Key Points of Challenges
- To find cause and effect relationships in neuro-ophthalmic disorders between the retina, the visual systems, and the entire brain.
- To assess and correlate the anatomical and functional similarities and differences between the eye and the brain in health and sickness
- To determine if a finding is clinically significant or incidental
Introduction
‘We can only see what we expect to see.’ This saying highlights the fundamental challenge that we face in exploring the function of the mind while also confined by the limitations of the mind, through its major sensory input, vision. The importance and challenges of this exploration is highlighted by the concerted efforts of scientists around the world and huge financial investments like the Human Connectome Project (Fields) and Allen Brain Atlas (Gilbert).
It is of utmost importance to be mindful of the major technical and clinical limitations of optical coherence tomography (OCT) and to use OCT-based imaging judiciously, correlating it continuously with the clinical findings. Therefore, it is helpful to consider its challenges by comparing it to the most frequently used neuro-imaging method, magnetic resonance imaging (MRI).
As future technological advances translate from bench to bedside, it is critical to bear in mind the limitations of these tools.
Challenges of Oct as a Neuro-Imaging Tool Using the MRI Simile
Challenges of optical coherence tomography-based imaging can be divided into technical and clinical (Table 1). The technical aspect is defined by the laws of physics and this has been reviewed by several excellent publications (Chen-Kardon, Coria, Watson, Lang x2). Therefore, this will only be reviewed briefly here and the emphasis will be placed on the clinical scenarios where expectations from OCT-based imaging are reasonable but still pending.
The first challenge is to become more familiar with the wide spectrum of pre-disease neuro-retinal findings that potentially affect OCT measurements after disease onset, so as not to label a “normal variant” as abnormal. Secondly, to understand the relevance of eye structural and functional findings in both health and disease of the brain. Because understanding the normal spectrum is a prerequisite to disentangle disease states, we must first better define the former in order to tackle the latter.
Presently, there are several unknown physiological variables that limit understanding cause and effect relationships between retinal measurements and brain states. For example, inability to determine the translaminar gradient across the lamina cribrosa due to the invasiveness of intracranial pressure (ICP) measurements results in uncertainty of its direct and indirect effect on OCT measurements (Lee).
Recently, promising MRI-based techniques in patients with idiopathic intracranial hypertension (IIH) have been described for indirect measurement of ICP (Saindane, Hu). Additionally, another mechanical factor that likely affects retinal nerve fiber layer (RNFL), and perhaps ganglion cell complex (GCC) thickness, is repetitive traction at the optic nerve head during eye movements, which has been implicated to cause peripapillary temporal choroidal atrophy and possibly normal tension glaucoma (Suh).
Furthermore, there is lack of information on anatomical variables due to great inter-individual variability in the number of retinal ganglion cells, which is estimated to vary between 0.7 to 1.5 million in the human retina (Kolb). Moreover, there could be structure and function discrepancy in the spectrum of normalcy, exemplified by the so-called normal neuroanatomical variants or incidentalomas on MRI brain scan, such as pineal cysts or enlarged cisterna magna that are often clinically insignificant (Ramji).
Therefore, in the era of personalized, so-called precision medicine, ophthalmo-anatomical variants must be defined. Although, some findings that were previously considered to be incidental on MRI brain, such as so-called Virchow-Robin or enlarged perivascular spaces, have later been implicated as pathological and possibly indicative of mild cognitive impairment (Niazi). Therefore, the most important challenge is to determine if a finding is clinically significant (contributory or relevant to the present clinical complaint and examination findings) or incidental. This is exemplified in considering patients who have an empty sella on MRI brain.
An empty sella can represent a secondary finding, an epiphenomenon (Saindane), as in the case of IIH or it can be a normal variant, as in the case of patients with normal ICP. Ultimately, this neuro-imaging finding must be interpreted within the clinical context, such as cases of normal opening pressure on lumbar puncture in patients with typical phenotypes and clinical history for the disease and who also have optic nerve head (ONH) swelling on fundus examination. Thus, in this example, the neuro-imaging information adds to the diagnosis of probable IIH (see Chapter 7: OCT and Optic disc edema).
Another frequently encountered example of this non-specificity of neuro-imaging on MRI brain is the so-called non-specific white matter disease (WMD) (Mohamed Habes). Often, this finding does not correlate with the patient’s symptoms but rather represents a wide differential diagnosis that ranges in severity from the more “severe” demyelinating or microvascular etiologies to the more ‘benign’ etiologies such as migraine.
Therefore, having the knowledge of patient specific symptoms and physical examination findings in combination with multimodal imaging offers the best available option to address the underlying pathophysiology. Analogy between OCT and MRI, particularly in the use of central nervous system (CNS) disorders is helpful to address the need for OCT as a neuroimaging tool. For example, perhaps the microvascular rarefaction seen on OCTA in several neurodegenerative disorders (Wangh, Lanzillo, Spain) is the retinal correlate of brain WMD.
In determining clinical significance, OCT findings are confounded by intra-individual (inter-eye) structural variability in the number of ganglion cells and their axons that make up the optic nerve and their correlation to the approximately 10,000 times more brain neurons (Herculano-Houzel S). In addition, the optic nerve is not only made of ganglion cell axons but are outnumbered by glial cells, similar to the brain where it is estimated that the ratio of glial cells to neurons is 10 to 1 (Fields).
In the brain, this enormous compaction of neurons occurs in the form of convoluted cortex of gray matter (thickness around 3 mm, total surface area 1.2-2.6 m2) in the cerebral hemispheres that is different from the eye (thickness around 0.5 mm, total surface area of the human retina is 1,094 mm2) (Kolb), where such compaction is not required resulting in the neuroretina having a relatively flat surface.
Also, recent discovery of eye lymphatic (Yucel) and brain glymphatic-lymphatic systems (Louveau) needs correlation and potentially further exploration by OCT-based imaging, as these systems likely play a part in the pathogenesis of degenerative disorders of the eye and brain by their waste product removal function.
Multimodal Imaging
One way to partially overcome some of these shortcomings of OCT-based imaging is to follow a multimodal approach (Saidha, Kuo), with the aim of subtracting complementary information about the clinical problem. However, the imaging-based information will need to be integrated with neuro-ophthalmic history and physical examination findings in order to decide whether it is relevant to the presenting complaint (Chapter 5.2.3).
While deciphering the eye and brain connections (structure) and visual perception (function), we rely on a growing number of sophisticated imaging technologies such as adaptive optics (Zwillinger, Meixner), computer tomography (Szatmary-orbital), fundus imaging (including autofluorescence, infrared reflectance, hyperspectral cameras) (More), magnetic resonance imaging (Neuroimaging of the visual system-Continuum-Szatmary), OCT (Mokbul, Oborwahrenbrock), optogenetics (Roska), calcium imaging with the utilization of 3D-photon imaging of neuronal and dendritic spine assemblies of the visual cortex (Hillier), scanning laser ophthalmoscopy (Mallery: OCT-SLO), ultrasonography, and/or a combination of these tools. All of these methods rely on the function of our own visual system; therefore, they are subject to its limitations.
However, when we combine ophthalmic and neuro-imaging structural with functional tools, such as electrophysiology (electroretinography, pupillometry, visual evoked potentials) (Meltzer), functional magnetic resonance imaging (fMRI), OCTA, and visual field testing, there is a higher likelihood that the limitations of each of these methods are overcome, or at least decreased. Therefore, both the neurology and the ophthalmology community has been using some combination of complementary technologies.
A frequent but difficult neuro-ophthalmic complaint is monocular transient visual loss (Biousse), when evaluation of orbital vascular supply is of interest but is a particularly difficult area to assess.
The use of OCT and OCTA is limited in this area secondary to its poor penetrance. Ultrasonography has better penetrance but limited field of view. Traditional magnetic resonance angiography (MRA), like 2D-and 3D time-of-flight (TOF) sequences are limited by suboptimal resolution and the fact that orbital adipose tissue, in which the vasculature is buried, has short T1 and long T2 relaxation times and therefore appears bright (hyperintense) on both T1- and T2 weighted sequences (MRA orbit).
Moreover, the usually applied fat suppression methods have significant artifacts that make them suboptimal for the assessment of small vessels such as the central retinal and medial and lateral posterior ciliary arteries that supply the retina and optic nerve, as well as having great interindividual variability (see Homology chapter).
Therefore, to achieve maximal MRA contrast between arteries and fat in retrobulbar space, the ideal way is to completely suppress fat signal while leaving blood signal high. Moreover, high resolution imaging is needed in order to reduce partial volume effects on small branches of the vessels (Graessl). MRA, even with the above limitations, complements OCTA in the visualization of the vascular network of the eye, especially with the use of contrast agents that improve vessel signals.
In daily clinical practice, there is also lack of sensitivity of these methods compared to the ideal research environment due to subject-based technical limitations, such as degraded quality due to inability to cooperate resulting in poor fixation and motion-induced artifacts, which is a major issue in patients with neurodegenerative disorders and visual fixation instability due to underlying disease such as optic neuropathy (Mallery).
Also, a major drawback of OCT-based neuroimaging is lack of specificity for not only a particular disorder but even for underlying pathophysiology such as compressive–infiltrative, hereditary, degenerative, inflammatory-infectious, traumatic, and vascular etiologies (Szatmary-eye pain, Chen-OCTA-chronic optic neuropathy).
Other major clinical disadvantages of OCT are that it is likely to miss early changes (Kupersmith x2), thereby limiting its role in determining the time course of disease, and that it provides non-localizing information (Symptoms and signs suggestive of orbital disease: localizing versus non-localizing (Al-Louzi: trans-synaptic degeneration, Vislisel, Zehnder).
Therefore, correlation with other structural and functional studies, such as visual field testing, is needed to localize and determine etiology (Blanch, Micieli). When evaluating complex neurodegenerative disorders, it is important to segment the entire retina and not only the neuroretina, as outer retinal changes have already been described (Al-Louzi pg 10). Also, subtracting additional information from the available datasets, such as shape (Wang J-Kai), texture, and other voxel-based morphometry (Bhavna pg 9,) is necessary due to the limited information thickness provides.
Oct Challenges in the Brain
In regards to the cerebral cortex, OCT-based imaging has the capability to visualize cortical neurons and microvasculature, especially of the deeper layers that are presently inaccessible by existing methods such as MRA and functional magnetic resonance imaging (fMRI). Cortical neurons are assembled into functional units consisting of horizontal layers and vertical columns, an example of such architecture is the interdigitated column system of the primary visual cortex (see Homology section). The superficial cortical layers are supplied by the so-called microvascular lobules (Adams).
The superficial cortical layers are supplied by the pial circulation in a layer-specific manner, so-called microvascular lobules (neurovascular units) consisting of a central larger venule surrounded by penetrating smaller arterioles (Adams). These microvascular lobules have an anastomotic system regulated by vascular sphincters by which local cortical perfusion is controlled. Interesting is the fact that microvascular lobules bear similar periodicity but do not necessarily coincide with the so-called cytochrome oxidase (CO) patches that define the metabolic activity of neurons (Adams).
An exception to this rule is the behavior of the granular layer of the sensory cortex, layer 4, which has the highest metabolic rate and richest microvascular supply, as judged by the red blood cell endogenous peroxidase activity. Such a robust activity is readily visible on blood-oxygen-level-dependent (BOLD) fMRI (Tian). Horton et al. raised the question as to how the so-called cytochrome oxidase (CO) patches that define the metabolic activity of neurons are able to maintain a 49% higher mean firing rate without having a richer microcirculation.
The answer is not yet known but it is hypothesized that it is secondary to faster blood flow and/or a higher oxygen extraction ratio compared to other areas. Because of this discrepancy, fMRI findings may not necessarily correlate with increased neuronal activity. Perhaps this question is best addressed in humans through multimodal manners, such as by combining in vivo non-invasive fMRI, 3D-photon microscopy, and even OCT and OCTA (Chen, Merkle).
A major advantage of OCT and OCTA over MRI is lower cost and therefore easier accessibility, and as such would be an ideal screening tool in neurodegenerative disorders. However, presently, high cost is still a limiting factor to academic research centers in some of the newer methods, such as OCTA.
Final Remarks and Take-Home Messages
In summary, OCT has already proven to offer invaluable information in several retinal and ONH disorders that is not available by any other method to date and has improved patient management and outcomes (Kupersmith: OCT-NORDIC). However, OCT-based imaging in the realm of complex brain anatomy, physiology, and pathophysiology, presently offers non-specific, non-localizing, and sometimes contradictory information emphasizing its limitations by only offering one piece of the puzzle.
In the era of precision medicine, future well-designed, prospective, controlled, and longitudinal studies with homogeneous populations are needed in both health and disease in order to be able to answer patient specific questions while filling out the current gaps of knowledge by utilizing high quality standardized scanning protocols and reporting (Malmqvist, Cruz-Herrasz). The future is bright as pieces of the brain-eye connection puzzle are falling into place through technologies such as OCT-based neuro-imaging.
Keeping in mind though that the usefulness of these tools in clinical practice require familiarity with detailed anatomy and physiology and pathophysiology of the entire visual system. Moreover, the acquired high quality OCT data to be clinically useful (establish cause and effect relationship, provide prognostic value, follow treatment effect) needs to be scrutinized and correlated with clinical findings (JJ Chen -cause and prognosis, Feucht, Kupersmith, Button- treatment effect).
Nota bene and future directions:
- The manufacturer’s normal database needs to be better defined and routinely corrected for refractive errors and axial length
- Individual eye and brain layers should not be looked at in isolation
- To assess the past, present and predict the future of eye and brain states by OCT, longitudinal studies and correlation with histopathology and in vivo multimodal imaging is needed
- Disease process specific and localizing neuro-i
