- Open Access
Multifocal ERG wavelet packet decomposition applied to glaucoma diagnosis
© Miguel-Jiménez et al; licensee BioMed Central Ltd. 2011
- Received: 11 February 2011
- Accepted: 17 May 2011
- Published: 17 May 2011
Glaucoma is the second-leading cause of blindness worldwide and early diagnosis is essential to its treatment. Current clinical methods based on multifocal electroretinography (mfERG) essentially involve measurement of amplitudes and latencies and assume standard signal morphology. This paper presents a new method based on wavelet packet analysis of global-flash multifocal electroretinogram signals.
This study comprised twenty-five patients diagnosed with OAG and twenty-five control subjects. Their mfERG recordings data were used to develop the algorithm method based on wavelet packet analysis. By reconstructing the third wavelet packet contained in the fourth decomposition level (ADAA4) of the mfERG recording, it is possible to obtain a signal from which to extract a marker in the 60-80 ms time interval.
The marker found comprises oscillatory potentials with a negative-slope basal line in the case of glaucomatous recordings and a positive-slope basal line in the case of normal signals. Application of the optimal threshold calculated in the validation cases showed that the technique proposed achieved a sensitivity of 0.81 and validation specificity of 0.73.
This new method based on mfERG analysis may be reliable enough to detect functional deficits that are not apparent using current automated perimetry tests. As new stimulation and analysis protocols develop, mfERG has the potential to become a useful tool in early detection of glaucoma-related functional deficits.
- Discrete Wavelet Transform
- Wavelet Packet
- Wavelet Packet Decomposition
- Humphrey Visual Field
Alternative approaches using objective measures of glaucomatous neuropathy that do not rely on psycho-physiological or structural testing have been investigated in recent years. One approach has been to use electroretinography (ERG) to measure the changes in electrical activity generated by retinal ganglion cell bodies or axons in glaucoma .
Use of ERG to detect glaucoma requires isolation of specific components related to ganglion cell responses. Several ERG techniques involving measurement of light-adapted (photopic) and dark-adapted (scotopic) full-field flash electroretinograms have been investigated. This research into use of ERG in experimental glaucoma has produced clear evidence to suggest that electro-physiological tools can detect early functional changes in glaucoma .
A potentially more effective procedure is multifocal ERG (mfERG) , which takes simultaneous recordings of focal responses from over 100 different retinal regions and uses them to produce topographic representations of retinal response components.
The most common methods used to analyse the mfERG signal are based on amplitude and latency waveform analysis. For example, in subjects with primary open-angle glaucoma OAG, the amplitudes decrease while the latencies may increase . Other approaches have used structural pattern analysis  to extract waveform identity patterns that may then be classified using a neural network. Zhou et al. have used the matching pursuit analysis method, a time-frequency analysis, to identify and characterize oscillatory potentials in the mfERG recording in primates .
The current paper represents a continuation of our previously published work .
The patients, methods of analysis and the results are new. Both studies have the same goal (glaucoma detection), but use different analysis tools: Discrete Wavelet Transform (DWT) in , versus Discrete Wavelet Packet Transform (DWPT) in this work. DWPT is an extension of the DWT to the full binary tree . In the discrete wavelet packet transform, both the scaling and wavelet coefficients are subject to the high-pass and low-pass filtering when computing the next layer scaling and wavelet coefficients. DWPT permits the detail functions to be further split into two or more subbands , which offers a richer signal analysis (discontinuity in higher derivatives, self-similarity, etc.) .
The markers obtained in both works are clearly different. In the previous work we obtained a marker based on the latency of a valley and another marker based on the latency of an edge. In this paper we obtain a marker based on the slope of the baseline of some oscillatory potentials.
This paper studies application of the wavelet packet transform in mfERG analysis. By reconstructing the third wavelet packet contained in the fourth decomposition level (ADAA4) of the mfERG recording, it is possible to obtain a signal from which to extract a marker in the 60-80 ms time interval. This marker comprises oscillatory potentials with a negative-slope basal line in the case of glaucomatous recordings and a positive-slope basal line in the case of normal recordings and it can be reliably used to differentiate between normal and glaucomatous mfERG waveform signals.
This study comprised twenty-five patients diagnosed with OAG and twenty-five control subjects (mean age: 47.5 (SD +/-2.5) for control group, 50.73 (SD +/- 3.8) for glaucoma group). For the purposes of analysis, normal and abnormal waveform databases were created from control subjects' and patients' mfERG recordings. These data were used to develop the algorithm method based on wavelet packet analysis.
Abnormal mfERG signals from glaucomatous patients were selected based on the same criteria followed in . Informed consent was obtained from all participants. The University of Alcalá approved all the protocols and the study was conducted in accordance with the tenets of the Declaration of Helsinki.
Control subjects with normal eyes were included in this study to establish an age-matched norm. All control subjects' eyes had an intraocular pressure (IOP) of 21 mmHg or less (with no history of increased intraocular pressure). Control subjects were screened by means of an ophthalmoscopic examination to confirm the healthy appearance of the optic disc and all had normal Humphrey Visual Field (HVF) test results. All patients' eyes IOP were kept under 21 mmHg with glaucoma eye drops.
The stimulus was viewed through pupils (minimum diameter of 7 mm) pharmacologically dilated with tropicamide (1%). A Burian-Allen bipolar contact lens electrode (Hansen Ophthalmics, Iowa City, IA) was placed on the eye after administering a topical anaesthetic (0.5% proparacaine). Residual spherical refractive error was corrected with the VERIS™ refractor/camera unit mounted on the stimulating monitor. Fixation stability and alignment of the patient's pupil with the refractor's optics were monitored with a built-in infrared camera. Each monocular recording lasted approximately 9 minutes (m-sequence exponent m = 13). For patient comfort, the recording was taken in 16 segments of about 30 seconds each. Segments contaminated by eye movements were discarded and rerecorded. Signals were amplified with a Grass Neurodata Model 12 amplifier system (Grass Telefactor, NH) with a gain of 50,000, band-pass filters (10-300 Hz) and a sampling interval of 0.83 ms (1200 Hz).
Data were analysed off-line using the VERIS™ Science 5.1 software. Artefacts due to blinks and eye movements were removed from the data using two iterations of the VERIS™ artefact removal algorithm. One iteration of spatial filtering (averaging each focal waveform with 30% of its six neighbours) was applied to increase the signal-to-noise ratio. The response epoch selected for the analysis comprised the induced component (60-90 ms) judged to have the largest contribution from the optic nerve head component (ONHC). ONHC decomposition analysis and waveform quality assessment were performed using the VERIS™ 5.1 pro software. Further wavelet signal analysis was performed in MATLAB (MathWorks Inc, Natick, MA).
Only glaucomatous sectors from patients affected by glaucoma made up the abnormal database, the total number of abnormal sectors was 723. Recordings from different numbers of patients could contribute to each sector. Sectors 1 and 2 had the least number of contributing records (3 each), sector 20 had the highest number of records (24) (SD = 5.33). The normal database was made up of 1400 sectors (25 controls, one eye per control, 56 sectors per eye). The minimum and maximum number of sectors per patient was 14 and 37, respectively (SD = 7.6).
mfERG wavelet packet decomposition analysis
Thus, the values for scale j compress or stretch the wavelet. When stretched, the wavelet covers a larger time scale and is able to follow slower changes in the signal (related to low frequencies). When compressed, the wavelet captures finer signal details (related to high frequencies). The parameter k produces a translation of the wavelet in the time domain. The discrete wavelet transform [11–13] analyses the signal at different frequencies with different resolutions, using regions with windowing of different sizes and obtaining a two-dimensional time-frequency function as a result.
Results obtained using the marker within the ADAA4 wavelet packet
PPV = 0.52
NPV = 0.91
Sensitivity = 0.81
Specificity = 0.73
The global-flash mfERG paradigm protocol used in this study provides a reliable and objective measure of visual loss in glaucomatous patients. This stimulation paradigm was able to extract a large ONHC contribution from the mfERG responses, thereby making it easier to detect waveform abnormalities.
Analysis of the signals obtained from wavelet packet decomposition showed a clear repetitive pattern in the signal reconstructed from wavelet packet ADAA4 in the time interval running from 60-80 ms within the induced component. Applying the previously described analysis reveals a variation in the value of the slope of the basal line (Figure 8). In the case of recordings of normal sectors, the pattern consisted of a quasi-sinusoidal waveform section with a rising basal line, while recordings of glaucomatous sectors produced a falling basal line. The slope of the basal line was approximated by the method of least squares in the 60-80 ms time interval.
Application of the optimal threshold calculated in the validation cases showed good sensitivity and specificity. Nevertheless, a small percentage of sectors were still classified incorrectly (Table 1). In this respect, use of different types of amplitude and latency analysis on similar mfERG signals have also shown good sensitivity and specificity [15–17].
Studies of nerve fibre layer thickness have shown that glaucomatous damage can be present in the visual field hemifield with normal achromatic sensitivity . In a recent study using FDP, it has been shown that in patients with OAG with established hemifield defects, 41% of 49 hemifields with apparently normal fields produced abnormal FDP results . Also, several studies using SWAP show that this perimetric technique may be able to detect visual field defects before white-on-white perimetry in cases of suspected glaucoma and may detect earlier progression of visual field defects in glaucoma patients .
The principal purposes of this study were to develop a new mfERG-paradigm glaucoma analysis protocol and to gain a better understanding of how the mfERG and HVF techniques compare. However, this paper does not try to determine, at this stage, whether mfERG or automated achromatic perimetry is better at detecting glaucomatous damage. The authors are aware that such a longitudinal study would require a larger group of control subjects and patients tested with both techniques so that specificity, sensitivity and likelihood ratios could be correctly determined.
This study provides evidence that this new mfERG analysis method may be reliable enough to detect and map functional deficits that are not apparent using current automated perimetry tests. As new stimulation and analysis protocols develop, the authors believe that mfERG has the potential to become a useful tool in early detection of glaucoma-related functional deficits, as well as in longitudinal assessment of the same.
Supported by grants from FIS and FISCAM 09.
- Hood DC, Odel JG, Winn BJ: The multifocal visual evoked potential. J Neuroophthalmol 2003, 23: 279–289.View ArticleGoogle Scholar
- Harwerth RS, Crawford MLJ, Frishman LJ, Viswanathan S, Smith EL, Carter-Dawson L: Visual field defects and neural losses from experimental glaucoma. Prog Retin Eye Res 2002, 21: 91–125. 10.1016/S1350-9462(01)00022-2View ArticleGoogle Scholar
- Sutter EE, Tran D: The field topography of ERG components in man-I. The photopic luminance response. Vision Res 1992, 32: 433–446. 10.1016/0042-6989(92)90235-BView ArticleGoogle Scholar
- Hasegawa S, Takagi M, Usui T: Waveform changes of the first-order multifocal electroretinogram in patients with glaucoma. Invest Ophthalmol Vis Sci 2000, 41(6):1597–1603.Google Scholar
- Miguel-Jiménez JM, Ortega S, Boquete L, Rodríguez-Ascariz JM, Blanco R: Multifocal electroretinography: structural pattern analysis and early glaucoma detection. Electronics Letters 2009, 45(22):1113–1115. 10.1049/el.2009.1195View ArticleGoogle Scholar
- Zhou W, Rangaswamy N, Ktonas P, Frishman LJ: Oscillatory potentials of the slow-sequence multifocal ERG in primates extracted using the Matching Pursuit method. Vision Research 2007, 47(15):2021–2036. 10.1016/j.visres.2007.03.021View ArticleGoogle Scholar
- Miguel-Jiménez JM, Boquete L, Ortega S, Rodríguez-Ascariz JM, Blanco R: Glaucoma detection by wavelet-based analysis of the global flash multifocal electroretinogram. Medical Engineering & Physics 2010, 32(6):617–622. 10.1016/j.medengphy.2010.02.019View ArticleGoogle Scholar
- Zhang SW, Huang DS, Wang SL: A method of tumor classification based on wavelet packet transforms and neighborhood rough set. Computers in Biology and Medicine 2010, 40(4):430–437. 10.1016/j.compbiomed.2010.02.007MathSciNetView ArticleGoogle Scholar
- Samarb VJ, Bopardikarc A, Raod R, Swartz K: Wavelet Analysis of Neuroelectric Waveforms: A Conceptual Tutorial. Brain and Language 1999, 66(1):7–60. 10.1006/brln.1998.2024View ArticleGoogle Scholar
- Garcia C, Tziritas G: Face Detection Using Quantized Skin Color Regions Merging and Wavelet Packet Analysis. IEEE Transactions on multimedia 1999, 1(3):264–277. 10.1109/6046.784465View ArticleGoogle Scholar
- Daubechies I: The wavelet transform, time-frequency localization and signal analysis. IEEE Transactions on Information Theory 1990, 36(5):961–1005. 10.1109/18.57199MathSciNetView ArticleGoogle Scholar
- Kaiser G: A friendly guide to wavelets. Birkhauser, Boston 1994.Google Scholar
- Mallat S: A wavelet tour of signal processing. Second edition. Academic Press, San Diego, CA; 1999.Google Scholar
- Misiti M, Misiti Y, Oppenheim G, Poggi JM: Wavelet toolbox. The Mathworks, Inc; 1997.Google Scholar
- Chu PH, Chan HH, Brown B: Glaucoma detection is facilitated by luminance modulation of the global flash multifocal electroretinogram. Invest Ophthalmol Vis Sci 2006, 47(3):929–37. 10.1167/iovs.05-0891View ArticleGoogle Scholar
- Chu PH, Chan HH, Brown B: Luminance-modulated adaptation of global flash mfERG: fellow eye losses in asymmetric glaucoma. Invest Ophthalmol Vis Sci 2007, 48(6):2626–33. 10.1167/iovs.06-0962View ArticleGoogle Scholar
- Fortune B, Wang L, Bui BV, Cull G, Dong J, Cioffi GA: Local ganglion cell contributions to the macaque electroretinogram revealed by experimental nerve fiber layer bundle defect. Invest Ophthalmol Vis Sci 2003, 44: 4567–4579. 10.1167/iovs.03-0200View ArticleGoogle Scholar
- Kook MS, Sung K, Kim S, Park R, Wang W: Study of retinal nerve fiber layer thickness in eyes with high tension glaucoma and hemifield defect. Br J Ophthalmol 2001, 85: 1167–1170. 10.1136/bjo.85.10.1167View ArticleGoogle Scholar
- Wu LL, Suzuki Y, Kunimatsu S, Araie M, Iwase A, Tomita G: Frequency doubling technology and confocal scanning ophthalmoscopic optic disc analysis in open-angle glaucoma with hemifield defect. J Glaucoma 2001, 10: 256–260. 10.1097/00061198-200108000-00002View ArticleGoogle Scholar
- Bayer AU, Erb C: Short-wavelength automated perimetry, frequency doubling technology perimetry and pattern electroretinography for prediction of progressive glaucomatous standard visual field defects. Ophthalmology 2002, 109: 1009–1017. 10.1016/S0161-6420(02)01015-1View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.