| Academic Open Internet Journal ISSN 1311-4360 |
Volume 17, 2006 |
IDENTIFICATION OF INTERMEDIATE LATENCIES IN TRANSIENT VISUAL EVOKED POTENTIALS
Dr.R.Sivakumar M.E., Ph.D Assistant Professor*
Dr.G.Ravindran M.E., Ph.D Professor**
*Department of Electronics and Communication Engineering,
Sri Krishna College of Engineering and Technology, Kuniamuthur, Coimbatore, Tamilnadu, India 641 008
** Centre For Medical Electronics, College of Engineering, Guindy Anna University, Chennai- 600 025, India, Asia.
Mailing address: Department of Electronics and Communication Engineering, Sri Krishna College of Engineering and Technology, Kuniamuthur, Coimbatore, Tamilnadu, India 641 008
E-mail: sivarkumar@mailcity.com, , rajagopalsivakumar@hotmail.com
Phone: +91-422- 2678001 ext. 230, Fax: +91-422- 2678012
Corresponding author: R.Sivakumar
Topic area: Biomedical signal processing
ABSTRACT
Transient Visual Evoked Potential (TVEP) is an important diagnostic test for specific ophthalmological and neurological disorders. The precision of clinical interpretation depends on the amount of information available. A more detailed study on spectral components of intermediate latency waveforms has been carried out. The method proposed in this paper is based on the magnitude difference between the dominant spectral components and adjust spectral component value that is very close to the dominant spectral component. The correlation between the magnitudes difference and the corresponding intermediate latency has been identified. The advantage of this method is that one can directly identify the latency more precisely than the time domain averaging method.
Key Words: Transient visual evoked potential, Spectral components, latency, averaging.
1 INTRODUCTION
The use of sensory evoked brain potentials in the study of attention and cognitive processing has a long history [1-4]. The Transient Visual Evoked Potential (TVEP) is an important diagnostic test for specific ophthalmological and neurological disorders [5-11]. VEP recordings are obtained in a simple and non-invasive way. The precision of clinical interpretation depends on the amount of information available. This requires long periods of stimulation. TVEP investigation focused on the dominant peaks N75 P100 N135, a negative deflection followed by a positive and then a negative deflection. The peak P100 occurring about 100 msec following the stimulation in all the normal patients. The amplitude and latencies of these peaks are measured directly from the signal. Quantification of these latency changes can contribute to the detection of possible abnormalities [12,13]. This requires the precise definition of the starting and the end points. Latency measure depends on the point at which the latency is calculated and usually the peak presents irregularities, so that interpolation is then required. The EP signal is always accompanied by the ongoing EEG signal, which is considered as noise in EP analysis. The SNR may be as low as –10dB. Overcoming the effects of noise becomes a major issue in EP analysis.
Many researchers have described a variety of approaches to extract the evoked potential from the background EEG [14-17]. Traditionally, the clinical use of VEPs based on visual reading. Given the fact that the useful data are completely buried in the ongoing EEG, averaging techniques are usually applied to estimate the VEP. Conventional methods of detection of visual anomalies, based on TVEPs require long periods of testing and averaging. Hence the problem of patient fatigue affects the accuracy of the results. These factors imply that the analysis in the time domain, based on amplitude and latency, is not reliable.
The failure of time domain analysis has compelled researchers to investigate the frequency domain characteristics of the VEP response. According to a working hypothesis published earlier [18], EPs are considered as stimulus – induced EEG rhythmicities. Accordingly, it is advantageous to analyze EPs in the frequency domain. The development of the FFT algorithm has facilitated the estimation of spectral functions [19]. Investigation of the frequency domain characteristics of VEP’s is an attractive analytic approach because it allows detection of suitable waveform abnormalities that may escape detection with normal latency measurements [20-22].
Several researchers have proposed methods using both the TVEP and SSVEP. Most of these methods utilize the latency and amplitude of P100 values of TVEPs to identify the abnormality and SSVEPs were Fourier analyzed, and phase and amplitude of the second harmonic response were measured. In most of the methods Fourier analysis was applied to only SSVEP [23,24].
Previous studies have made extensive use of "transient" evoked potentials, which are computed by averaging a large number of repetitive responses to separate the desired signal from concurrent "noise." Only few studies adapted the spectral analysis to TVEP. Apaydin et al (1993) studied the effects of oxygenated free radicals on VEP spectral components in experimental diabetes using TVEP. Kulkarni and Udpikar (1995)[25] recorded the thirty normal subjects at a flash rate of 1.8 Hz. Their result showed that the dominant frequencies in the range of 4 to 16 Hz. Finally they suggested that the power spectrum analysis of VEP could be used as a non-invasive, objective technique to assess the stage of in any ocular diseases.
Agar et al (2000)[26], Yargicoglu et al (1996)[27] and Yargicoglu et al (1997)[28] studied the effect of Chronic Cadmium exposure on VEP and EEG spectral components on Swiss albino rats. Amplitude maximal were obtained in the 2-4, 4-7, 8-13, 14-20, 20.5-36Hz frequency bands. None of the above methods correlated the latency with spectral components.
Our previous studies have confirmed that presence of latencies of 100msec, 120msec, 140msec, and 160msec can be identified precisely using the spectral components [29,30]. However, intermediate latencies between 100ms and 120ms or between 120ms and 130ms,etc. such as 112,123,145 etc the results shown that the intermediate latency cannot be predicted precisely. To overcome this problem, a method has been proposed in this paper for identifying intermediate latency values.
A more detailed study on spectral components of intermediate latency waveforms have been carried out in this paper. The method proposed in this paper is based on the magnitude difference between the dominant spectral components and adjust spectral component value that is very close to the dominant spectral component. The correlation between the magnitudes difference and the corresponding intermediate latency has been identified in this paper.
The advantage of this method is that one can directly identify the latency more precisely than the time domain averaging method.
2 MATERIAL AND METHOD
2.1.1 Subjects
Experiments were carried with subjects in the Neurology Department of the leading Medical Institute from 2002 to 2004. 550 cases with complete data were analyzed. Of these patients, 100 normal and 400 abnormal subjects (35 females and 65 males in the age group of 39 – 52 years-mean age 48) were taken for further analysis. In 400 abnormal subjects, 50 subjects had Multiple sclerosis (MS), 30 subjects had diminished vision, 20 subjects had Motor neural disorder and 300 subjects had diabetic retinopathy.
2.1.2 Patient preparation
The local institutional human experimentation committee approval has been obtained before the procedure. The written consent has also been obtained from each subject after complete explanation of the nature of study and possible consequence of the study. Subjects were requested to take routine medications in the morning of the procedure, including prescription ophthalmics, washing the hair the previous night (to facilitate electrode placement) and they were requested to eat shortly before the exam (to avoid relative hypoglycemia) and corrective lenses have also been brought to the testing room.
2.1.3 Equipment
Nicolet Viking IV – Pattern-shift stimulator television screen, signal amplifier with filters, computer system for averaging were used for the analysis (Figure 1).
2.1.4 TVEP Recording
TVEP was performed in a specially equipped electro diagnostic procedure room (darkened, sound attenuated room). Initially, the patient was made to sit comfortably approximately 1 meter away from the pattern-shift screen. Subjects were placed in front of a black and white checkerboard pattern displayed on a video monitor. The checks alternate black/white to white/black at a rate of approximately twice per second. Every time the pattern alternates, the patient's visual system generates an electrical response that was detected and was recorded by surface electrodes, which were placed on the scalp overlaying the occipital and parietal regions with reference electrodes in the ear (Figure 2). The patient was asked to focus his gaze onto the center of the screen. Each eye was tested separately (monocular testing).
2.1.5 Stimulation Pattern
The visual stimuli were checkerboard patterns (contrast 70%, mean luminance 110 cd/m2) generated on a TV monitor and reversed in contrast at the rate of two reversals per second (Figure 3). At the viewing distance of 114 cm , the check edges subtended 15 minutes of visual angle and the screen of the monitor subtended 12.5°. The refraction of all subjects was corrected for the viewing distance. The stimulation was monocular, with occlusion of the contra lateral eye.
2.1.6 Electrodes and Electrode Placement
Cup-shaped
Ag/AgCl electrodes were fixed with collodion in the following positions:
active electrode at Oz, reference electrode at Fpz, ground on the
left ear (Figure 4). The interelectrode resistance was kept below 3
k
.
The bioelectric signal was amplified (gain 20,000), filtered (band-pass,
1–100 Hz), and 75 events free from artifacts were averaged for
every trial (Odom et al 2004). The analysis time for each trial was 250
msec.
2.1.7 Eye Blink Removal
A common artifact that corrupts the TVEP data is eye blinks. This problem has been solved by an amplitude threshold method. The TVEP signals with magnitude above 50microvolts are assumed to be contaminated with eye blinks and are discarded from the experimental study and additional trials were conducted as replacements.
2.1.8 Data set description:
The experimental data was collected in terms of blocks of trials. First trial was the time period of 250msec before the onset of stimulation. Remaining trials are 250msec each after the onset of stimulation. One block of trial was the continuous collection of 20 trials displayed one after other. In a typical experiment, 3-4 blocks of trials were recorded. In the block of trials, the eye blink trials were eliminated.
2.1.9 P100 latency measurements
For each subject 75 trials were carried and corresponding waveforms were stored in the system hardware. From the 75 trials, 70 artifact free trials waveforms were selected and using the averaging method the all-70 trials were averaged to get the TVEP waveform. By manually moving the cursor over the averaged waveform the characteristic points such as N75, P100 and N145 were identified and corresponding latency values were identified. Only P100 values were taken for further analysis. The subjects TVEP data were divided into three groups based on their latency values
1. 100 Normal TVEP with the latency value of 100msec
2. 150 abnormal TVEP with latencies of 120, 140, 160, 180 and 200msec
3. 250 abnormal TVEP with intermediate latencies such 110, 115, 123, 133, 156, etc latency waveforms.
2.1.10 Spectral components identification and Peak Search
The spectral components of the above 3 groups were analyzed separately. The spectral components for each subject were identified using the Welch’s averaged periodogram method. The dominant spectral components and corresponding magnitude values were extracted. The spectral components (FS1, FS2) adjacent to the dominant spectral component on either side separated by 1Hz magnitudes (AS1, AS2) were also extracted. The adjacent spectral component having larger magnitude was selected for further analysis and the adjacent spectral component with lesser magnitude eliminated from further analysis. For example, figure 5 shows the spectrum plot for abnormal subject with dominant spectral component at 3Hz, and the adjacent spectral components at 2Hz(FS1) and 4Hz(FS2). It is shown that the 2Hz having larger magnitude than the 4Hz. Therefore 2Hz magnitude used for further analysis.
2.1.11 Magnitude difference and Intermediate latency
The magnitude differences between the dominant peak and the adjacent peak close to the dominant peak have been computed. Based on the first dominant peak frequency and the adjacent peak frequency the mid latency value were fixed, such as for the dominant peak at 3Hz and the adjacent peak at 2Hz the Mid latency fixed at 110msec and for the dominant peak at 3Hz and adjacent peak at 4Hz the mid latency fixed at 130 msec. Table 1 shows all possible combinations of first dominant, adjacent peak values and corresponding mid latency values.
The actual latency has been computed using the dominant peak spectral component and the magnitude difference between the dominant and adjacent peak.
Magnitude difference = First dominant Peak magnitude Adjacent peak magnitude
Actual Latency = Mid Latency ± Additional Latency
Based on the frequency value of adjacent peak with the dominant peak frequency the magnitude difference is called positive (P) or negative (N) magnitude difference. If the adjacent peak is lesser frequency than the dominant peak, then that magnitude called as positive and if the adjacent peak frequency is higher than the dominant frequency then the magnitudes is called as negative magnitude difference. Based on the magnitude difference, an additional latency has been identified, which is to be added to the base latency to obtain the actual intermediate latency. For positive magnitude difference, additional latency is added to the base latency and for negative magnitude difference additional latency is subtracted from the base latency to obtain the intermediate latency. Table 2 shows the positive magnitude difference values and corresponding latency to be added to the base latency. Table 3 shows the negative magnitude difference value and corresponding latency to be subtracted from the base latency.
The complete algorithm is shown in the Figure 6. These intermediate latencies have been compared with the latency identified by the averaging method.
3 RESULTS AND DISCUSSION
It has been observed that for the latency value of 100, 120, 140, 160, 180, 200, 220msec only the dominant spectral component peak and negligible adjacent peak amplitude are obtained. For the intermediate latencies, it has been observed that the adjacent peak has the considerable magnitude values. Figures 7 and 8 show the spectrum of 120msec and 113msecc TVEP waveforms. From the figures it has observed that the 120msec waveform spectrum has negligible adjacent peak magnitude and 128msec waveform spectrum has considerable magnitude value.
Table 4 shows the intermediate latency calculation for the dominant spectral components at 3Hz and secondary peak at 2Hz with positive magnitude difference. Table 5 shows the intermediate latency calculation for the dominant spectral component at 2Hz and the secondary peak at 3Hz with negative magnitude difference. It has been shown that as the magnitude difference increases, adding or subtracting latency difference from the mid latency obtains the intermediate latency.
The same procedures were repeated for the remaining ranges also i.e. 120-140, 140-160, 160-180, 180-200, and 200-220msec. It has been found that the intermediate latency calculation using the spectral components and magnitude difference methods exactly coincides with the results of the averaging method (P = 0.001). Thus using the proposed method, the spectral peak and corresponding latency have been obtained to the required precision. The results have been compared with 250 practical cases and were found to be consistent. Hence the application of the above method would definitely increase the accuracy of diagnosis.
4 REFERENCES
1. Spong P., Haider M. and Lindsley D.B. (1965) ‘Selective attentiveness and cortical evoked responses to visual and auditory stimuli’, Science, Vol. 148, pp. 395-397.
2. Naatanen R. (1975) ‘Selective attention and evoked potentials in humans – a critical review’, Biological Psychology, Vol. 2, pp. 237-307.
3. Regan D. (1989) ‘Human Brain Electro physiology: Evoked potentials and Evoked Magnetic Fields in Science and Medicine’, Amsterdam, Elsevier, pp. 41-42.
4. Collura T.A. (2001) ‘Human steady state Visual Evoked potential and Auditory potential Components During a Selective Discrimination Task’, Journal of Neuropathy, Vol. 1, pp. 1-13.
5. Plant G.T. (1983) ‘Transient visually evoked potentials to sinusoidal gratings in optic neuritis’, J Neurol Neurosurg Psychiatry., Vol. 46, No.12, pp. 1125-1133.
6. Chiappa K.H. (1990) ‘Pattern shift visual evoked potential methodology’, Evoked potentials in Clinical Medicine, K.H.Chiappa, Ed. Newyork: Raven Press, pp. 37-110.
7. Misra J.K. and Kalith (1999) ‘Clinical Neurophysiology’, I.Churchill Livingstone Pvt Ltd, New Delhi.
8. AbdelMageed M.A.S. and Assem H.M. (2002) ‘Electroretinography and visual evoked potential in children with IDDM’, Proceedings of ISLRR Vision 2002 Conference, pp.150.
9. Lauritzen L., Jorgensen M.H. and Michaelsen K.F. (2004) ‘Test-retest reliability of swept visual evoked potential measurements of infants visual acuity and contrast sensitivity’, Pediatric Research, Vol. 55, pp. 701-708.
10. Momose K., Kiyosawa M., Nemoto N., Mochizuki M. and Yu J.J. (2004) ‘ PRBS-determined temporal frequency characteristics of VEP in glaucoma’, Documenta Ophthalmologica, Vol. 108, pp. 41-46.
11. Suttle C.M. and Turner A.M. (2004) ‘Transient pattern Visual Evoked Potentials in children with Down’s syndrome’, Ophthalmic and Physiological Optics, Vol. 24, No. 2, pp. 91.
12. Nogawa T., Katayama K., Okuda H. and Uchida M. (1991) ‘Changes in the latency of the maximum positive peak of visual evoked potential during anesthesia’, Nippon Geka Hokan, Vol. 60, No.3, pp. 143-153.
13. Xu S., Wagner H., Joo F., Cohn R. and Klatzo I. (1992) ‘Reduced latency of the visual evoked potential cortical response following cryogenic injury to cerebral cortex – A neuroexcitatory phenomenon’, Neurological Research, Vol. 14, pp. 233-235.
14. Fotopoulos S., Bezerianos A.and Laskaris N. (1995) ‘Latency measurement improvement of P100 complex in visual evoked potentials by FMH filters’, IEEE Transactions on Biomedical Engineering, Vol. 42, No. 4, pp. 424 –428.
15. Kong X. and Thakor N.V. (1996) ‘Adaptive estimation of latency changes in evoked potentials’, IEEE Transaction on Biomedical Engineering, Vol.43, pp. 189-197.
16. Gevins A.S., Morgan N.H., Bressler S.L., Doyle J.C. and Cutillo B.A. (1986) ‘ Improved event-related potential estimation using statistical pattern classification’, Electroencencephalogr., Clin. Neurophysiol., Vol. 64, pp. 177-186.
17. Davila C.E., Srebro R. and Ghaleb I.A. (1998) ‘Optimal Detection of Visual Evoked Potentials’, IEEE Transaction on Biomedical Engineering, Vol. 45, No. 6, pp. 800 –803.
18. Basar E. (1980) ‘ EEG-Brain Dynamics’, Elsevier, Amsterdam.
19. Kelly S., Burke D., Chazal P.D. and Reilly R. (2002) ‘Parametric Models and Spectral analysis for Classification in Brain-Computer Interface’, Proc. Of 14th International Conference on DSP, pp. 1-4.
20. Skuse N.F. and Burke D, (1990) ‘Power spectrum and optimal filtering for visual evoked potentials to pattern reversal’, Electroencephalogr Clin Neurophysiol., Vol. 77, No.3, pp. 199-204.
21. Apaydin C., Oguz Y., Agar A., Yargicoglu P., Demir N. and Aksu G. (1993) ‘Visual Evoked potential and optic nerve histopathology in normal and diabetic rates and effect of ginkgo biloba extract’, Acta Ophthalmol., Vol. 71, No. 5, pp. 623-628.
22. Kramarenko A.V. and Tan U. (2002) ‘Validity of Spectral analysis of evoked potentials in brain research’, International Journal of Neuroscience, Vol. 112, pp. 489-499.
23. Tobimatsu S. and Kato M.M. (1996) ‘The effect of binocular stimulation on each component of transient and steady-state VEPs’, Electroencephalogr Clin Neurophysiol., Vol. 100 No.3, pp. 177-83.
24. Nakayama M. (1994) ‘Transient and steady-state electroretinograms and visual evoked potentials to pattern and uniform-field stimulation in humans’, Fukuoka Igaku Zasshi., Vol. 85, No. 7, pp. 225-234.
25. Kulkarni G.R. and Udpikar V. (1995) ‘An integrated facility for data acquisition and analysis of biomedical signals. Case studies on VEP, IVS’, Proc. 14th Conference of the Biomedical Engineering Society of India, pp. 67 – 68.
26. Agar A., Yargicoglu P., Aktekin B., Edremitlioglu M. and Kara C. (2000) ‘The effect of cadmium and experimental diabetes on EEG spectral data’, J. Basic Clin. Physiol. Pharmacol., Vol.11, pp. 17–28.
27. Yargicoglu, P., Agar A., Izgut-Uysal V.N., Senturk U.K. and Oguz Y. (1996) ‘Effect of Chronic Cadmium exposure on VEP and EEG spectral components’, International Journal of Neuroscience, Vol. 85, pp. 173-184.
28. Yargicoglu, P., Agar A., Oguz Y., Izgut-Uysal V.N., Senturk U.K. and Oner G(1997) ‘The effect of Development Exposure to Cadmium (cd) on Visual Evoked Potentials (VEPs) and Lipid Peroxidation’, Neurotoxicology and Teratology, Vol. 19, No. 3, pp. 213-219.
29. Sivakumar R. and Ravindran G. (2004) ‘Automatic Discrimination of Abnormal Subjects Using the Visual Evoked Potential Spectral Components’, International Journal of Biomedicine and Biotechnology, Vol. 4, No. 1, pp.5-9.
30. Sivakumar R. and Ravindran G. (2002) ‘Visual evoked potential spectral component analysis’, Proc. 2nd European Medical & Biological Engineering Conference - EMBEC’02, Austria, pp. 18-20.
Figure 1 Nicolet Viking IV machine
Figure 2 10-20 Electrode System with Electrode position
Figure 3 Pattern Shift Stimulation TV Screen
Figure 4 Electrode locations
Figure 5 Spectral plot with adjacent components at 2Hz and 4Hz
Table 1 Mid-latency Values
| |
S.No |
First dominant peak(Hz) |
Adjacent dominant peak(Hz) |
Mid Latency (msec) |
|
1. |
3 |
2 |
110 |
|
2. |
3 |
4 |
130 |
|
3. |
4 |
3 |
130 |
|
4. |
4 |
5 |
150 |
|
5. |
5 |
4 |
150 |
|
6. |
5 |
6 |
170 |
|
7. |
6 |
5 |
170 |
|
8. |
6 |
7 |
190 |
|
9. |
7 |
6 |
190 |
|
10. |
7 |
8 |
210 |
|
11. |
8 |
7 |
210 |
Table 2 Positive Magnitude Difference vs. Addition latency
S.No |
Magnitude difference |
Additional Latency (msec) |
| 1. |
0,0.1,0.2 |
0 |
| 2. |
0.3,0.4 |
1 |
| 3. |
0.5,0.6 |
2 |
| 4. |
0.7,0.8 |
3 |
| 5. |
0.9,1 |
4 |
| 6. |
1.1,1.2 |
5 |
| 7. |
1.3,1.4 |
6 |
| 8. |
1.5,1.6 |
7 |
| 9. |
1.7,1.8 |
8 |
| 10. |
1.9,2 |
9 |
| 11. |
>20 |
10 |
Table 3 Negative Magnitude Difference vs. Addition latency
S.No |
Magnitude difference |
Additional Latency(msec) |
| 1. |
0,0.1,0.2 |
0 |
| 2. |
0.3,0.4 |
1 |
| 3. |
0.5,0.6 |
2 |
| 4. |
0.7,0.8 |
3 |
| 5. |
0.9,1 |
4 |
| 6. |
1.1,1.2 |
5 |
| 7. |
1.3,1.4 |
6 |
| 8. |
1.5,1.6 |
7 |
| 9. |
1.7,1.8 |
8 |
| 10. |
1.9,2 |
9 |
| 11. |
>20 |
10 |
Figure 6 Intermediate Latency Identification Flow Chart
Figure 7 120msec Latency waveform spectrum
Figure 8 - 113msec Latency waveform spectrum
Table 4 Intermediate Latency Calculation with positive magnitude difference
S.No |
Magnitude difference |
Intermediate latency = Base latency + Addl. Latency (msec) |
| 1. |
0,0.1,0.2 |
110 + 0 = 100 |
| 2. |
0.3,0.4 |
110 + 1 = 111 |
| 3. |
0.5,0.6 |
110 + 2 = 112 |
| 4. |
0.7,0.8 |
110 + 3 = 113 |
| 5. |
0.9,1 |
110 + 4 = 114 |
| 6. |
1.1,1.2 |
110 + 5 = 115 |
| 7. |
1.3,1.4 |
110 + 6 = 116 |
| 8. |
1.5,1.6 |
110 + 7 = 117 |
| 9. |
1.7,1.8 |
110 + 8 = 118 |
| 10. |
1.9,2 |
110 + 9 = 119 |
| 11. |
>20 |
110 + 10 =120 |
Table 5 Intermediate Latency Calculation with Negative magnitude difference
S.No |
Magnitude difference |
Intermediate latency = Base latency + |
|
1. |
0,0.1,0.2 |
110 – 0 = 110 |
|
2. |
0.3,0.4 |
110 – 1 = 109 |
|
3. |
0.5,0.6 |
110 – 2 = 108 |
|
4. |
0.7,0.8 |
110 – 3 =107 |
|
5. |
0.9,1 |
110 – 4 =106 |
|
6. |
1.1,1.2 |
110 – 5 =105 |
|
7. |
1.3,1.4 |
110 – 6 =104 |
|
8. |
1.5,1.6 |
110 – 7 = 103 |
|
9. |
1.7,1.8 |
110 – 8 = 102 |
|
10. |
1.9,2 |
110 – 9 = 101 |
|
11. |
>20 |
110 – 10 = 100 |
Technical College - Bourgas,
All rights reserved,
© March, 2000