Clinical electrophysiology of the visual system provides an important adjunct to the bedside evaluation of patients with neurologic diseases. It provides the only means available for objective assessment of visual function, especially of the retina and optic nerve, and requires less active participation than subjective evaluations such as perimetry and visual acuity testing. Electrophysiologic techniques can also provide localizing information in the visual pathway and—given the functional organization of the visual system—adjustment of stimulus parameters can be used to define the specific cell type or subpathways affected by an injury. Electrophysiologic techniques play an important role in discriminating diseases of the retina and optic nerve when the distinction cannot be made on standard clinical grounds alone. They also may have use in distinguishing types of injuries of the optic nerve (e.g., glaucomatous, inflammatory, or metabolic), based on assessments tailored to the evaluation of particular functional systems (such as motion or color). The utility of visual electrophysiology is established firmly in the diagnosis and monitoring of patients with ophthalmologic conditions, including neurologic conditions with well-recognized ophthalmologic manifestations such as multiple sclerosis (MS). In addition, clinical research suggests that visual electrophysiology could be used to investigate other neuroinflammatory and neurodegenerative diseases, with the visual system used as a model.
Anatomy of the anterior visual system
The neurosensory retina is one of the three traditional anatomic subdivisions of the central nervous system (the others being brain and spinal cord). The projections of the retina are characterized extremely well and make up a significant volume of the cerebral hemispheres. Of our senses, vision has the greatest amount of cortical surface area dedicated to processing its output and up to 40 percent of the brain has been estimated to be devoted to this effort.
At peak function, the human eye can discriminate up to 10 million shades of color, is sensitive to the flux of a single photon in a darkened room, and can detect flashes of light as brief as 10 to 20 msec. These physiologic features of the human visual system are embedded in the anatomic substrate that makes up the retina, visual pathways, visual cortex, and accessory visual areas. The retina is comprised of 110 million cells divided into at least ten anatomic layers and three primary subsets of neurons. More than 50 different cell types have been described, and ten different neurotransmitters play an important role in retinal physiology. Photoreceptors are the primary sensory neurons of the retina and the most abundant of all the retinal cell types (approximately 90 million rods and 5 million cones on average). They are positioned in the deepest layers of the retina (closest to the retinal pigment epithelium, choroid, and sclera), meaning that light has to traverse all the other layers of the retina before interacting with a photoreceptor ( Fig. 22-1 ). Photoreceptors evidence intrinsic photosensitivity and participate in basic phototransduction—turning light into neuronal signal. Photoreceptors respond to light by converting 11-cis-retinal to 11-trans-retinal, resulting in the hyperpolarization of the neuron and leading to a reduction in release of glutamate from its synaptic terminal.
Photoreceptors are subdivided further into two classes of cells: rods and cones (which are distinguished from one another both by the morphology of their outer segments and by the type of opsin they contain). Rods are highly sensitive to light and evidence peak function in low light conditions. Cones, by contrast, are color sensitive and provide their predominant perceptual input under bright-light (photopic) conditions. In normal individuals there are three cone subtypes (short-, medium-, and long-wavelength), each tuned to a different peak spectral sensitivity within the visible spectrum. Cone and rod distribution is not uniform throughout the retina, as cones are concentrated in the fovea, making up nearly all the photoreceptors in the central 1 degree of the retina, with a steep decline in density out to about 3 degrees eccentricity, beyond which they are distributed evenly throughout the retina. Rods, by contrast, have a peak density around 6 to 8 degrees from the center of the retina, but also show declining density with greater eccentricity outside of this boundary.
The middle layers of the retina consist of interneurons that perform the initial processing and organizing of visual information, including the distribution of visual data into parallel processing streams. Photoreceptors synapse with the cells of the inner nuclear layer (bipolar and horizontal cells) in the outer plexiform layer of the retina. Horizontal cells provide a wide neural network for inter-retinal feedback and processing of visual information (see Fig. 22-1 ). Their organization is likely the basis for the antagonistic, concentric, center-surround response characteristics of bipolar and retinal ganglion cells. Signals arising from photons striking photoreceptors in the receptive field of a single bipolar cell or retinal ganglion cell converge on two separate pools, some of which enhance neuronal firing and others of which inhibit neuronal responses. These stripes of inhibitory and excitatory input are organized in a target-like pattern and their size is not symmetric across the retina. Neurons with larger receptive fields and thereby larger “centers” and “surrounds” can be found at greater eccentricity from central vision. The center-surround organization of retinal neurons likely plays an important part in the sensitivity of the retina to the pattern-reversal electrophysiologic techniques described later. Furthermore, there are many subtypes of bipolar cells, each of which synapses with either groups of rods or cones exclusively. Bipolar cells can be characterized as ON or OFF based on how they respond to photoreceptor hyperpolarization and their response to the reduction of glutamate input that results (depolarizing in the case of ON-bipolar cells).
Finally, retinal ganglion cells constitute the primary output neurons of the retina. They receive input from the bipolar cells (via amacrine cells) and synapse primarily in the thalamus, as well as in the tectum and hypothalamus. There are approximately 1.2 million retinal ganglion cells in the human retina, 70 percent of which are found in the central 30 degrees of the retina. It is important to note that retinal ganglion cells typically receive input from a large number of bipolar cells (which, in turn, have received input from a number of photoreceptors). As a consequence, the receptive field of retinal ganglion cells frequently is substantially larger than the corresponding receptive field of the photoreceptors or bipolar cells that provide its input (see Fig. 22-1 ). In addition, ganglion (and bipolar cells) do not respond equally to a stimulus at all points in their receptive field because of their center-surround organization.
There are four primary types of retinal ganglion cells. Magnocellular (also known as alpha or Y-type) retinal ganglion cells have a large receptive field and a large axonal diameter (and hence faster conduction of their action potential). They are concentrated outside of the macula and are tuned primarily for detection of contrast and motion. Parvocellular retinal ganglion cells, conversely, have a smaller receptive field, a smaller axonal diameter, and a relatively slower conduction time, and are concentrated in the macula. Many parvocellular retinal ganglion cells are tuned also for the discrimination of color. Koniocellular retinal ganglion cells are the smallest in size and least well characterized. Finally, a class of intrinsically photosensitive retinal ganglion cells containing melanopsin are distributed evenly throughout the retina. These melanopsin-containing retinal ganglion cells likely underlie subjective luminance sensitivity, provide some—if not most—of the afferent input for the pupillary response, and help entrain the circadian rhythm. In addition to participating in the continued segregation and analysis of visual data, retinal ganglion cells carry retinal information to the brain for further processing.
After traversing the inner surface of the retina, retinal ganglion cell axons coalesce into the optic nerve, which carries visual information through the orbit and into the brain. The axons are myelinated once they exit the eye posterior to the lamina cribrosa. Visual information from the two eyes arrives together at the optic chiasm before undergoing an anatomic separation that reflects their functional segregation by hemifield. These axons travel via the optic tract to synapse in the thalamus (sensory visual information), hypothalamus (circadian rhythm), and midbrain tectum (luminance sensitivity and pupillary response). Each of the first three subtypes of retinal ganglion cells synapse in different layers of the lateral geniculate nucleus (LGN) of the thalamus, where visual information is processed further before being taken to the visual cortex for additional processing and decoding. Only a small number of melanopsin-containing retinal ganglion cells synapse in the intergeniculate leaflet of the LGN; the majority of them synapse in the midbrain and hypothalamus.
The predominant subnucleus of the LGN (the dorsal LGN) receives retinal input segregated by eye and retinal ganglion cell subtype. Inputs from magnocellular retinal ganglion cells and parvocellular retinal ganglion cells are separated into distinct lamellae, while ON and OFF retinal ganglion cells are also separated in less distinct laminae within these lamellae. The interneurons of the dorsal LGN are also involved in signal processing of visual information. In contrast to retinal ganglion cells, neurons in the LGN are selectively responsive to stimuli with a particular orientation or direction of motion. However, similar to retinal ganglion cells, relay neurons in the dorsal LGN display a center-surround response profile, which means that the LGN enhances the discriminating power of the retinal input and allows for small variations in the visual scene to be highlighted by the visual cortex. Dorsal LGN cells send their axons primarily to the primary visual cortex (Brodmann area 17), which surrounds the calcarine fissure in the occipital cortex. This is also referred to as the striate cortex because of the unique Gennari stripe that can be identified on histologic cross-section and which itself provides horizontal interconnections between areas of the primary visual cortex.
Outputs from the dorsal LGN primarily terminate in layer 4 of the striate cortex (one of the so-called granular cortical layers). The striate cortex (also called V1 or primary visual cortex) dedicates 65 percent of its area to the central 15 degrees of the visual field (out of approximately 150 degrees and 120 degrees across the horizontal and vertical meridians, respectively). Primary visual cortex maintains the general retinotopic organization of visual information with visual information from central vision directly including the depths and banks of the calcarine fissure. Cortex dedicated to the peripheral field includes the surrounding cortical areas anteriorly to the parieto-occipital fissure and posteriorly along the surface of the occipital pole. In total, striate cortex comprises approximately 4 percent of the whole human cortical surface, and more than half of this is dedicated to macular vision. Visual information arising from the portion of the retina subserving central vision is therefore relatively “over-represented,” which has important consequences for the visual evoked potential.
Visual information is deciphered and processed further in accessory visual cortex, which includes a wide swath of neighboring cortex in areas that have been distinguished based on functional determinants (V2–V6). These areas do not have the precise anatomic and histologic landmarks of the Gennari stripe and are characterized best via functional characterization in the primate, although lesional studies demonstrate their existence in humans. Outside V1, the stream of visual information changes significantly and no longer is characterized by the serial relay of visual information and stepwise modulation of visual data. Instead, extrastriate visual processing reflects a loose hierarchy with ample feedback and feedforward loops of interconnection. Although the retinotopic organization is maintained in some of these regions, others demonstrate neurons with extremely large receptive fields tuned for particular features of the visual scene, such as faces (including emotional state), animals, or rates of motion. Two classic pathways arising from the visual cortex include the so-called dorsal stream carrying visual information to the parietal lobes for spatial identification (“Where?”) and the ventral stream carrying information to the temporal lobes for object identification (“What?”). In addition, much of the primary visual cortex is comprised of layers (besides layer 4C) that receive inhibitory and modulating inputs from extrastriate areas that help to focus visual processing power on features deemed salient.
When considering the implications of the organizational structure of the visual system for clinical electrophysiology, it is important to remember that visual perception is not the consequence of parallel stimulation of afferent neurons, but that each layer of the visual system influences how much of the arriving visual input is transmitted to the next relay. Furthermore, our visual system is designed to discriminate variation in the visual scene, not strictly to quantify the intensity of visual stimuli. This objective is encoded in the architecture and physiology of the system for processing visual information. Understanding this purpose for our visual system is important in understanding how a particular visual stimulus will drive an electrophysiologic response and how injury might impact the response recorded.
Clinical electrophysiology of the visual system
Visual evoked potentials (VEPs) are cortically generated electrical potentials recorded over the scalp in response to a visual stimulus. Electroretinograms (ERGs) are retinally generated potentials recorded from the cornea or periorbital skin. However, the size of the potentials elicited in both the VEP and ERG is small compared with ambient electrical signals that reflect the ongoing activity of brain, heart, and muscle. Therefore, signal-averaging techniques are employed to help resolve both the VEP and the ERG from background electrical potentials related to nonstimulus activity or ambient conditions (“noise”). The amount of signal averaging required is entirely dependent on the signal-to-noise ratio (SNR), but a reasonable rule of thumb for pattern-reversal full-field VEPs is that 50 to 250 averaged samples are required to achieve an adequate SNR. Averaging paradigms are determined partially by the stimulus and specific technique being employed, as averaging is not always required for ERG acquired with a contact lens electrode (and less averaging typically is required for the ERG than VEP). In general, averaging is achieved by recording the responses to repeated stimulation, and clinical studies require several minutes of recording time. The noise reduction provided by averaging can be calculated by the inverse of the square root of the number of averages and there are therefore diminishing returns afforded by increased averaging.
Basic Aspects of the ERG
ERG can be recorded from the corneal surface by a contact lens embedded with a corneal ring electrode and a conjunctival reference, or from periorbital skin by small foil electrodes attached at the lateral canthi. Dermal electrodes significantly improve patient comfort and are recommended particularly for children and infants, but usually require additional signal averaging to optimize SNR.
The standard ERG stimulus is a high-intensity light stimulus such as a strobe light or a Ganzfeld (German for “whole-field”) stimulator (a hollow dome to which a patient affixes the gaze to obtain a brief but uniform stimulus to the entire retina). The stimulus used can be modulated to optimize assessment of rod or cone response. Cones have a shorter refractory period and therefore a higher-frequency stimulus can be used (30 Hz) to aid in the assessment of their function. (Rods provide minimal response at a frequency greater than 20 Hz.) It is particularly important to utilize methods to optimize cone responses, as there are over ten times more rods than cones, and the rod response on ERG is an order of magnitude greater than the cone response. Colored filters can also be employed to isolate rod and cone responses further. Rod responses have their peak spectral sensitivity in near blue light (510-nm wavelength) and cones as a group have their peak sensitivity in the yellow range (560 nm). Furthermore, red cones exhibit a response with little to no contamination from rods, given the limited overlap in peak sensitivities. In addition, ambient lighting conditions can be varied to help distinguish rod and cone responses. This is significant because the retinal diseases for which ERG is clinically useful have differential effects on rod and cone populations.
The ERG typically is recorded with a patient’s pupils dilated after the application of mydriatics. This helps to ensure uniform illumination of the retina and prevent pupillary responses from influencing the character of the ERG tracing. The ERG can also be recorded after light adaptation (photopic) to quench the effects of rods, or dark adaptation (scotopic) to enhance rod contribution. There is a fair amount of variability in the protocols employed but, as an example, dark adaptation usually requires that the patient sit in a darkened room for 30 minutes prior to recording.
Although there are differing conventions among ophthalmologists and neurologists, ERG waveforms frequently are described as negative when the tracing dips below the baseline and positive when they cross above it ( Fig. 22-2 ). ERG typically is analyzed by evaluating the first negative deflection (a-wave) and the subsequent positive deflection (b-wave). In photopic recordings, especially with a bright flash as stimulus, a high-amplitude positive wave may precede the a-wave. The a-wave reflects the depolarization of photoreceptors en masse (rods or cones predominantly depending on stimulus parameters and preconditioning state), and the b-wave reflects the activation of the retinal interneurons from the inner nuclear layer. The times to peak of the a- and b-waves are referred to as “implicit time” rather than “latency.” In normal individuals the a-wave is observed at around 15 msec and the b-wave at around 30 to 35 msec in photopic ERGs. In photopic recordings the implicit times are usually shorter (faster) but the amplitude of the response is usually lower than in dark-adapted ERGs. Additional c- and d-waves have been described and attributed to retinal pigment epithelium and OFF bipolar cells, respectively, but these are not assessed routinely in the clinic.
The ERG can also be recorded with an alternating checkerboard or other pattern-reversal stimulus. The pattern-reversal ERG (PERG) is described by the same conventions that govern flash ERG. (Positive deflections are recorded above the baseline, and negative below.) It is characterized by a negative wave at approximately 50 msec (N50) and a positive deflection at around 95 msec (P95). The P95 (also referred to as the PERG b-wave) is thought to reflect activation of the retinal ganglion cells, as some of these cells are tuned to sense spatial contrast, and the implicit time of the PERG implicates neuronal populations downstream from the photoreceptors. In an extremely large retrospective case series, it was reported that subjects with abnormal VEPs and known disease of the anterior visual pathway frequently lost the N95 component of their PERG but almost always showed complete preservation of the P50. However, PERG amplitudes are small and sometimes unreliable and, as a consequence, PERG has found application primarily as a research tool.
Flash and pattern-reversal VEPs
Prior to 1970, VEPs were only elicited using a diffuse stroboscopic or Ganzfeld flash as the stimulus. Additional stimulus approaches were developed subsequently, including checkerboard pattern-reversal and sinusoidal grating patterns. The flash VEP is much less sensitive than pattern-reversal VEPs for detecting abnormalities such as optic neuritis. In addition, flash VEP responses have higher interindividual variation and are more dependent on the patient’s state of arousal than pattern-reversal VEP, limiting their routine clinical utility. Their primary clinical use is restricted to patients who are unable to fixate or who are suspected of feigning complete blindness.
Transient and Steady-State VEPs
After stimulation, a transient, cortically generated response is tied to the relevant stimulus (the “transient VEP”) and consists of a series of potentials that are alternately positive and negative in polarity. They are labeled on the basis of their polarity and latency. With repeated high-frequency (over 4-Hz) stimulation, however, cortical responses remain constant—or nearly so—with respect to amplitude and phase (“steady-state VEP”). Regan likened transient evoked potentials to providing a “kick” and measuring the response, whereas steady-state responses are similar to “shaking a system gently” and studying the harmonic oscillation that develops. There is some evidence that steady-state and transient VEPs provide complementary data. The frequency of stimulation in the steady-state VEP can be varied to distinguish responses from the magnocellular (40 to 50 Hz) and parvocellular systems (15 to 20 Hz) in the macaque but this has not been demonstrated to have clinical utility in humans. Interestingly, the observed frequency at which patients with MS appear to have the most profound dysfunction with steady-state VEPs is similar to the frequency at which critical flicker fusion deficits are observed. However, the steady-state VEP is used primarily in evaluating subjects who cannot communicate or otherwise participate in examinations for the assessment of acuity (infants and small children); other applications are restricted largely to the research laboratory at this time.
Interpretation of the Transient Pattern-Reversal VEP
Peak response for the transient full-field pattern-reversal VEP occurs approximately 50 to 250 msec after the stimulus. By convention in clinical neurophysiology (as opposed to ophthalmology) laboratories, the recording arrangements for VEPs are such that electrical potentials that lead to an upward deflection are termed negative, while those with a downward slope are termed positive. There are two primary features to each deflection that can be described: (1) the time elapsed since the stimulus (latency) and (2) the magnitude of deflection from the baseline (amplitude). Normal ranges used for references are dependent on the size, luminance, contrast, and temporal characteristics of the stimulus (see next section). Varying other features of the stimulus, such as the color or shape, have less well-characterized influences on the response but also result in differences.
By convention, most recordings are evaluated using the “Queen Square” montage, which includes a midoccipital electrode placed 5 cm above the inion, referenced to a midfrontal electrode placed 12 cm above the nasion (MO–MF). To complete the montage, leads usually are also placed 5 cm to the left (LO) and right (RO) of the MO lead. This placement is obviously very similar to the Fz, Oz, O1, and O2 lead placement from the International 10–20 system and this array can be used as an alternative.
In most individuals, the first response of the full-field pattern-reversal VEP recorded midoccipitally is a negative deflection termed the N75. However, given lack of consistency of both the presence and latency of the N75, by convention full-field VEPs usually are assessed by evaluating the first major positive deflection that occurs at around 100 msec and is therefore designated the P100 component ( Fig. 22-3 ). This positive deflection does not necessarily have the largest amplitude, nor is it always the earliest positive response that is seen, but it is the most reproducible. Following the P100, the next negative deflection is referred to as the N145 (see Fig. 22-3 ). In patients with hemianopia, stimulation of the affected hemifield leads to absent responses or marked variation in the normal amplitude distribution in the contralateral hemifield (LO–MF or RO–MF), but a good response is seen ipsilateral to the “blind” hemifield ( Fig. 22-4, A and B ). However, this method is unreliable compared to dipole source localization, and the ease of obtaining standard automated perimetry has relegated this finding to one of academic interest.
The exact cortical source of the full-field VEP has not been determined unequivocally. It has been suggested that the N75 reflects input from the dorsal LGN to the striate cortex (via the optic radiations), whereas the P100 may reflect a secondary inhibitory response at V1 or excitatory outflow to the accessory visual cortical areas (V2 to V5). There have been several reports of VEPs persisting even in cases of bilateral destruction of the primary visual cortex, including in 4 of 19 patients with complete cortical blindness.
Latency of the VEP generally is defined as the time from the stimulus to a prespecified feature of the record, typically the peak of the component of interest. Approaches include measurement of latency to the point of the maximal positive (or negative) deflection but this has the drawback that flat or noisy components may be challenging to characterize. An alternative approach is to measure the latency to peak by interpolation—in the case of the P100, a straight line is drawn from the downward slope of the N75 and the upward slope of the N145, and the P100 latency is measured at the point of their intersection. The delineation of abnormal latency for the P100 on full-field VEP is laboratory specific. It is of critical importance that the same methods used for establishment of laboratory references are employed at the time of clinical analysis. Any deviation from a standard approach should be explained clearly in a laboratory report.
Latency delay of the full-field VEP is often interpreted as evidence of demyelinating injury to the visual pathway ( Fig. 22-5 ). Abnormality of latency is defined routinely as a value exceeding the mean by more than 2.5 standard deviations. Assuming a gaussian distribution for the reference population, this means that 99.4 percent of normal individuals should fall within this reference range. Given this stringent requirement for abnormality, it is important to note that some individuals with injury to the visual pathway still may be classified as normal. As a consequence, most investigators also evaluate interocular latency differences to improve sensitivity. The optimal cut-off for interocular latency should be defined within an individual clinical laboratory but standard values range from 6 to 10 msec. Consideration must be given to anterior segment (cataract or other ocular media opacities) or outer retinal diseases (such as diabetic eye disease or myopic degeneration) before changes are attributed to optic nerve disease. This is an additional strength to the combined approach, described later.
P100 amplitude is highly variable and, as a consequence, delineation of amplitude abnormality in the full-field VEP is not based on a simple numerical value. There is a high degree of interindividual variability in amplitude on pattern-reversal VEPs in healthy subjects, and the range of observed values is not subject to a gaussian distribution, making it difficult to establish normal values. Furthermore, there may be interocular differences in amplitude of up to 200 percent and repeated VEPs in the same individual may show variability in amplitude of a similar extent as well, depending on the subject’s state of arousal and other patient- and condition-specific factors. Accordingly, caution should be exercised in characterizing a recording as abnormal based on amplitude criteria alone.
Technical Issues for Recording VEPs
As with all clinical electrophysiologic recordings, careful electrode placement with attention to impedance reduction is critical to generating reliable and useful recordings. Because of responses that occur around 60 Hz, bandpass filters at this frequency are not recommended, so ambient electrical noise must be kept to a minimum.
Luminance and Contrast
Luminance is defined as the intensity of light from the visible spectrum per unit area traveling in a given direction (usually expressed in candelas/meter 2 [cd/m 2 ]). Ambient lighting conditions (and preconditioning) can have a significant effect on recordings, and maintenance of luminance conditions is critical for undertaking a reliable study. The luminance of both the stimulus and the ambient conditions should be monitored. The necessary conditions for appropriate ERG recordings are set by the type of study being performed (described earlier). VEPs preferably are recorded under ambient photopic conditions in a standard, normally illuminated room. Patients and normal subjects (laboratory references) should be recorded under the same lighting conditions. The mean luminance at the center of the field is recommended to be at least 50 cd/m 2 but optimally should approximate background luminance. The luminance of the display is also an important factor influencing the VEP, and regular calibration of the system or the use of self-calibrating units is recommended.
Contrast is the luminance difference between two adjacent elements in the visual scene and is defined by the following equation:
C = L max − L min / L max + L min