1] S.R. Benbadis, W.C. LaFrance, G.D. Papandonatos Jr., K. Korabathina, K. Lin, H.C. Kraemer, Interrater reliability of EEG-video monitoring, Neurology 73 (2009) 843–846. [2] M.H. Bjork, T. Sand, G. Brathen, O.M. Linaker, G. Morken, B.M. Nilsen, A.E. Vaaler, Quantitative EEG findings in patients with acute, brief depression combined with other fluctuating psychiatric symptoms: a controlled study from an acute psychiatric department, BMC Psychiatry 8 (2008) 89. [3] G.G. Brown, D.H. Mathalon, H. Stern, J. Ford, B. Mueller, D.N. Greve, G. McCarthy, J. Voyvodic, G. Glover, M. Diaz, E. Yetter, I.B. Ozyurt, K.W. Jorgensen, C.G. Wible, J.A. Turner, W.K. Thompson, S.G. Potkin, Multisite reliability of cognitive BOLD data, Neuroimage 54 (2011) 2163–2175. [4] R. Cannon, D. Baldwin, J. Lubar, Self-perception and experiential schemata in the addicted brain, Applied Psychophysiology and Biofeedback 33 (2008) 223–238. [5] R.L. Cannon, M. Congedo, J.F. Lubar, T. Hutchens, Differentiating a network of executive attention: LORETA neurofeedback in anterior cingulate and dorsolateral prefrontal cortices, International Journal of Neuroscience 119 (2009) 404–441. [6] R. Cannon, C. Kerson, A. Hampshire, sLORETA and fMRI detection of medial prefrontal default network anomalies in adult ADHD, Journal of Neurotherapy 15 (2011) 358–373. [7] R. Cannon, J. Lubar, M. Congedo, K. Thornton, K. Towler, T. Hutchens, The effects of neurofeedback training in the cognitive division of the anterior cingulate gyrus, International Journal of Neuroscience 117 (2007) 337–357. [8] R.L. Cannon, M.K. Crane, P.D. Campbell, J.H. Dougherty, D.R. Baldwin, J.D. Effler, L.S. Phillips, F. Hare, M. Zachary, K.E. Cox, D.J. Di Loreto, A 9-year-old boy with multifocal encephalomalacia: EEG Loreta and lifespan database, magnetic resonance imaging and neuropsychological agreement, Journal of Neurotherapy 15 (2011) 3–17. [9] K.L. Coburn, E.C. Lauterbach, N.N. Boutros, K.J. Black, D.B. Arciniegas, C.E. Coffey, The value of quantitative electroencephalography in clinical psychiatry: a report by the Committee on Research of the American Neuropsychiatric Association, Journal of Neuropsychiatry and Clinical Neurosciences 18 (2006) 460–500. [10] M. Corsi-Cabrera, L. Galindo-Vilchis, Y. del-Rio-Portilla, C. Arce, J. Ramos-Loyo, Within-subject reliability and inter-session stability of EEG power and coherent activity in women evaluated monthly over nine months, Clinical Neurophysiology 118 (2007) 9–21. [11] P. Coutin-Churchman, Y. Anez, M. Uzcategui, L. Alvarez, F. Vergara, L. Mendez, R. Fleitas, Quantitative spectral analysis of EEG in psychiatry revisited: drawing signs out of numbers in a clinical setting, Clinical Neurophysiology 114 (2003) 2294–2306. [12] S. Grossberg, M. Versace, Spikes, synchrony, and attentive learning by laminar thalamocortical circuits, Brain Research 1218 (2008) 278–312. [13] S. Hu, M. Stead, Q. Dai, G.A. Worrell, On the recording reference contribution to EEG correlation, phase synchrony, and coherence, IEEE Transactions on Systems Man and Cybernetics Part B: Cybernetics 40 (2010) 1294–1304. [14] J.R. Hughes, E.R. John, Conventional and quantitative electroencephalography in psychiatry, Journal of Neuropsychiatry and Clinical Neurosciences 11 (1999) 190–208. [15] K.O. McGraw, S.P. Wong, Forming inferences about some intraclass correlation coefficients, Psychological Methods 1 (1996) 30–46. [16] T.R. Oakes, D.A. Pizzagalli, A.M. Hendrick, K.A. Horras, C.L. Larson, H.C. Abercrombie, S.M. Schaefer, J.V. Koger, R.J. Davidson, Functional coupling of simultaneous electrical and metabolic activity in the human brain, Human Brain Mapping 21 (2004) 257–270.
R.L. Cannon et al. / Neuroscience Letters 518 (2012) 27–31 [17] R.D. Pascual-Marqui, Standardized low-resolution brain electromagnetic tomography (sLORETA): technical details, Methods and Findings in Experimental and Clinical Pharmacology 24 (Suppl. D) (2002) 5–12. [18] R.D. Pascual-Marqui, C.M. Michel, D. Lehmann, Low resolution electromagnetic tomography: a new method for localizing electrical activity in the brain, International Journal of Psychophysiology 18 (1994) 49–65. [19] D.A. Pizzagalli, T.R. Oakes, R.J. Davidson, Coupling of theta activity and glucose metabolism in the human rostral anterior cingulate cortex: an EEG/PET study of normal and depressed subjects, Psychophysiology 40 (2003) 939–949. [20] L.S. Prichep, K.R. Alper, L. Sverdlov, S.C. Kowalik, E.R. John, H. Merkin, M.L. Tom, B. Howard, M.S. Rosenthal, Outcome related electrophysiological subtypes of cocaine dependence, Clinical Electroencephalography 33 (2002) 8–20. [21] B. Saletu, G. Saletu-Zyhlarz, P. Anderer, N. Brandstatter, R. Frey, G. Gruber, G. Klosch, M. Mandl, J. Grunberger, L. Linzmayer, Nonorganic insomnia in generalized anxiety disorder. 2. Comparative studies on sleep, awakening, daytime vigilance and anxiety under lorazepam plus diphenhydramine (Somnium) versus lorazepam alone, utilizing clinical, polysomnographic and EEG mapping methods, Neuropsychobiology 36 (1997) 130–152. [22] S.C. Sanei, J.A. EEG Signal Processing, John Wiley & Sons, Ltd., West Sussex, 2007.
31
[23] R.W. Thatcher, C. Biver, J.F. Gomez, D. North, R. Curtin, R.A. Walker, A. Salazar, Estimation of the EEG power spectrum using MRI T(2) relaxation time in traumatic brain injury, Clinical Neurophysiology 112 (2001) 1729–1745. [24] R.W. Thatcher, C.J. Biver, D.M. North, Quantitative EEG and the Frye and Daubert standards of admissibility, Clinical Electroencephalography 34 (2003) 39–53. [25] R.W. Thatcher, D. North, C. Biver, Evaluation and validity of a LORETA normative EEG database, Clinical EEG and Neuroscience 36 (2005) 116–122. [26] R.W. Thatcher, D.M. North, R.T. Curtin, R.A. Walker, C.J. Biver, J.F. Gomez, A.M. Salazar, An EEG severity index of traumatic brain injury, Journal of Neuropsychiatry and Clinical Neurosciences 13 (2001) 77–87. [27] S.U. Zuidema, A.L. Buursema, M.G. Gerritsen, K.C. Oosterwal, M.M. Smits, R.T. Koopmans, J.F. de Jonghe, Assessing neuropsychiatric symptoms in nursing home patients with dementia: reliability and Reliable Change Index of the Neuropsychiatric Inventory and the Cohen-Mansfield Agitation Inventory, International Journal of Geriatric Psychiatry 26 (2011) 127–134. [28] D. Zumsteg, D.M. Andrade, R.A. Wennberg, Source localization of small sharp spikes: low resolution electromagnetic tomography (LORETA) reveals two distinct cortical sources, Clinical Neurophysiology 117 (2006) 1380–1387.