**2.1 Imaging the cone mosaic in diabetic retinopathy**

Biological systems are characterized by symmetrical spatial arrangements as in hexagonal systems. These structures can have equal angles and sides. Even when they are not clearly delimited [52], Voronoi diagrams can generate the limits between elements, taking the starting points as generators. It is known that the regular hexagon is the maximum-sided polygon that could be used to cover a plane without overlaps. The immediate advantage of this organization is the maximization of number of neighbors of each element and the consequent optimization of cell signaling, resolution, and isotropy (photoreceptors, retinal neurons).

#### *2.1.1 Cone parameters*

#### *2.1.1.1 Photoreceptor density*

The density of photoreceptors is usually assessed by dividing the number of detected cones by the analyzed area. This method implies a normal distribution within the analyzed area and will underestimate the density if this area overlaps over regions for which no data are available. These limitations motivate the need for Voronoi local density analysis [53]. The density of photoreceptors can be expressed in metric (cones/mm2 ) or angular (cones/degree) units. The value of the antero-posterior axis can influence the values expressed in metric units. At a certain eccentricity, the distance from the fovea may be different depending on the antero-posterior axis. For example, an eccentricity of 2° corresponds to a distance between 0.53 and 0.64 mm from the fovea, given an axis of 22–26 mm. As the axis gets longer, the retinal area increases and the cones density gets lower when expressed in metric units [54].

#### *2.1.1.2 Cone spacing*

This indicator is useful to be used in conjunction with photoreceptor density. It considers only the closest photoreceptor for each cone detected in the region of interest, regardless of the measured distance [17]. This can lead to high values for isolated cells. On the other hand, it has been shown that the distance between photoreceptors is less sensitive to subsampling (compared to cones) and to the correlation with a certain pathology. Thus, photoreceptor spacing overestimates the overall health of the photoreceptor mosaic (e.g., a mosaic with sporadic photoreceptor losses may have normal distances between photoreceptors, but in the presence of an abnormal density). In conclusion, in order to correctly depict a mosaic, all biomarkers should be considered [55].

#### *2.1.1.3 Voronoi diagrams*

Given a finite set of two or more points in an Euclidean plane, we associate all locations of that plane with the nearest element of the set. The result is a tessellation of the plane in a series of regions associated with the elements of the set of points. This tiling is called the Voronoi diagram generated by the set of points, while the regions that make up the Voronoi diagram are called Voronoi polygons [56]. Li and collaborators showed that less than 30% of Voronoi polygons are non-hexagonal in the vicinity of the fovea and that their percentage increases to 50–60% at higher eccentricities [57]. Cones at higher eccentricities are unevenly distributed and protect the visual system from perceiving a distorted signal [58].

AO ophthalmoscopy is an accessible tool for clinicians to visualize the human retina *in vivo* [59, 60]. Cones parameters have been analyzed in healthy, adult population [15, 61, 62] and correlated to several factors (age, gender, refractive error, antero-posterior axis, race, ethnicity). These studies showed a variability of photoreceptor densities in normal population [15, 63], which makes it difficult to detect small variabilities of this parameter in comparative studies. The inclusion of all photoreceptor parameters (distance from the nearest photoreceptor and Voronoi diagrams—with the analysis of their spatial arrangement) in study protocols promises to provide more conclusive results [17]. Nevertheless, the first measurements of cone density come from the postmortem histological analysis of the human retina [64]. It has been shown that in the center of the fovea, there is a density of 199,000 cells/mm2 , which decreases to about 20,000 cells/mm2 to 1 mm from the center of the fovea.

#### *2.1.2 Cone parameters in diabetic retinopathy*

AO ophthalmoscopy has been used to assess the parafoveal cone parameters in diabetic patients (type I or type II diabetes) and controls in various studies (**Figure 3**) [16, 17, 48, 66, 67].

Lombardo et al. [17] studied the differences between cone parameters at 1.5 degrees eccentricity in patients diagnosed with DM I without DR or with nonproliferative DR and in healthy subjects. Their results showed that cone density was higher in the control group as compared with the study group, on both vertical and horizontal meridians. Another study [47, 68] showed that cone density was 10% lower in type I diabetic subjects with no DR than in controls. Moreover, cone densities were also investigated in type II diabetic patients with or without DR and in healthy subjects, at 0 and 2 degrees eccentricity [16]. The results showed a positive correlation of the extent of cone loss on AO imaging with DR severity in

*Adaptive Optics Imaging Technique in Diabetic Retinopathy DOI: http://dx.doi.org/10.5772/intechopen.101266*

**Figure 3.**

*Examples of regions of interest (red square) at 2, 3, and 4 degrees nasally analyzed in a control subject (a) and in a diabetic patient (b). (Reproduced from [65]. Copyright by the ©Romanian Society of Ophthalmology.)*

type II diabetic patients. On the other hand, Tan and collaborators [69] found no cone densities differences at 7 degrees eccentricity between patients with DM I and healthy subjects, probably due to the short duration of diabetes in the study group.

In addition to this, cone spacing was found to significantly increase in diabetic patients, when compared with healthy controls [70]. Results show lessening of hexagonal mosaics of cones (Voronoi 6 tiles) with increased severity of DR.

The horizontal meridian was proved to have a higher average cone density than the vertical one in both controls and type I diabetic patients with no DR [47, 68]. However, the asymmetry was higher in the control group. Another study reported this pattern in normal subjects [15]. This asymmetry, also called "horizontal–vertical anisotropy," could be related to the way we are using our vision. When reading, the horizontal meridian is more in demand than the vertical one. Psychophysical studies have shown a higher contrast sensitivity and spatial resolution of the horizontal meridian as compared with the vertical one at a given eccentricity [71]. Further studies including more subjects are needed to describe the cone parameters in diabetic and age-matched volunteers.

#### **2.2 Imaging the retinal microvasculature in diabetic retinopathy**

Retinal microcirculation is a network of arterioles, capillaries, and venules with diameters that do not exceed 150 μm, whose main function is to assure and regulate an optimal tissue perfusion [72]. Microcirculation alterations, also called microangiopathy, lead to organ damage and clinical events in patients with DM.

A microvascular network is assessed from a structural and functional point of view, taking into consideration both its topology and geometric abnormalities [73].

**Figure 4.**

*Image of the retinal artery of a patient with diabetes mellitus, with visualization of the walls (short arrow) and lumen (long arrow), employing AO detect artery software.*

The retinal vascular branching model is similar to a fractal, having a complex pattern of distribution, where each part has similar features to the main structure. DR is associated with a decreased fractal dimension, which is probably correlated with the shortcoming of the retinal circulation [74]. The network geometry is appreciated using the length and the diameters of the vessels, by calculating different derived parameters.

Vascular-addressing diseases encounter different patterns of vascular remodeling. The inward eutrophic remodeling, seen in stages 1 and 2 of essential hypertension, is characterized by a reduced lumen diameter, with consecutive rearrangement of the surrounding smooth muscle cells, but without a marked growth response [75]. The media-to-lumen ratio is increased due to reduced lumen diameter and external diameter of the wall and due to increased media thickness, with minimum changes in the total amount of wall tissue. All these result in decreased vasodilation potential of the vessel and altered arteriolar distensibility [76, 77]. In contrast, the hypertrophic remodeling, found in diabetes mellitus and in long-standing, severe and secondary hypertension, exhibits a significant growth response with vascular smooth muscle cell hypertrophy or hyperplasia [78, 79]. This leads to the increase of both media-to-lumen ratio and the media cross-sectional area of the vessel [80].

The bedside morphological assessment of human retinal microcirculation is not facile, the gold standard being the media-to-lumen ratio evaluated by wire or pressure micromyography on bioptic samples [81]. Recently, noninvasive techniques for the evaluation of retinal arterioles prove rather good agreement with micromyographic measurements, in particular scanning laser Doppler flowmetry and adaptive optics [81, 82]. Preliminary data suggest that AO has a substantial advantage over Doppler flowmetry, having a better correlation with the gold standard, but invasive procedure [81].

Adaptive optics ophthalmoscopy uses a cutting-edge technique, visualizing the retinal arterioles lumen as a bright band, while the walls correspond to the darker neighboring regions (**Figure 4**). The wall thickness of the blood vessels depends on the vessels' size, with large lumens requiring thicker walls. The AO-related vascular parameters of interest are:


Zaleska-Żmijewska et al. [48] found statistically significant differences between the control group and the prediabetic group with respect to LD and to WLR. Furthermore, the same scientific team demonstrated that WLR, WCSA, and the average WT exhibit significant differences between the control group and type II diabetic patients [70]. An AOSLO study confirms that retinal arterial WT is significantly greater in patients with type 2 diabetes than in controls [83].

When it comes to DM type I, the average capillary LD proved to be significantly narrower in eyes with nonproliferative diabetic retinopathy and type I DM, when compared with healthy subjects (6.27 ± 1.63 μm versus 7.31 ± 1.59 μm, p = 0.002) [50]. When investigating the retinal microcirculation changes in type I and II diabetic patients without retinopathy, among all studied parameters, only WLR was significantly different between the control group and each group of diabetic patients taken individually, while no statistically significant differences were found between the two diabetic groups [47, 68].

New data characterize retinal arterioles according to DR classification, showing that LD, WT, and WLR significantly correlate with the stage of DR [84]. AO was even used to establish the effects of a multinutrient complex on retinal microvasculature in diabetic patients. After 3 months of food supplements administration, WLR, WT, and WCSA had significantly decreased, when compared with initial observations [85].

AO proved its potential to detect retinal microvascular changes in prediabetic subjects and diabetic patients, as well as to reveal differences between the diabetic groups. Providing useful information about the topological and geometrical features of the retinal microvasculature from early onset of diabetic disease, AO has a promising role in the future, providing valuable prognostic, diagnostic, and therapy-related information in diabetic retinopathy.
