Clinical Neurophysiology 119 (2008) 2185–2187

Contents lists available at ScienceDirect

Clinical Neurophysiology

journal homepage: www.elsevier .com/locate /c l inph

32-Channel banana-avg montage is better than 16-channel double bananamontage to detect epileptiform discharges in routine EEGs

Juan Ochoa a,b,*, Walter Gonzalez a, Ramon Bautista a,b, John DeCerce a,b

a Neuroscience Institute, Department of Neurology, University of Florida/Shands Jacksonville, 580 West 8th Street, Tower I, 9th Floor, Jacksonville, FL 32209, USAb Comprehensive Epilepsy Program, University of Florida, Jacksonville, FL 32209, USA

a r t i c l e i n f o

Article history:Accepted 20 June 2008Available online 26 August 2008

Keywords:EEGEpilepsyElectroencephalogramSpike detectionMontage.

1388-2457/$34.00 Published by Elsevier Ireland Ltd.doi:10.1016/j.clinph.2008.06.010

* Corresponding author. Address: Neuroscience Instogy, University of Florida/Shands Jacksonville, 580 WFloor, Jacksonville, FL 32209, USA. Tel.: +1 904 244 98

E-mail address: Juan.ochoa@jax.ufl.edu (J. Ochoa).

a b s t r a c t

Objective: We designed a study, comparing the yield of standard 16-channel longitudinal bipolar mon-tage (double banana) versus a combined 32-channel longitudinal bipolar plus average referential mon-tage (banana-plus), to detect epileptic abnormalities.Methods: We selected 25 consecutive routine EEG samples with a diagnosis of spike or sharp waves in thetemporal regions and 25 consecutive focal slowing and 50 normal EEGs. A total of 100 samples wereprinted in both montages and randomized for reading. Thirty independent EEG readers blinded fromthe EEG diagnosis were invited to participate.Results: Twenty-two readers successfully completed the test for a total of 4400 answers collected foranalysis. The average sensitivity to detect epileptiform discharges for 16 and 32-channel montageswas 36.5% and 61%, respectively (P < 0.001). The specificity was not significantly different for 16 and32-channel montages, 95.4% and 92.4%, respectively (P = 0.27). However more readers detected false pos-itives on 32-channel montage compared to 16-channel montage.Conclusions: 32-Channel banana-plus montage has a better yield to detect epileptic abnormalities thanthe 16-channel double banana montage.Significance: Residents and EEG fellows could improve EEG-reading accuracy if taught on a combined 32-channel montage.

Published by Elsevier Ireland Ltd. on behalf of International Federation of Clinical Neurophysiology.

1. Introduction

Routine electroencephalogram (EEG) is the most useful tool toconfirm the diagnosis of epilepsy. Approximately 90% of patientswith epilepsy have epileptiform discharges between attacks (Bin-nie and Stefan, 1999). These abnormalities include spikes, sharpwaves, spike-wave complex, focal slowing and electrographicseizures. A waking EEG of at least 30-min duration with hyper-ventilation and photic stimulation has a sensitivity to detect epi-leptic discharges around 50%, in adults with epilepsy. Serialrecords may enhance the yield up to 85% (Salinsky et al.,1987). Although these findings suggest the presence of underly-ing epileptic disorder, the prevalence of epileptiform abnormali-ties between people without epilepsy is about 2–3% (Binnie andStefan, 1999). Sensitivity during the first EEG for detecting epi-leptiform discharges is reported in a range from 29% to 56% inpatients with epilepsy (Gilbert et al., 2006). Techniques currently

on behalf of International Federatio

itute, Department of Neurol-est 8th Street, Tower I, 9th96; fax: +1 904 244 9543.

used to increase the yield of epileptiform abnormalities in theEEG laboratory are hyperventilation, photostimulation, sleepdeprivation and repetitive EEG (Veldhuizen et al., 1983; Guara-nha et al., 2005; Mendez and Brenner, 2006). Nowadays routineEEGs are recorded using digital devices, allowing readers tomanipulate the settings of the machine including the display ofchannels, montage, sensitivity and filters, whereas the traditionalanalog machines would not allow any changes and have a lim-ited number of channels (Worrell et al., 2002). Current recom-mendations by the American Clinical Neurophysiology Societyfor a good technical acquisition of digital EEGs include recordingnot less than 16-channels of simultaneous recording using bothbipolar and referential montages during acquisition (AmericanClinical Neurophysiology Society, 2006). These connotations werecreated to improve the detection of different abnormalitiesincluding epileptiform discharges and to obtain standard param-eters for easier interpretation independent of the EEG’s reader.

Double banana or longitudinal bipolar montage is a commonmontage used routinely and it is alternated with any other referen-tial montage (Worrell et al., 2002). Bipolar montages depend onpotential difference of adjacent recording electrodes; hence a focalabnormality with a large field may have little potential difference

n of Clinical Neurophysiology.

2186 J. Ochoa et al. / Clinical Neurophysiology 119 (2008) 2185–2187

between the two recording electrodes resulting in a low wave dis-play or even cancellation (Burgess and Collura, 2002). Referentialmontages (common referential montages include ears, vertex,and common average) on the contrary use a reference electrodethat is ideally virtually silent, enhancing the display of potentialsfrom the recording areas; however, electrical contamination ofthe reference is common and it may yield to a wrong interpreta-tion. Consequently, a combination of a bipolar montage and a ref-erential montage in the same display may provide a betterdiscrimination of abnormal focal EEG abnormalities.

2. Methods

We designed a study, comparing the yield of standard 16-chan-nel longitudinal bipolar montage (double banana) versus a com-bined 32-channel longitudinal bipolar plus average referentialmontage (banana-avg), to detect epileptiform abnormalities. Insti-tutional IRB approved this study and written informed consent wasobtained from all readers before the study entry.

2.1. Materials

Twenty-five consecutive routine EEG samples, with a diagno-sis of spike or sharp waves in the temporal regions (right andleft) and 25 consecutive focal slowing and 50 normal EEGs (con-trols) were selected from our EEG archives at The University ofFlorida Neuroscience Institute EEG database. EEG diagnosis wasconfirmed by a board-certified electroencephalographer. A singlerepresentative sample was printed both in a 16-channel doublebanana montage and a 32-channel combined (banana-avg) mon-tage. The banana-avg montage contains 16-channel longitudinalbipolar combined with 16-channel common average reference.The reference montage was defined as the average of F3, F4,C3, C4, F7, F8, T3, T4, P3, P4, T5 and T6. Electrodes with frequentcontamination including Fp1–Fp2, A1–A2, O1–O2 and FZ–CZ–PZwere excluded from the average. A total of 200 EEG samples(100 samples in double banana montage and 100 samples in ba-nana-plus montage) were included for the study. Selected pageswere randomly assigned and printed at standard letter papersize, Each EEG sample contains only one abnormality and wasclassified either epileptic or not epileptic.

2.2. EEG specifications

All EEGs were recorded on Nihon Kohden machines. The samplerate was 240/s with 12-bit resolution. All studies used the 10–20International System of electrode placement. All records had a sen-sitivity of 7 mV/mm, HFF 70 Hz, TC 0.1 s. AC filter off, and free of

Fig. 1. Sensitivity and specificity for 2

major artifacts. Readers performed a direct visual analysis of theEEG on printed sample pages.

2.3. Randomization

A total of 200 EEG samples were included. Each EEG sample wasrandomized in order to prevent pairing equal samples of bothmontages and identified with a randomized number using excelrandomization software.

2.4. Population

Thirty independent EEG readers were directly requested to par-ticipate in this study. The readers were clinical epileptologists se-lected from the American Epilepsy Society. All readers wereblinded from the EEG diagnosis and had no previous exposure tothe banana-plus montage.

3. Results

Twenty-three out of 30 invited readers accepted to participate.A total of 4400 answers were collected for the analysis correspond-ing to 22 readers. One reader was excluded because his answerswere marked wrongly in the answer-sheet during the transcriptionprocess. We used a two-tailed t-test for the statistical analysiscomparing 16 versus 32-channel montage for each individual read-er. The average sensitivity to detect epileptiform discharges for 16and 32-channel montages was 36.5% and 61%, respectively(P < 0.001). The specificity was not significantly different for 16and 32-channel montages (95.4% and 92.4%, respectivelyP = 0.27) see Fig. 1. Furthermore, there were no differences on po-sitive predictive values (PPV) (16-channel 83% and 32-channel81%) but negative predictive values (NPV) were better for the bana-na-plus montage (16-channel = 81% and 32-channel = 88%P < 0.01). Further analysis using categorical data revealed that 11readers detected more false epileptiform abnormalities when read-ing on a 32-channel montage compared to only two readers whodetected more false positives on a 16-channel montage. Althoughnine readers showed no difference in the detection of false posi-tives, McNemar test suggested that false detection of epileptiformabnormalities was more frequent when reading on a 32-channelmontage.

Inter-reader agreement was calculated using Kappa test aver-age; the banana-avg group showed a moderate agreement (0.57)and it was better than double banana montage group who had alow agreement (0.367). The results showed that the answers onthe 32-channel montage were more homogeneous than whenusing only 16-channel double banana.

2 readers (16 and 32-channels).

J. Ochoa et al. / Clinical Neurophysiology 119 (2008) 2185–2187 2187

4. Discussion

This study demonstrated that displaying 32-channels andcombining a bipolar with a referential montage increase theyield for the detection of epileptic discharges on routine EEGswithout significantly affecting the specificity of the test. How-ever there is a potential risk of detecting more false positiveson 32-channel montage as found on the McNemar analysis.There was a greater inter-reader agreement when using the32-channel montage. The banana-plus montage thus appears tobe a good alternative for routine EEG reading. The main limita-tion of this study is that a single-page EEG display does not rep-resent the standard for EEG reading and it is unclear if thedifference found in this study would be maintained after readingthe entire record in a traditional way. In addition, given the var-iability of readers’ thresholds to call an epileptiform activity, thegold standard to assess sensitivity of the test cannot be clearlydefined. To overcome this limitation, the data were also analyzedusing each reader as their own control to assess the differencebetween the samples regardless of the diagnosis, still maintain-ing a significant difference between 16 and 32-channelmontages.

Bipolar montages display better EEG data without the contam-ination of the average montage whereas average montages displaybetter the EEG data that are attenuated by large field distributionsand technical limitations such as short interelectrode distance orsalt bridge. There is not added cost or recording time to obtainEEGs using a 32-channel display. Traditional EEGers who wereused to 16-channel display recordings may have some difficultyadapting to read a busier display and may slow down their readingspeed. We believe that it would not take long to get used to a com-bined montage for an experienced EEGer. EEG fellows and resi-

dents should be taught to read on a 32-channel display with acombined bipolar and referential montage.

Conflict of interest

None of the authors has any conflict of interest to disclose.

Acknowledgments

We thanks Dr. Peter Wludyka for his support in the statisticalanalysis. We confirm that we have read the Journal’s position on is-sues involved in ethical publication and affirm that this report isconsistent with those guidelines.

References

American Clinical Neurophysiology Society. Guideline 6: a proposal for standardmontages to be used in clinical EEG. J Clin Neurophysiol 2006;23:111–7.

Binnie CD, Stefan H. Modern electroencephalography: its role in epilepsymanagement. Clin Neurophysiol 1999;110:1671–97.

Burgess RC, Collura TF. Polarity, localization, and field determination inelectroencephalography. In: Wyllie E, editor. The treatment ofepilepsy. Philadelphia: Lippincot Williams and Wilkins; 2002. p. 230–50.

Gilbert DL, Sethuraman G, Kotagal U, Buncher CR. Meta-analysis of EEG testperformance shows wide variation among studies. Neurology 2006;60:564–70.

Guaranha MS, Garzon E, Buchpiguel CA, Tazima S, Yacubian EM, Sakamoto AC.Hyperventilation revisited: physiological effects and efficacy on focal seizureactivation in the era of video-EEG monitoring. Epilepsia 2005;46:69–75.

Mendez OE, Brenner RP. Increasing the yield of EEG. J Clin Neurophysiol2006;23:282–93.

Salinsky M, Kanter R, Dasheiff RM. Effectiveness of multiple EEGs in supporting thediagnosis of epilepsy: an operational curve. Epilepsia 1987;28:331–4.

Veldhuizen R, Binnie CD, Beintema DJ. The effect of sleep deprivation on the EEG inepilepsy. Electroencephalogr Clin Neurophysiol 1983;55:505–12.

Worrell GA, Lagerlund TD, Buchhalter JR. Role and limitations of routine andambulatory scalp electroencephalography in diagnosis and managing seizures.Mayo Clin Proc 2002;77:991–8.