Retinal signal processing involves different levels of adaptation processing and these adaptations start in the photoreceptor layer followed by post-receptoral feedback mechanisms. It has been claimed that the non-linear mechanisms in the retina arise predominantly from the inner retina.54 Although the second-order kernel response reflects the mechanisms of temporal interactions in the retina, there is still conflict over the hypothesis that the components in the second-order kernel response reflect the activities of the retinal ganglion cells.55,56
Nevertheless, several studies have used the mfERG to access the physiological response of the (presumably damaged) ganglion cells to detect signs of glaucomatous damage in terms of amplitude57 and implicit time.58 In addition, the changes in amplitude primarily affect the central retina,57 where the latency of mfERG responses showed a significant negative correlation with the mean sensitivity (dB) of static perimetry.58 The amplitude of the mfERG response is also reduced in patients with ocular hypertension59 and the second-order kernel response is also abnormal even in patients with glaucoma or glaucoma suspects with a normal visual field.60 Sakemi, Yoshii and Okisaka61 found that neither the first- nor the second-order kernels of the mfERG showed any changes correlated with glaucomatous visual field abnormalities and questioned its relationship with inner retinal responses. In fact, outer retinal activity has also been found to make a contribution to second-order responses,36,62 and this contribution may complicate the interpretation of the retinal changes in glaucoma. Even though the latency changes have been reported to be more sensitive than amplitude changes in showing glaucomatous visual field defects,58 it has been noted that mfERG changes do not precede glaucomatous defects diagnosed using static perimetry.
How do the retinal ganglion cells contribute to the mfERG response? This is an important question to consider before applying the mfERG to the detection of glaucoma. An experimental model of glaucoma in primates resulting in a loss of retinal ganglion cells showed a marked attenuation of both the first- and second-order mfERG responses63 and the mfERG amplitudes were highly correlated with the density of the surviving retinal ganglion cells. Other studies37,64 demonstrated the inner retinal contribution to the monkey mfERG by recording before and after intra-vitreal injections of N-methyl-D-aspartic acid (NMDA) and tetrodotoxin. Tetrodotoxin blocks the sodium-based action potentials of ganglion and amacrine cells and substantially alters the mfERG from monkeys.62 Further treatment with NMDA removes amacrine cell activity and feedback components, as it depolarises the post-synaptic membranes of ganglion and amacrine cells. These studies support the view that mfERG responses reflect the contribution from the ganglion cells. The mfERG waveform in the monkey is quite different from that in humans even for the first-order kernel response. The first-order kernel responses in the monkey have large naso-temporal variation with double peaks in the waveform,35 but these are not obvious in the human mfERG.
The effects of experimental glaucoma in the monkey on first-order and second-order mfERG responses were similar to those seen under the effects of tetrodotoxin and NMDA.37 The naso-temporal variation and oscillatory potentials also disappeared in the experimental glaucomatous eyes. This suggests that the spiking activity of inner retinal neurons is the cause of the naso-temporal variation across the retina. As shown by a comparison of first- and second-order kernels in primate mfERG, the inner retina has a relatively greater input into the second-order mfERG response.62 In a laser-induced experimental glaucoma model, the second-order mfERG responses were more sensitive to glaucomatous changes.65 Therefore, a sound approach in modification of the protocol or analysis of the mfERG would be to enhance the contribution of inner retinal activity in the mfERG to detect glaucomatous defects.
Optic nerve head component in mfERG
Most previous studies have applied the conventional (fast flickering) mfERG with high contrast stimulation to test for glaucomatous dysfunction;57,59 however, no simple correlation of the topographical mfERG changes and the retinal dysfunction observed in visual field defects has been found.61,66 Thus, a range of stimulation paradigms has been proposed to improve the situation.
One of the most important studies was done by Sutter and Bearse,54 who demonstrated that the human mfERG response contained a component attributable to ganglion cell activity. They used a mathematical algorithm to extract a component with a latency, which increased in proportion to the estimated length of the ganglion cell axons from the site of stimulation to the optic nerve head. They speculated that this component (the so-called optic nerve head component) originated from the ganglion cell axons. They found that glaucomatous damage can reduce the magnitude of this component. This optic nerve head component theory was supported by data from the monkey, where the tetrodotoxin-sensitive component from the mfERG waveform was similar to the optic nerve head component.67 In addition, the marked naso-temporal variation of the monkey mfERG was eliminated by pharmacological suppression of inner retinal activity,37 suggesting that the optic nerve head component derived from ganglion cells is likely to be related to the naso-temporal variation of the mfERG. Although this component exists in the mfERG response, it is quite difficult to observe in most records because it varies in appearance and it is not easy to extract it from the complex waveform of the first-order kernel.
Low contrast mfERG
Several previous studies have reported that the contrast sensitivity in glaucoma is significantly affected, especially at low contrast.68,69 A low contrast stimulus presentation has been proposed for the mfERG and when the stimulus contrast is reduced to 50 per cent, the naso-temporal variation in the first-order kernel in human mfERG becomes obvious.38 As naso-temporal variations in waveform have been hypothesised to arise from ganglion cell activity, it appears that contrast attenuation of the stimulus can increase the relative proportion of the responses coming from the inner retina in the human mfERG.70 Therefore, in an attempt to obtain a better mfERG response from the inner retina, mfERG recordings with reduced stimulus contrast have been applied to detect glaucomatous changes.71 At this contrast level, the human mfERG waveform is similar to that of the monkey, with oscillatory components on the ascending and descending limbs of the first positive peak of the low contrast mfERG. Additionally, for subjects with glaucoma, the later oscillatory component (that is, the one on the descending limb) seems to be reduced in magnitude71,72 (Figure 4). This is not a universal finding and in those with clear abolition of oscillatory components in the mfERG waveform, the changes do not correspond to the localised visual field losses. Although the first-order kernel responses for low-stimulus contrast recordings show obvious changes in glaucoma,73 the sensitivity of these recordings is insufficient to detect inner retinal disease. In addition, the second-order kernel responses are very noisy for low contrast stimulation and are very difficult to study.
Figure 4. (A) Visual field results from two patients with glaucoma (G1 and G2) using the Humphrey Field Analyser II (Central 30-2 threshold measurement). (B) Diminished oscillatory component on the descending limb of the low contrast multifocal electroretinogram waveform (grey circle) from glaucomatous patients compared with that from normal subject (N1).60
Download figure to PowerPoint
Slow-sequence (slow flash stimulation) mfERG
In the conventional fast mfERG, the stimulus is generally displayed on a monitor with a frame rate of 75 Hz, which means that the time interval between two successive flickering stimuli is approximately 13.3 ms. In 1995, Wu and Sutter74 introduced a slow multifocal stimulation to analyse the topographic distribution and non-linearities of oscillatory potentials (OP) in the retinal response. The OP generated from slow multifocal stimulation are likely to be related to feedback from the inner retina. Sano and colleagues75 further modified this slow mfERG paradigm (that is, slow-sequence mfERG) and observed a new wavelet in the mfERG, as the m-sequence presentation for mfERG is slowed by interleaving three grey frames between the presentations of flickering stimuli (Figure 5). The insertion of these additional frames increases the time interval between two successive m-sequence frames by a factor of four. The reported positive wavelet appeared on the descending limb of the first positive peak of this slow-sequence mfERG. The amplitude of this wavelet increased significantly when 30 further grey frames were added between the m-sequence frames. Furthermore, the amplitudes of this unmasked wavelet obtained from the nasal retina were significantly larger than those from the temporal retina, and this naso-temporal variation is considered to be related to the distance of the stimulated area from the optic nerve head. The characteristic of this small wavelet, the so-called s-wave, was not observed in eyes with optic neuritis in mfERG recordings at any presentation frame rate; however, it was shown in unaffected eyes in all patients with unilateral optic neuritis. This wavelet reappeared with recovery from the disease and its recovery was significantly correlated with the recovery of visual acuity. Thus, these findings support the idea that the new wavelet originates from the ganglion cells.75 The oscillatory component on the descending limb of the low contrast mfERG waveform and the new wavelet generated from the slow-sequence mfERG are very similar in terms of the appearance, amplitude and implicit time. This suggests that both applications enhance or unmask inner retinal activity and are most likely to measure the same response in the mfERG, although the stimulation protocols are totally different.
Figure 5. (A) Schematic diagram showing the flash stimulation sequence of the slow-sequence (slow flickering stimulation, MOOO) multifocal electroretinogram (mfERG). (B) The first-order kernel of the slow-sequence mfERG from the central (rings 1 to 2) and peripheral (rings 3 to 6) regions.
Download figure to PowerPoint
Recently, a series of high frequency oscillatory potentials following the dominant component of the first-order kernel response was found in the slow-sequence mfERG stimulation with three dark frames76 (Figure 5). This protocol is different from the above as the luminance of the inserted frames is different, that is, one using grey frames (about 60 cd/m2)75 and the other using dark frames (less than 1.0 cd/m2).76 Under this protocol, a change of the oscillatory potentials was observed in patients with glaucoma. The oscillatory potentials reduced significantly in the central field and in the nasal field for patients with normal tension glaucoma, allowing 85 per cent sensitivity for differentiation from normal subjects. The oscillatory potentials can be divided into fast (143 Hz) and slow (77 Hz) wavelets.77 In monkeys, the fast oscillatory potentials are significantly larger in the temporal than in the nasal retina77,78 and they are reduced in experimental glaucoma with moderate correlation with local visual field sensitivity.77 This suggests that the fast oscillatory potentials are also likely to be related to the activity of retinal ganglion cells.79 All of these slow-sequence mfERG protocols with the insertion of a number of blank frames can minimise the involvement of second-order or higher-order kernel responses and provide an uncontaminated first-order kernel response. This relatively clean first-order kernel response contains less temporal interaction from the retina but probably has a degree of non-linear activity due to spatial interaction. Hence, it can provide different results in glaucoma detection compared with flash ERG without non-linear interaction.
Global flash stimulation mfERG
The above-mentioned applications show a degree of effectiveness in the detection of glaucomatous defects; the higher order interactions involving temporal processing of retinal responses leading to adaptation are of interest. These interactions are one of the core strengths of the mfERG.
The use of higher-order kernels is a method by which the interaction of successive flashes may be examined. The second-order kernel is the simplest analysis containing non-linearity in the retina. A previous study80 suggested that the inner retina involves adaptation mechanism(s) and thus dysfunction of the inner retina may alter the adaptation mechanism(s). The second-order kernel response, depending on the effect of the preceding flash, has also been applied to the study of glaucomatous damage;81 however, it does not show a correlation to visual field loss in glaucoma.
The first-order kernel response is more complex than it appears as the second-order kernel response contributes to the later part of the first-order kernel response81 due to the adaptation influences from the preceding flash. If this non-linear response in the first-order kernel can be enhanced and measured, it would be useful for recording inner retinal activity. Hence, an alternative mfERG protocol using a global flash, which provokes an interaction containing an enhanced non-linearity has been developed.80 The hypothesis here is that if the global flash does not produce an adaptive effect, it would not contribute to the mfERG response because its contribution would be cancelled when the focal responses are extracted (Figure 6).
Figure 6. Schematic diagram for the signal derivation of the first-order kernels of the global flash stimulation (MOF) in the multifocal electroretinogram (mfERG). The bottom waveform is the first-order kernel of the global flash mfERG. It contains two components: direct component (DC) and induced component (IC).
Download figure to PowerPoint
In this global flash stimulation, there are two components in the mfERG response: a direct component (DC) and an induced component (IC) (Figure 6). The direct component is assumed analogous to a conventional mfERG response, while the induced component is the change of the response to the global flash produced by the prior local flash. This non-linear induced component represents adaptive changes in the response and the changes may be generated by the inner retina.80 Different protocols have been proposed to enhance this non-linear component in the retina and most studies have used one periodic global flash interposed between two consecutive focal flashes,82–86 while others have used two87 or three periodic global flashes88 between consecutive focal flash stimulations. Although there are different ways of inserting the global flash in the m-sequence, the aims are the same, namely, to enhance the adaptive effect in the retina so as to measure the inner retinal contribution. It has been claimed that the global flash technique exhibits a large optic nerve head component80,89 and the naso-temporal asymmetry of the induced component response is observed in the human global flash mfERG.83 It has also been shown that the optic nerve head component can be extracted from the induced component epoch and that the loss of the optic nerve head component in glaucoma is more apparent when using the technique. In other studies, the induced component has been shown to be reduced in glaucoma.87,88 Although the amplitude of the induced component response from the nasal retina was most affected in glaucoma,88 a small oscillation in the induced component from the temporal retina was also found to be sensitive in the detection of glaucoma.83 A recent study reported that the induced component of the superior temporal retinal region is the most sensitive parameter for glaucoma differentiation.87 The similar outcomes obtained from different studies indicate that the various global flash protocols can enhance adaptation in the retina for the detection of glaucomatous defects.
The induced component response in the ‘two global flash’ stimulation paradigm can provide a sensitivity of 85 per cent and a specificity of 80 per cent in the detection of POAG.87 This finding is similar to another study83 using a ‘one global flash’ stimulation protocol, which reported a sensitivity of 75 per cent and a specificity of 83 per cent. Thus, the induced component response in the human mfERG has been confirmed to be efficient in detecting glaucoma but the correlation of the induced component response with the corresponding visual field defects in glaucoma is not yet well defined.83,88 In addition, the relatively large inter-subject variation of the induced component response limits its possibility for the assessment of localised glaucomatous damage in individual patients.82
In contrast, the characteristics of the direct component in the global flash mfERG have not been widely studied. The direct component is sensitive to changes in diabetic retinopathy82 and age-related maculopathy84 and it has been suggested to show a contribution from the optic nerve head component.80 Although the direct component has been suggested to be analogous to a conventional mfERG response,80 the direct component also reflects a certain level of adaptive change produced by the periodic global flashes85 because there are temporal interactions between the focal flashes and the periodic global flashes in this paradigm. These interactions are reflected in the change of shape of the direct component from the response of the conventional mfERG.83
A mfERG protocol combining both luminance-modulation (contrast) stimulation and global flash stimulation has recently been proposed.90 This modified global flash mfERG paradigm with various contrast stimuli has been designed to measure temporal adaptive changes in the retina (Figure 7). The low contrast stimulation unmasks the oscillatory component from the inner retina and the global flash stimulation enhances the temporal adaptation in the retina. This combination protocol is believed to further advance the mfERG in the detection of glaucomatous damage. The direct and induced components of the contrast response functions in this modified global flash mfERG showed different characteristics. The direct component responses in normal subjects remain steady at mid- and high-contrast levels, but in subjects with glaucoma the direct component responses show a significant reduction in amplitude at mid-contrast levels and a mild reduction at high-contrast levels (Figure 8); however, the induced component responses show a larger reduction in amplitude only under high-contrast conditions. Subjects with glaucoma showed a loss of contrast saturation in the direct component contrast response function, which is most likely caused by impairment of the fast-adaptation mechanism in the retina. Quantifying this loss by calculating the area under the direct component contrast response function (adaptive index) provides a measure of the intrinsic response changes with contrast levels. Moreover, the adaptive index factors out baseline response amplitude variation and minimises the effect of inter-subject variation of the response amplitude. The adaptive index has been shown to provide good differentiation between normal subjects and glaucomatous patients with a sensitivity of 93 per cent and a specificity of 95 per cent.90 The adaptive index also illustrated a good correlation with the glaucomatous visual field defects and this has not been reported previously.
Figure 7. (A) Schematic diagram showing the luminance modulation (contrast) for the global flash multifocal electroretinogram (mfERG) stimulation (MOFO). (B) The first-order kernel of the global flash mfERG at different contrast levels. Both the direct component (DC) and the indirect component (IC) increase in magnitude as the stimulus contrast increases.
Download figure to PowerPoint
Figure 8. The luminance-modulated (contrast) response function of the grouped average (A) direct component (DC) and (B) indirect component (IC) responses (ring 4 to ring 6) from normal subjects and glaucomatous patients. The luminance-modulated response function for the DC in normal subjects shows a saturation characteristic as the stimulus contrast increases; the response function for the IC increases linearly with the stimulus contrast. Glaucoma subjects show a decrease in both DC and IC response amplitudes at all contrast levels. Bars indicate standard deviation.
Download figure to PowerPoint
Furthermore, clinically normal fellow eyes of patients with unilateral glaucoma were found to have impaired adaptation in the retina.91 The adaptive index in the fellow eyes was severely reduced and close to the value from the glaucomatous eyes. Thus, fellow eyes that were clinically normal had already shown abnormal changes in the retinal adaptive mechanism and this allows these fellow eyes to be differentiated from normal by the value of the adaptive index. Significant reductions in the adaptive index occur before any defined visual field loss in the fellow eyes of patients with unilateral glaucoma.92 These findings confirm that an impaired retinal adaptive mechanism occurs before observed visual field abnormalities in patients at high risk of glaucoma. As the mfERG is a tool used to study the temporal processing of the retinal responses leading to adaptation, appropriate protocols in mfERG measurement can help to advance the detection of glaucomatous defects as well as assist in early diagnosis.
A question is raised of how the induced component could be useful in assessing inner retinal function if the direct component is also reduced. Since the direct component is believed to come from the outer to mid-retinal layers and the induced component is believed to come from the inner retinal layers,93 the magnitude of the direct component should have an influence on the magnitude of the induced component; however, the characteristics of the direct component and the induced component in the contrast response functions in glaucoma patients are totally different compared with normal subjects. The function of the direct component in glaucomatous subjects was significantly decreased at the mid-contrast level, while the function of the induced component in glaucomatous subjects was found to have a significant reduction at the high-contrast level. The different patterns in response reduction of direct and induced components imply that the changes of these two components in glaucoma have a certain level of independence. A recent study investigated the cellular contributions to the global flash mfERG by pharmacological dissection in porcine eyes.93 The inner retinal activity partially contributes to the direct component with superimposed regular oscillation-like wavelets. Hence, eye diseases (such as glaucoma) involving damage to the inner retinal layers may reduce oscillation-like wavelets contributing to the direct component and ultimately alter the characteristics of the direct component contrast response function. For retinal diseases with outer retinal dysfunction it cannot be excluded that the magnitude of the induced component may be affected primarily by the direct component.
Time-frequency analysis of the mfERG
Recently, another approach has been taken to the analysis of mfERG data. The usual analysis is the measurement of the response in terms of peak-to-peak or root-mean-square of the amplitude. The more recent analysis is a wavelet analysis, in which the oscillatory potentials are extracted from the mfERG waveform. The frequency of the oscillatory potentials are studied as an indicator of glaucomatous damage.79,94 This analytic method can apply in different stimulation protocols of mfERG, for example, slow-sequence stimulation79 and global flash stimulation.94 Although there are different methods for wavelet analysis, oscillatory potentials are the target of the mfERG response and these are most likely to be contributed from the inner retina74 and related to glaucoma.76
Comparison with mfVEP
Another multifocal technique for the detection of glaucoma is the multifocal visual evoked potentials (mfVEP), which was introduced in 1994.95 Numerous studies have reported good correlation between the results of mfVEP and the findings of visual field defects in glaucoma.96–105 There is no evidence to demonstrate the superiority of mfVEP over mfERG in the detection of glaucoma. The mfERG and the mfVEP are objective perimetric techniques, which measure responses from the retina and from the retina through to the visual cortex, respectively. They are limited by different factors, which affect sensitivity and specificity. Therefore, various approaches and protocols have been introduced to improve their usage in a clinical situation. In addition, both techniques are influenced by other physiological changes, for example, outer retinal changes would affect the findings of mfERG as well as mfVEP, and optic nerve or optic tract lesions other than glaucoma would affect the results of mfVEP. At present, neither mfERG nor mfVEP can give an adequate diagnosis of glaucoma. With further advances and in conjunction with other clinical tests, both techniques will help to provide accurate diagnosis and monitoring of glaucoma.