In-Vivo Confocal Evaluation of Corneal Nerves in Patients with Diabetic Neuropathy BY

Purpose: The objective of the first part of this study was to determine if the corneal nerve parameters of mean nerve count, total nerve length, percent primary nerve character, and tortuosity of nerves in the subbasal nerve plexus varied amongst three groups of patients: those with severe diabetic neuropathy, those with diabetes and mild to no clinical evidence of neuropathy, and a healthy control group. The second part of this study focused on correlating nerve tortuosity values with mean nerve count, total nerve length, and primary nerve character. Methods: Confocal data for patients with severe diabetic neuropathy (n = 7), diabetes and mild or no neuropathy (n = 15), and healthy controls (n = 9) were collected with the Heidelberg Retina Tomograph confocal microscope and analyzed using NeuronJ and CCMetric. Image analysis included determining the mean nerve count, total nerve length, percentage of primary nerves, and tortuosity values for each scan. Analysis of variance, Tukey’s post-hoc test, and the Pearson Coefficient of Determination were used as tools for statistical analysis. Results: There was a statistically significant difference (p < 0.05) in the mean nerve count between the healthy control group and the severe neuropathy group. The healthy control group expressed a strong linear relationships when comparing tortuosity with mean nerve count (R = 0.77) and mean nerve length (R = 0.64), while the severe neuropathy group expressed its lowest values of mean nerve count and total nerve length at the highest and lowest tortuosity values. Conclusions: The results support the hypothesis that corneal nerves, as imaged by confocal microscopy, correlate to, and can be used as diagnostic markers for, systemic diabetic neuropathy. A larger patient sample size will increase confidence and reliability of the data, and an automated MATLAB script currently being developed will speed up the analysis process. Biophysics Undergraduate Honors Thesis Raval 3 Introduction Over 25 million people in the United States suffer from diabetes, and about 60-70% of diabetics express mild to severe forms of neuropathy [1]. In the past, the only method of directly examining the peripheral nerves was to conduct skin or nerve biopsies, which were uncomfortable and invasive. The benefit of using ophthalmic markers for early detection of diabetic peripheral neuropathy is that the procedure is non-invasive, cost-efficient, and clinically accessible. The first objective of this research is to compare the characteristics of corneal nerves – mean count, mean length, primary nerve character, and tortuosity – in patients with diabetes and severe neuropathy, diabetes and mild to no clinical evidence of neuropathy, and a healthy control group. The second objective is to determine how nerve tortuosity, a parameter that has not been studied extensively in the field, correlates with nerve count, length, and character. The ultimate goal of this study is to utilize the results as a quantitative diagnostic that can indicate the presence of diabetic neuropathy in patients before full expression of clinical symptoms of the disease arises. The cornea can be easily examined in an outpatient setting using corneal confocal microscopy, an in-vivo procedure that involves shining a pinpoint laser into the patient’s eye, progressively scanning through the layers of the cornea, and transmitting the scans to a digital imaging program to quantify the data. The presence of significant differences between groups of patients could hint at a pathophysiological mechanism relating the systemic condition of diabetes to the status of the corneal nerves. If significant correlations between diabetic neuropathy and corneal nerve parameters can be quantified, a model for the quantitative relationship between the severity of diabetic neuropathy and corneal nerve parameters can be established. This diagnostic model Biophysics Undergraduate Honors Thesis Raval 4 could be utilized during routine doctor appointments or at regular eye exams to non-invasively test for early detection of diabetic neuropathy. It is hypothesized that corneal nerves can be used as effective diagnostic markers for diabetic peripheral neuropathy, and that increased severity of neuropathy will correlate with lower nerve counts and lower total nerve lengths due to greater nerve damage in the cornea. It is also hypothesized that increased neuropathy will correlate with lower tortuosity values and a higher percentage of primary nerve character relative to secondary nerve character, due to the neuropathic destruction of the more tortuous secondary nerve branches that disrupts nerve-tonerve communication. Finally, it is hypothesized that strong correlations will exist between tortuosity, nerve count, and nerve length in healthy patients but will not exist in neuropathy patients, as increased nerve damage may reduce the ability of the nerves to sense their surroundings and adapt to the presence of neighboring nerves by becoming more tortuous and compact. Biophysics Undergraduate Honors Thesis Raval 5 Background The cornea is the outermost layer of the front of the eye [2]. Completely transparent, its purpose is to refract light onto the retina in the posterior chamber of the eye. The human cornea is comprised of five major layers: the epithelium, Bowman’s layer, the stroma, Descemet’s membrane, and the endothelium. The epithelium is located on the outside of the cornea and makes up approximately 10% of the cornea’s overall thickness [2]. The main functions of the epithelium are to protect the inner layers of the cornea from external debris such as dust particles and bacteria, and to provide a surface that is capable of absorbing the oxygen and cellular nutrients of tears and transmitting these nutrients to the inner layers. The bottom layer of the epithelium, the basement membrane, serves as the anchor for epithelial cells. Directly beneath the basement membrane lies Bowman’s layer, an 8-14 μm thick, acellular layer composed of collagen [3]. The exact function of Bowman’s layer has yet to be discovered, but it has been hypothesized that proper functioning of the human cornea is not dependent on this layer, as patients who have undergone excimer laser photorefractive keratectomy to remove Bowman’s layer have not experienced any adverse effects [3,4]. The third layer, the stroma, accounts for 90% of the entire thickness of the cornea, and it is mainly comprised of water (~78%) and collagen (~16%) [2]. Stromal collagen is especially important for the maintenance of corneal structure, strength, and elasticity [2]. Descemet’s membrane, a thin sheet of collagen fibers, lies directly behind the posterior stroma and serves as a protection against injury and infection. The innermost layer of the cornea, the endothelium, is responsible for pumping out the excess fluid that leaks into the stroma from inside the eye, thereby safeguarding against edema and potential blindness [2]. Biophysics Undergraduate Honors Thesis Raval 6 Between the basal epithelium and Bowman’s layer lies the subbasal, or subepithelial, nerve plexus, the corneal region responsible for epithelial innervation [5]. The subepithelial region of the human cornea contains a radiating pattern of bundles of nerve fibers that originate from stromal nerves and pass through small openings in Bowman’s layer, converging in spiral-like patterns about 1-2 mm away from the corneal apex, the outermost point of the cornea [6]. The terminal endings of these nerve fibers reach the epithelial layer but are too small to be seen with contemporary confocal technology [6]. The nerve plexus consists of highly dense sensory and autonomic nerve networks, the former of which constitute the ophthalmic branch of the Figure 1: The five primary layers of the human cornea: epithelium, Bowman’s membrane, stroma, Descemet’s membrane, and endothelium [2]. Biophysics Undergraduate Honors Thesis Raval 7 trigeminal nerve [7]. Mechanical, thermal and chemical stimulation of these sensory nerves is the primary cause of pain in the human eye [7]. The autonomic network contains both sympathetic fibers, which are derived from the superior cervical ganglion, and parasympathetic fibers, which are derived from the ciliary ganglion [7]. The high density of nerve endings in the cornea is derived from the posterior ciliary nerves, and the cornea is approximately 100 times more sensitive than the neighboring conjunctiva [8]. If these nerve endings are destroyed, the cells of the epithelial layer will swell and produce basal lamina, which inhibits mitosis and directly causes apoptosis of epithelial cells [8]. Figure 2: Subbasal nerve plexus, immediately anterior to Bowman’s layer, as imaged by a confocal microscope [6]. The hyperreflectivity of Bowman’s layer allows for easy characterization and visualization of the nerve plexus [9]. While nerves may also be found in the stromal layer, they are often much thicker than nerves in the subbasal nerve plexus, and can be differentiated from subbasal nerves as such [9]: Biophysics Undergraduate Honors Thesis Raval 8 Figure 3: Stromal nerves are thicker than their subbasal counterparts [9] An in-vivo study of corneal nerve morphology using a Confoscan slit-scanning confocal microscope showed that corneal nerves originate in the stroma with thick, linear trunks that extend both laterally throughout the stroma and towards the anterior part of the stroma [10]. Before these trunks reach the subbasal nerve plexus, they begin to thin out and form networks that extend towards, and penetrate through, Bowman’s layer. Diabetic neuropathy is nerve damage characteristic to patients with diabetes, as high blood sugar can damage nerves throughout the body [11]. This nerve damage is mainly the result of a combination of several systemic-based causes, including neurovascular, metabolic, and autoimmune factors. Prolonged exposure to high levels of blood glucose can damage the blood vessels that are responsible for carrying nutrients and oxygen throughout the body as well as cause inflammation in nerves [12]. The most common regions of nerve damage are in the nerves of the lower extremities, but studies have shown that this peripheral degeneration correlates with certain ophthalmic markers, such as the morphological degradation and reduced sensitivity of Biophysics Undergraduate Honors Thesis Raval 9 corneal nerves, thinning of retinal nerve fibers, and peripheral field loss [13]. This project focuses on the first marker, corneal nerve degradation, and seeks to correlate key parameters of this effect, such as nerve count, nerve length, and tortuosity, with severity of diabetic neuropathy. In-vivo corneal confocal microscopy is a technique that can be used to obtain a high-resolution scan of the human cornea. While the slit lamp biomicroscope and ophthalmoscope are also used by clinicians to observe the cornea, they are unable to do so with high clarity, detail, or magnification, and so physiological studies of the cornea at the microscopic level have been limited to in-vitro observations [14]. Light biomicroscopy of corneal layers results in poorly resolved images, because light is reflected from structures above and below the layer of interest [9]. For example, if a biomicroscope was used to examine a corneal lesion in the anterior stromal layer, light reflected from the surrounding layers – the epithelium, the tear film, and even the endothelium – would cause light contamination and defocus the target layer [9]. The corneal confocal microscope works on the principle of point illumination, whereby a light source and camera can be used simultaneously to illuminate and image a specific region of tissue in coplanar fashion [4]. The resolution of the image captured by a confocal microscope is therefore much higher than that of a traditional fluorescence microscope, which shines light evenly over an entire region rather than focusing it on a particular point. Since confocal microscopy only deals with one point of a sample at a time, the field of view from a single image will be negligibly small. To account for this potential problem, the confocal microscope “scans” over a small region of tissue by illuminating it with thousands of tiny points of light, so that a high-resolution, substantial field-of-view image of the region can be constructed via amalgamation of all of the Biophysics Undergraduate Honors Thesis Raval 10 points in the plane. The confocal microscope can sequentially raster across each plane through the z-axis, thus producing a series of cross-sectional images of the sample: The United States Food and Drug Administration (FDA) has approved three types of confocal microscopes for use in the clinical setting: the tandem scanning confocal microscope (TSCM), the scanning-slit confocal microscope (SSCM), and the laser scanning confocal microscope (LSCM) [16]. In the TSCM, the illumination and detection beams do not follow the same path; rather, they travel tandem to one another through separate but identical apertures [17]. The main advantages of the TSCM include its ability to image peripheral corneal lesions, capture real-time images at up to 30 frames per second, and scan through the z-axis with high resolution [16, 18]. It does this via a Nipkow disc, which rotates at about 900 revolutions per minute and has 64,000 pinholes arranged in spiral fashion [18]. The SSCM, which works by eliminating out-of-focus light from the scan via an array detector, is the most popular clinical confocal microscope due to its user-friendly image analysis software [16]. Wide variations in subbasal nerve density have Figure 4: A cross-sectional view of the human cornea as viewed by a corneal confocal microscope. The epithelium (scans 1-4), nerve plexus (scan 5), anterior and posterior stroma (scans 6 and 7), and endothelium (scan 8) can be seen as the microscope scans through the cornea [15]. Biophysics Undergraduate Honors Thesis Raval 11 been reported by several studies depending on the type of confocal microscope used. The TSCM and the SSCM have measured nerve densities of healthy patients in the range from 5.5 μm/mm to 11.1 μm/mm, while the LSCM has measured densities up to 21.7 μm/mm [6]. The reason for these discrepancies lies in the differences in field brightness and image contrast amongst microscopes; the LSCM and SSCM tend to have greater brightness and contrast than the TSCM [6]. In addition, due to the high-resolution capabilities of the LSCM, it is able to detect smallercaliber nerves, which is why its density estimates can be up to twice those of the SSCM [6]. Clear nerve detection is more difficult at the edges of images due to a decrease in field brightness [6], but error can be minimized if measurements are made in consistent fashion and the regions of interest for each image have the same areas. The most recent confocal microscope and the one that has been used for this study is the LSCM (Heidelberg Retina Tomograph II Rostock Corneal Module, HRTII) [4]. The main advantage of using the HRTII over more traditional microscopes like the Confoscan series is that it is able to produce a series of extremely high-resolution, thin-layer images of the cornea. HRTII generates high-contrast images at a wavelength of 670nm [6, 15]. Figure 5 below shows how a beam of laser light is directed through an initial aperture and a partial mirror, passes through a lens, and is directed through the focal plane of the lens onto a small section of the patient’s cornea. This light is then reflected off the cornea and travels back through the lens and through the partial mirror. Another aperture with a small pinhole is placed adjacent to the mirror such that only the light that is coplanar with the focal plane of the lens is able to pass through to the detector. This is specifically what makes confocal microscopy unique; the additional pinhole placed within the focal plane before the detector eliminates any out-of-plane light rays from flooding the detector. Biophysics Undergraduate Honors Thesis Raval 12 Figure 5: Basic layout of corneal confocal microscopy [6]. Left: the light rays at the focal point of the lens reflect off the cornea and are transmitted through the conjugate aperture to the detector. Right: the light rays that are not reflected off the object plane (dotted lines) are not transmitted through the conjugate aperture to the detector. The left side of Figure 5 shows how light that is reflected at the focal point of the objective lens is “co-focused” and confocal to the detector. The right side of Figure 5 depicts the light rays that are out of the plane of the focal point of the objective lens (slightly anterior or posterior to the focal point) as solid lines that are defocused at the detector [6]. In other words, these rays are not coplanar to the lens and so the light rays reflected off this plane will not be transmitted through the pinhole to the detector. The lens can therefore be adjusted such that the cornea of the patient is located precisely at its focal point. The brightness of confocal images is dependent on several factors, such as the intensity of the laser, the radius of the confocal aperture, and the light scatter and specular reflectivity [6]. Small particles in the epithelial and stromal layers contribute to light scatter, while sudden changes in refractive index, which occur at the boundaries of the epithelial and endothelial layers, contribute to specular reflection [6]. Statistical analysis of the data obtained from NeuronJ and CCMetrics will consist of a One-Way Analysis of Variance (ANOVA), Tukey’s post-hoc test, and the Pearson Coefficient of Biophysics Undergraduate Honors Thesis Raval 13 Determination. The first two tools, ANOVA and Tukey’s post-hoc test, will be used on the data obtained from the first part of this study, where the corneal nerve parameters of mean nerve count, total nerve length, tortuosity, and percentage of primary nerve character, are compared across patient groups. The third tool, the Pearson Coefficient of Determination (R value), will be used during the multivariable portion of this study, where tortuosity is plotted against mean nerve count, total nerve length, and percentage of primary nerve character for each patient group. When comparing two samples, the t-test can be successfully utilized between the groups. However, the problem with using the t-test when more than two groups are involved, as is the case in this study, is that false positives tend to arise more often, and an error of 1 – (0.95) results, where n is the number of separate t-tests being conducted. One-way ANOVA corrects this error by comparing all three means simultaneously instead of pair-wise, thereby eliminating the added type I error probability that arises from performing multiple t-tests. ANOVA works by generating an F statistic, or F value, which is defined as the variance between sample means divided by the combined variance within each sample as shown in Equation 1 below [20]. The p-value can be determined directly from the result of the F test via a correspondence table; the program Windows OriginLab, automatically converts this F value to its corresponding p-value.