Enhanced Depth Imaging Optical Coherence Tomography of Ultrasonographically Flat Choroidal Nevi Demonstrates 5 Distinct Patterns

Abstract

Purpose:

To characterize the enhanced-depth imaging optical coherence tomography (EDI-OCT) features of ultrasonographically flat choroidal nevi.

Design:

Retrospective observational study

Participants:

Patients with ultrasonographically flat choroidal nevi identified over 3-year period.

Methods:

Comprehensive eye examinations and multimodal imaging were performed every 6 −12 months.

Main Outcome Measures:

Lesion thickness and features, EDI-OCT findings

Results:

102 nevi (98 patients) were included. Median follow-up was 9 (range: 0–144) months and 1–8 (mean: 2.1) OCTs were performed. On OCT, nevi were classified into 5 subtypes: 3.2% were “not visible,” 37.2% had “hyperreflectivity confined within normal choroidal thickness,” 16.0% had characteristic anteriorly-bowed hyperreflectivity with “discrete” borders and cascading edges, 14.9% were “flat with posterior bowing,” and 28.7% were “elevated.” 1 nevus, which was of the “elevated” type, demonstrated clinically insignificant growth (<10% in thickness) after 22 months of follow-up.

Conclusions:

There are 5 distinct EDI-OCT patterns of choroidal nevi that appear flat on ultrasonography, and many demonstrated thickness and elevation not measurable on B-scan. Lesion thickness can be measured using EDI-OCT.

INTRODUCTION

Choroidal nevi are the most common benign tumors of the ocular fundus. 6.5% of the adult Caucasian population were found to have choroidal nevi in the Blue Mountains Eye Study.¹ The annual rate of malignant transformation of choroidal nevi in the United States Caucasian population has been estimated to be 1 in 8845.² Factors predictive of growth into melanoma include thickness greater than 2 mm, subretinal fluid (SRF), symptoms, orange pigment, margin near disc, ultrasonographic hollowness, and absence of halo or drusen.³ Uveal melanoma carries a significant risk of morbidity and mortality. Thus, detection and monitoring of choroidal nevi are of paramount importance, as nevi are the precursor lesions to many choroidal melanomas. Beyond the clinical and ultrasonographic features, optical coherence tomography (OCT) has improved the ability of clinicians to detect subtle changes in choroidal lesions, such as the presence of small amounts of subretinal fluid.

Shields et al. demonstrated the utility of time-domain (TD) OCT in imaging the retina overlying choroidal nevi in 120 consecutive patients but recognized that TD-OCT findings were limited to the anterior surface of each nevus without penetration into the mass.⁴ With improved speed, sensitivity, and axial resolution, the advent of spectral domain OCT (SD-OCT) has allowed for better detection of retinal and retinal pigment epithelial (RPE) changes secondary to choroidal tumors. Nonetheless, SD-OCT findings are limited to the anterior aspect of choroidal tumors and unable to fully characterize intrinsic features.⁵

The enhanced-depth imaging (EDI) modification of SD-OCT has allowed for high-resolution cross-sectional imaging of choroidal lesions.⁶ Margolis and Spaide evaluated normal eyes and were able to measure choroidal thickness using EDI-OCT.⁷ A significant advantage of EDI-OCT over traditional SD-OCT in the evaluation of choroidal nevi is in monitoring those lesions that cannot be visualized on ultrasonography. Choroidal lesions that are thin enough to appear ultrasonographically flat are difficult to delineate, measure, and reliably follow with B-scan ultrasonography. Torres et al. demonstrated that small choroidal tumors undetectable on ultrasonography could be objectively measured using EDI-OCT.⁸ Though the features of choroidal nevi have been described previously by Shah et al.,⁹ a study focusing on flat choroidal nevi using EDI-OCT has not been undertaken previously. EDI-OCT allows for more precise delineation of boundaries of flat choroidal nevi when compared to ultrasonography. Being able to reliably characterize and follow choroidal nevi over time is of critical importance to detect high-risk changes early. In so doing, transformation to choroidal melanoma can be detected early and treated early. It has been shown that the risk of metastasis from choroidal melanoma is related to lesion thickness, on a millimeter-by-millimeter basis,¹⁰ and that the risk of radiation retinopathy is directly related to the size of the tumor and its location at the time that it is treated.¹¹ Therefore, developing strategies for the early identification of malignant transformation of nevi may reduce the side effects of treatment and may also save lives. We sought to characterize the EDI-OCT features of ultrasonographically flat choroidal nevi and to determine if EDI-OCT can detect early changes in lesion thickness measurements over time.

METHODS

We performed a retrospective observational study approved by the Vanderbilt University Institutional Review Board, which was designed to identify all patients who had ultrasonographically flat choroidal nevi in one or both eyes evaluated in the Ocular Oncology Clinic, Vanderbilt Eye Institute, Nashville, Tennessee between November 2013 and March 2017. The research was conducted in accordance with the Health Insurance Portability and Accountability Act and adhered to the tenets of the Declaration of Helsinki. A standardized protocol was followed wherein patients had a comprehensive eye examination with dilated fundoscopy as well as fundus photography, fundus autofluorescence (FAF), B-scan ultrasonography, and EDI SD-OCT using Heidelberg Spectralis (Heidelberg Engineering, Heidelberg, Germany) every 6–12 months. EDI was performed using a technique comparable to that described by Spaide et al.⁶ All of the procedures were performed as part of standard of care. Informed consent was obtained from all patients for examination and testing. Patients with choroidal nevi of thickness > 1 mm on B-scan ultrasonography were excluded. Patients with pigmented choroidal lesions suggestive of choroidal melanocytosis and not typical of nevi were excluded. Patients with choroidal nevi who did not undergo complete imaging protocol including EDI-OCT were excluded.

Demographic characteristics, including age and sex, were recorded. Presence of symptoms (blurred vision, photopsias, metamorphopsia), the number of choroidal nevi, laterality of eye involved, clinical appearance, findings of multimodal imaging, any change compared to previous appearance on imaging, and length of follow-up, were recorded. Loss to follow-up was defined as lack of re-evaluation ≥ 12 months from last clinic visit. Specifically, the clinical features recorded included: degree of pigmentation, location (macular or extramacular), juxtapapillary (within 3 mm of optic disc margin) or not, largest basal diameter (in millimeters), presence of halo, presence of drusen, and presence of orange pigment (lipofuscin). Infrared (IR) features recorded included: whether the lesion was visible or not on infrared photography and other associated features (such as whether the lesion was dark, pale, or white; presence of flecks). FAF features recorded included: whether the lesion was visible or not and the presence of hyper-autofluorescent flecks.

EDI-OCT characteristics recorded included: number of EDI-OCT scans performed, presence of drusen, presence of subretinal fluid, presence of “discrete” appearance (anteriorly bowed hyperreflectivity cascading at lesion edges), posterior shadowing, and lesion thickness. Lesion thickness was measured with calipers utilizing a method comparable to that described by Shah et al.⁹ Lesion thickness including choroid was measured from the sclerochoroidal junction to the outermost boundary or base of RPE. Lesions were classified into one of 5 groups based on EDI-OCT appearance. Growth was considered significant if ≥ 20% increase in lesion thickness on EDI SD-OCT was noted. All imaging was interpreted and confirmed by two fellowship-trained retina specialists. Statistical analysis consisted of two-tailed Fisher’s exact tests, which were conducted using GraphPad InStat 3 for Windows, GraphPad Software (La Jolla, California, United States). A p-value <0.05 was considered statistically significant.

RESULTS

One hundred and two flat choroidal nevi in 98 patients (53 men, 45 women) were identified. The mean patient age was 61 years (range: 15–90 years). Seven (6.9%) were classified as halo choroidal nevi and 5 (4.9%) were amelanotic. Mean and median number of flat nevi were 1 (range: 1 – 3). Clinical features are detailed in Table 1.

Table 1:

Clinical Characteristics of 102 Flat Choroidal Melanocytic Lesions

Features	Number
Clinical Appearance
Nevus	62 (60.8%)
Ephelis	40 (39.2%)
Symptoms
Positive	6 (5.8%)
None	96 (94.1%)
Color
Pigmented	90 (88.2%)
Amelanotic	5 (4.9%)
Halo	7 (6.9%)
Location
Macular	53 (51.9%)
Extramacular	49 (48.0%)
Juxtapapillary (border within 3 mm of disc)	29 (28.4%)
Non-juxtapapillary	73 (71.6%)
Mean Largest Basal Diameter (mm)	2.9 (0.8 – 11.3)
Drusen
Present	33 (32.4%)
Absent	69 (67.6%)
Orange Pigment
Present	14 (13.6%)
Absent	88 (86.3%)
Infrared Appearance
White	26 (25.5%)
Pale	42 (40.8%)
Dark	13 (12.6%)
Not visible	16 (15.5%)
With flecks	5 (4.9%)
Not possible/not available	5 (4.9%)
Hyper-autofluorescent flecks	24 (23.3%)
Drusen	13 (12.6%)
Orange pigment	10 (9.7%)
Both	1 (1.0%)
Subretinal Fluid
Present	8 (7.8%)
Absent	94 (92.2%)

92.2% (94/102) had reliable EDI-OCT performed. 7.8% (8/102) of nevi did not have EDI-OCT of the nevus available for review, almost always because imaging did not capture the nevus due to its far peripheral location. The mean number of EDI-OCT scans performed was 2.1±1.5 (range: 1–8) over a median follow-up period of 9 months (range: 0–144 months). Fifty-two choroidal nevi in 51 patients were evaluated with at least 2 EDI-OCT scans longitudinally over time during the study period. There was a 23.5% (23/98 patients) rate of loss to follow-up for 12 months or more since last clinic visit.

Based on EDI-OCT findings, it became apparent that each choroidal lesion could be classified into one of the following distinct OCT patterns: 3/94 nevi (3.2%) were not visible on EDI-OCT imaging (Type 0), 35/94 nevi (37.2%) demonstrated hyperreflectivity confined within normal choroidal thickness (Type 1), 15/94 (16.0%) had anteriorly bowed hyperreflectivity cascading at discrete lesion edges with dark posterior shadowing (“discrete,” Type 2), for 14/94 (14.9%) the surface was flat but there was “posterior bowing” with scleral excavation (Type 3), and 27/94 (28.7%) had actual elevation of the surface of the nevus (Type 4). OCT features and patterns are detailed in Table 2. Figure 1 illustrates representative photos and EDI-OCTs of all 5 different subtypes. For all nevi, mean thickness was 336.4 microns + 143.9 microns. The relationship of EDI-OCT types with clinical characteristics of the choroidal lesions are detailed in Table 3.

Table 2:

Enhanced-depth Imaging Spectral-domain Optical Coherence Tomography Features of Ultrasonographically Flat Choroidal Lesions

Features	Number
Mean choroidal lesion thickness	336.4 microns (n = 88 measurable)
EDI SD-OCT Pattern (no OCT available in 8/102 lesions)
Type 0 = Not visible	3/94 (3.2%)
Type 1 = Hyperreflective but confined within normal choroidal thickness	35/94 (37.2%)
Type 2 = Discrete (waterfall appearance at lesion edges)	15/94 (16.0%)
Type 3 = Posteriorly bowed	14/94 (14.9%)
Type 4 = Elevated	27/94 (28.7%)
Posterior Shadowing (not available/not possible in 10/102)
Yes	66/92 (71.7%)
No	26/92 (28.3%)
Drusen
Yes	29/94 (30.9%)
No	65/94 (69.1%)

Figure 1: Five distinct EDI-OCT patterns of ultrasonographically flat choroidal lesions (red arrows used to demarcate boundaries).

A-D) Type 0 (EDI-OCT invisible) lesion. A) Pigmented ephelis without drusen or orange pigment. Not visible on IR (B), EDI-OCT (C), or B-scan ultrasound (D).

E-H) Type 1 “hyperreflective” lesion. E) Pigmented lesion without drusen or orange pigment. F) White on IR; G) Hyperreflective but confined within normal choroidal thickness on EDI-OCT. H) Flat on B-scan ultrasound.

I-L) Type 2 “discrete” lesion: I) Pigmented nevus with drusen and orange pigment. J) Pale on IR. K) Discrete (anteriorly bowed hyperreflectivity cascading at lesion edges) on EDI-OCT. L) Flat on B-scan ultrasound.

M-P) Type 3 “posteriorly bowed” lesion. M) Pigmented nevus with drusen and no orange pigment. N) Pale on IR. O) Posteriorly Bowed on EDI-OCT. P) Flat on B-scan ultrasound.

Q-T) Type 4 “elevated” lesion. Q) Pigmented lesion without drusen or orange pigment. R) Dark on IR. S) Elevated on EDI-OCT. T) Flat on B-scan ultrasound.

*IR = infrared (imaging); EDI-OCT = enhanced depth imaging optical coherence tomography

Table 3:

Enhanced-depth Imaging Spectral-domain Optical Coherence Tomography Features Correlated With Clinical Characteristics of Ultrasonographically Flat Choroidal Lesions

Type	LBD (mm)	Pigmentation	Juxtapapillary	Symptoms	Drusen	OP	SRF	IR	AF
0	4	Lightly pigmented	33%	33%	33%	0%	0%	Invisible	Invisible
1	1.5	Pigmented or lightly pigmented	40%	3%	23%	3%	0%	51% pale 34% white 11% invisible 3% dark	90% invisible
2	3.6	Pigmented or lightly pigmented	33%	0%	53%	20%	0%	53% white 40% pale 7% dark	80% invisible 20% hypo-AF
3	3.5	Pigmented or lightly pigmented; 21% with halo	29%	14%	43%	21%	21%	43% pale 29% white 21% invisible 7% dark w/halo	67% invisible 14% hypo-AF 14% with hyper-AF flecks
4	3.6	70% pigmented or lightly pigmented; 19% amelanotic 11% with halo	22%	7%	41%	26%	19%	37% dark (10% w/halo) 33% pale 19% invisible 11% white	64% invisible 32% hypo-AF

^*Type 0 = Not visible (n=3), Type 1 = Hyperreflective but confined within normal choroidal thickness (n = 35), Type 2 = Discrete (anteriorly bowed hyperreflectivity cascading at lesion edges) (n = 15), Type 3 = Posteriorly bowed (n = 14), Type 4 = Elevated (n = 27); LBD = mean largest basal diameter in millimeters, OP = orange pigment, SRF = subretinal fluid, IR = infrared (imaging), AF = autofluorescence (imaging)

Four percent (2/52) were noted to have a change on OCT or new appearance of known clinical risk factors for growth during the observation period. One nevus developed new hyper-autofluorescent spots and an increase in subretinal fluid but did not increase in size. One lesion increased in thickness by <10% at follow-up 22 months later. Both of these are shown in Figure 2.

Figure 2: Changes in flat choroidal lesions over time demonstrated on EDI-OCT A-B).

Juxtapapillary pigmented nevus with orange pigment and scant SRF. 6.5 months later, new hyper-autofluorescent spots and increased SRF developed but there was no increase in lesion size. The SRF is marked with a red asterisk.

C-D) Lightly pigmented nevus of elevated subtype (Type 4) with drusen, orange pigment, and SRF. Lesion thickness increased from 537 microns to 589 microns.

*EDI-OCT = enhanced depth imaging optical coherence tomography; SRF = subretinal fluid

EDI-OCT type 3 or 4 was associated with the presence of orange pigment (OR = 3.95, 95% CI: 1.1 – 13.7; p-value = 0.038). Of the 41 EDI-OCT type 3 or 4 lesions, 24.4% (10/41) had orange pigment. Conversely, of the 14 lesions with orange pigment, 71.4% (10/14) were classified as either Type 3 or 4 on EDI-OCT. Nineteen and a half percent (8/41) of Type 3 or 4 lesions had associated SRF, and conversely, of the 8 lesions with associated SRF, 100% (8/8) were classified as EDI-OCT type 3 or 4 (OR = 27.1, 95% CI: 1.5 – 486; p-value = 0.0009). Specifically, 62.5% (5/8) of these lesions with SRF were classified as EDI-OCT type 4 (OR = 4.80, 95% CI: 1.1 – 22.0; p-value = 0.04). 100% (2/2) of the choroidal lesions that had documented change over time were classified as EDI-OCT type 4, representing 7.4% (2/27) of all EDI-OCT type 4 lesions. There was a trend toward this being statistically significant (p-value = 0.08). There was no significant association between juxtapapillary location and EDI-OCT types.

DISCUSSION

We studied 102 choroidal melanocytic lesions labeled as ultrasonographically flat. Eight could not be imaged due to their far peripheral location. Of the remaining 94 nevi, 91 (96.8%) could be detected with EDI-OCT. One of the important risk factors for nevus growth identified by Shields et al. is tumor thickness greater than 2 mm.³ Although ultrasonography has long been used to measure choroidal tumors, it is imprecise, especially for smaller lesions. Ultrasonography has been shown to overestimate choroidal melanoma thickness by at least 2 mm in 10% of cases when compared to histopathologic analysis.¹² Torres et al. demonstrated that choroidal tumors less than 1 mm in height that were deemed undetectable by ultrasonography could actually be measured with EDI-OCT.⁸ Since choroidal nevi should undergo appropriate monitoring based on clinical characteristics with the goal of early detection and treatment of those nevi that evolve into uveal melanomas, EDI-OCT may be more sensitive than ultrasonography in monitoring small nevi.

With the advancement of SD-OCT to include EDI, we can identify features beyond the anterior surface and can more definitively identify tumor boundaries, including the posterior boundaries of these flat lesions.⁸ We demonstrated the ability to consistently identify and measure the thickness of ultrasonographically flat choroidal lesions using EDI-OCT in 94% of nevi (3% not visible; 3% with difficult to discern posterior boundary due to heavy shadowing). This allows for an objective, quantitative metric that can be followed over time to monitor for growth, especially in high-risk patients. Francis et al. found EDI-OCT and swept-source OCT (SS-OCT) to be comparable in evaluation of predominantly-elevated choroidal nevi in most respects, including the ability to identify nevus-scleral interface and secondary retinal changes. Furthermore, they classified elevated choroidal nevi based on distention patterns into one of three OCT configurations – dome, plateau, or almond. The morphology of 90% (27/30) of these nevi were identifiable with ultrasonography.¹³

We propose a classification system consisting of five distinct EDI-OCT patterns. We hypothesize that both histopathologic characteristics of the choroidal lesions, as well as artifacts of imaging, contribute to the appearance of each type of lesion on EDI-OCT. We hypothesize that Type 0, which are not visible on EDI-OCT, correspond to lesions that actually represent focal hyperpigmentation of otherwise histologically-normal choroid without true nevus cells. Type 1 lesions, with hyperreflectivity confined within the normal boundaries of the choroid, may represent replacement of the choroid, or partial thickness replacement of the choroid, with nevus cells. This would be the choroidal equivalent to an ephelis (akin to what can be seen on the iris or on the skin), as opposed to a true nevus. The choroidal ephelis has been referenced previously.¹⁴ Type 2 lesions have a hyperreflective, anterior border with dark posterior shadowing. We initially posited that this hyperreflectivity may arise from dense pigmentation at the surface of the nevus [akin to what is seen with melanocytoma, and to a lesser degree, with congenital hypertrophy of the retinal pigment epithelium (CHRPE)]. However, we found that both darkly pigmented and lightly pigmented nevi (as well as occasional amelanotic nevi) could have a Type 2 appearance on EDI-OCT (Table 3). Reflectivity is the result of tissue density and interfaces. In this case, the apparent optical density could be altered by cell orientation and shape. Nevus cells are spindle cells, and the peripheral cells may be oriented in such a way as to increase the reflectivity. While this may be occurring around the entire nevus (or even within the center of the nevus), the anterior cells cast a dark shadow that obscures this effect below. For comparison, a similar effect is seen with SD-OCT imaging of melanocytomas, where the deep pigmentation is throughout the tumor on histopathology, but only the anterior border appears hyperreflective, with the underlying tumor obscured by shadowing form this overlying reflectivity. The effacement of the overlying choriocapillaris, as well as the compression of surface nevus cell, may also contribute to these interface changes at the anterior border, and ultimately contribute to the hyperreflectivity seen on EDI-OCT. The suggestion that the OCT appearance of Type 2 nevi may have more to do with imaging artifacts than histopathology is supported by the observation that the lesions have sharp lateral borders on OCT, and the hyperreflectivity cascades down the edges but does not wrap around posteriorly, due to the shadowing effect. In contrast to the above, we propose that Type 3 nevi represent true nevi, which just appear flat at the surface because the area of expansion extends posteriorly into the sclera causing excavation. Type 4 nevi are also true nevi that have some height to them, but the height is below the threshold of detectability on ultrasound, thus appearing “flat” on ultrasound, while being elevated on the higher-resolution EDI-OCT.

While both Type 2 nevi as well as Type 4 nevi may have an elevated anterior boundary, the most distinctive element of type 2 nevi is not the elevation (which is characteristic of all type 4 nevi) but rather the defining characteristic of Type 2 nevi is the prominent anterior hyperreflectivity with the associated posterior shadowing. Because of the posterior shadowing, it is not easy to delineate the posterior border. With Type 4 nevi, the posterior border can be easily delineated and thickness more easily measured due to absence of the prominent (and characteristic) anterior hyperreflectivity causing posterior shadowing. This distinction is clinically meaningful and not just academic, as it means that thickness (and changes in thickness) can be clearly measured in Type 4 lesions, while this may be more difficult in Type 2 lesions. Another difference is that type 2 lesions have very discrete lateral borders whereas Type 4 lesions gradually blend into normal thickness choroid. Because of this clear difference in the morphology at the edges, we do not feel that the differences between Type 2 and Type 4 lesions are solely due to an artifact of imaging and shadowing.

Though clinically flat with undetectable morphology on ultrasonography, 28.7% were found to be elevated on EDI-OCT. Critically, only type 4 (elevated) lesions demonstrated change over time or growth in our study. Though these ultrasonographically flat choroidal lesions could be measured and followed over time, they rarely demonstrated significant change or growth. Only 4% (2/52) of our cohort demonstrated change. One developed new hyper-autofluorescent spots and increased subretinal fluid with a nominal increase in thickness over 13 months of follow-up. The other lesion increased in thickness by approximately 10% over 22 months of follow-up. Both of these lesions were of the elevated subtype (EDI-OCT type 4), and both lesions also had many other high-risk features. The first had juxtapapillary location, orange pigment, subretinal fluid, no drusen, and no halo. The second had symptoms, orange pigment, subretinal fluid, and no halo.

The strengths of our study include good sample size, comprehensive/consistent clinical examination, and standardized imaging protocol. The limitations of our study include retrospective design, variable follow-up, and inability to image a small fraction of choroidal lesions that occurred in the far periphery with EDI-OCT. Though many patients had > 12 months of follow-up, it was not uniform across the cohort. Of the total cohort, 23.5% were lost to follow-up for 12 months or more from the last clinic visit. This may have induced bias in detection of change in choroidal lesions evaluated. Though the sample size is comparable to other large published cohorts, since growth is rare in this subgroup of nevi, larger studies are needed to fully address the question of monitoring growth over time using this imaging modality. Given the low rate of growth in these flat choroidal nevi and lack of long-term prospective imaging and clinical data for a large cohort of these patients, as well as the lack of any patients who actually developed malignant melanoma, caution is advised in relying too heavily on EDI-OCT as a sole tool in monitoring for growth of flat nevi, pending larger confirmatory studies.

In conclusion, we have created a novel classification system of different subtypes of ultrasonographically flat choroidal lesions based on their EDI-OCT appearance. More than a quarter of these ultrasonographically flat choroidal lesions demonstrated elevation on EDI-OCT, and many more of those that did not have surface elevation had thickness that was invisible on ultrasound due to bowing in the posterior direction. EDI-OCT type 3 or 4 tended to be associated with other known high-risk features predictive of growth of choroidal nevi such as presence of symptoms, orange pigment, and subretinal fluid, and lack of drusen. The two choroidal nevi that demonstrated change over follow-up were of the elevated EDI-OCT type 4 class, with minor growth detected in one Type 4 lesion. With the ability to measure lesion thickness using EDI-OCT, flat choroidal lesions can be monitored objectively and measured quantitatively, with the goal of early detection and treatment of lesions that undergo malignant transformation. Elevation on EDI-OCT may carry a greater risk of growth but a larger cohort with long-term study is needed to better define the risk. Conversely, the presence of an EDI-OCT type 0 or type 1 appearance may be reassuring. Physicians caring for patients with ultrasonographically flat choroidal nevi could be advised to consider obtaining measurements using EDI-OCT instead.