Enhanced Visualization of Subtle Outer Retinal Pathology by En Face Optical Coherence Tomography and Correlation with Multi-Modal Imaging

Abstract

Purpose: To present en face optical coherence tomography (OCT) images generated by graph-search theory algorithm-based custom software and examine correlation with other imaging modalities.

Methods: En face OCT images derived from high density OCT volumetric scans of 3 healthy subjects and 4 patients using a custom algorithm (graph-search theory) and commercial software (Heidelberg Eye Explorer software (Heidelberg Engineering)) were compared and correlated with near infrared reflectance, fundus autofluorescence, adaptive optics flood-illumination ophthalmoscopy (AO-FIO) and microperimetry.

Results: Commercial software was unable to generate accurate en face OCT images in eyes with retinal pigment epithelium (RPE) pathology due to segmentation error at the level of Bruch's membrane (BM). Accurate segmentation of the basal RPE and BM was achieved using custom software. The en face OCT images from eyes with isolated interdigitation or ellipsoid zone pathology were of similar quality between custom software and Heidelberg Eye Explorer software in the absence of any other significant outer retinal pathology. En face OCT images demonstrated angioid streaks, lesions of acute macular neuroretinopathy, hydroxychloroquine toxicity and Bietti crystalline deposits that correlated with other imaging modalities.

Conclusions: Graph-search theory algorithm helps to overcome the limitations of outer retinal segmentation inaccuracies in commercial software. En face OCT images can provide detailed topography of the reflectivity within a specific layer of the retina which correlates with other forms of fundus imaging. Our results highlight the need for standardization of image reflectivity to facilitate quantification of en face OCT images and longitudinal analysis.

Fig 1

En face optical coherence tomography images of the ellipsoid zone (A-I:V), interdigitation zone (B-I:V), retinal pigment epithelium (C-I:V) and Bruch’s membrane (D-I:V). Images from the first 3 columns are from a normal subject and the last column is from a patient with Bietti crystalline dystrophy. The scanning protocol covers a 15° (horizontal) × 10° (vertical) field of view on the retina. Each row (I:V) corresponds to increasing separation between consecutive B-scans: 11, 30, 60, 120 and 240 μm.

Fig 2

A 20° × 20° Near-infrared reflectance image NIR image (A) showing the region of the volumetric dense raster scan (black rectangle) and the location of the horizontal OCT B-scan (white line). Unprocessed OCT scan showed irregular elevation of the ellipsoid line due to a vitelliform lesion (B). Three layers of interest were segmented in this B-scan (C). The B-scan contrast is enhanced to reduce image quality variation between B-scans within the raster scan set (D). Using the base of the retinal pigment epithelium as a reference layer, the B-scan is flattened and a parallel boundary is chosen for creating a slab of pixels for generating the en face OCT image (E). The same procedure is repeated for consecutive B-scans to generate a reflectivity profile across the scanning area. The en face OCT image can be overlaid on the NIR image (F).

Fig 3

Multimodal imaging of the left eye of a healthy subject showing normal 15° × 10° Near-infrared reflectance NIR (A), infrared autofluorescence IRAF (B), microperimetry (C), blue-light autofluorescence BAF (D), adaptive optics flood illumination ophthalmoscopy AO-FIO cone montage and density maps overlaid on NIR (E,F), and en face OCT maps of the ellipsoid zone EZ (G), interdigitation zone IZ (H), and retinal pigment epithelium RPE (I). HE, Heidelberg Eye Explorer software (Heidelberg Engineering); CU, Custom-built software using graph-search theory algorithm. White dotted rectangle corresponds to region for which en face maps are generated.

Fig 4. Multimodal imaging showing the left eye of Case 1: angioid streaks secondary to pseudoxanthoma elasticum.

Near-infrared reflectance NIR (A), infrared autofluorescence IRAF (B), microperimetry (C), blue-light autofluorescence BAF (D), and adaptive optics flood illumination ophthalmoscopy AO-FIO cone montage and density maps overlaid on NIR image (E, F) of the central 15° × 10° showing the angioid streaks as irregular linear hypo-reflective lesions. The occasional streaks showing hypo-autofluorescence, relative scotoma in regions unaffected by angioid streaks and cone tip reflexes within the region of angioid streaks. Microperimetry overlaid on cone tip reflex montage (G) showed reduced cone densities compared with normative values, both in regions with reduced (H:I) and normal sensitivity (H:II-III). B-scans flattened to the base of retinal pigment epithelium RPE (I) are used for generating en face OCT images of the ellipsoid zone (J), apical portion of the RPE (K) and basal portion of the RPE (L). Inaccuracy in automated segmentation of the basal RPE by HE resulted in artefact in the ellipsoid zone and RPE en face OCT images (red arrow in K, L and M). The linear pattern of angioid streaks were only visible in the en face OCT image derived from the basal portion of the RPE (L). HE, Heidelberg Eye Explorer software (Heidelberg Engineering); CU, Custom-built software using graph-search theory algorithm. White dotted rectangle corresponds to region for which en face OCT images are generated.

Fig 5

Multimodal imaging showing the left eye of Case 2: acute macular neuroretinopathy at presentation (A-G), and 6 months (H-N) of follow-up. Near-infrared reflectance NIR (A,H), adaptive optics flood illumination ophthalmoscopy AO-FIO density map overlaid on NIR (B,I,) and magnified cone images (D,K), microperimetry (C,J), OCT B-scans in the region of relative scotoma (E,L) and en face OCT images of ellipsoid (F,M), and interdigitation zones (G,N) showed partial recovery. Both commercial and custom software were able to demonstrate the topography of ellipsoid and interdigitation zone injury.

Fig 6. Multimodal imaging showing the left eye of Case 3: early hydroxychloroquine toxicity.

20° × 20° Near-infrared reflectance NIR (A) reveals a subtle hyper-reflective foveal signal while infrared autofluorescence IRAF (B), microperimetry (C) and blue-light autofluorescence BAF (D) are within normal limits. Adaptive optics flood illumination ophthalmoscopy AO-FIO density map overlaid on NIR (E) and cone image (F) reveals a transition zone in the parafovea where cone reflexes are lost (E). OCT B-scan (G) through the foveal center demonstrates preserved (i.) and attenuated (ii.) interdigitation zone at parafovea. The en face OCT image reveals a para- and perifoveal concentric zone of reduced reflectivity from the ellipsoid zone EZ (H) and the interdigitation zone IZ (I). Retinal pigment epithelium RPE layer en face OCT appeared normal (J). In this case, HE is able to accurately identify the contour of Bruch’s membrane and hence the en face OCT images are similar to the one generated by our custom software (CU). White dotted square corresponds to region for which en face maps are generated.

Fig 7. Multimodal imaging showing the left eye of Case 4: Bietti crystalline dystrophy.

Although crystals are seen, the boundary of retinal degeneration is not clearly visualized on Near-infrared reflectance NIR (A), infrared autofluorescence IRAF (B) or blue-light autofluorescence BAF (C). Microperimetry (D) demonstrates preserved retinal sensitivity within 1° of the center of fixation indicating the presence of foveal photoreceptor cells supported by an island of RPE. Crystals are also visualized with adaptive optics flood-illumination ophthalmoscopy AO-FIO (E) but the poor signal from cones precluded AO-FIO cone density mapping (F). The OCT B-scans (G) reveal atrophy of the RPE and outer retinal layers. As a direct consequence, the errors in automated segmentation of the Bruch’s membrane by HE lead to artefacts in the en face OCT images of the ellipsoid-RPE slab (H-HE) and the Bruch’s membrane (I-HE). Our custom algorithm was able to generate ellipsoid-RPE (H-CU) and Bruch’s membrane (I-CU) en face OCT images. The well-defined central island of ellipsoid-RPE correlated well with the region of retinal sensitivity on microperimetry (J). The distribution of crystals seen on AO imaging (E, yellow arrows) mirrors the lesions seen on en face OCT of Bruch’s membrane (I-CU, K, yellow arrows).