Stephanie J. Chiu

Automatic segmentation of seven retinal layers in SDOCT images congruent with expert manual segmentation

Segmentation of anatomical and pathological structures in ophthalmic images is crucial for the diagnosis and study of ocular diseases. However, manual segmentation is often a time-consuming and subjective process. This paper presents an automatic approach for segmenting retinal layers in Spectral Domain Optical Coherence Tomography images using graph theory and dynamic programming. Results show that this method accurately segments eight retinal layer boundaries in normal adult eyes more closely to an expert grader as compared to a second expert grader.

Validated Automatic Segmentation of AMD Pathology Including Drusen and Geographic Atrophy in SD-OCT Images

PURPOSE To automatically segment retinal spectral domain optical coherence tomography (SD-OCT) images of eyes with age-related macular degeneration (AMD) and various levels of image quality to advance the study of retinal pigment epithelium (RPE)+drusen complex (RPEDC) volume changes indicative of AMD progression. METHODS A general segmentation framework based on graph theory and dynamic programming was used to segment three retinal boundaries in SD-OCT images of eyes with drusen and geographic atrophy (GA). A validation study for eyes with nonneovascular AMD was conducted, forming subgroups based on scan quality and presence of GA. To test for accuracy, the layer thickness results from two certified graders were compared against automatic segmentation results for 220 B-scans across 20 patients. For reproducibility, automatic layer volumes were compared that were generated from 0° versus 90° scans in five volumes with drusen. RESULTS The mean differences in the measured thicknesses of the total retina and RPEDC layers were 4.2 ± 2.8 and 3.2 ± 2.6 μm for automatic versus manual segmentation. When the 0° and 90° datasets were compared, the mean differences in the calculated total retina and RPEDC volumes were 0.28% ± 0.28% and 1.60% ± 1.57%, respectively. The average segmentation time per image was 1.7 seconds automatically versus 3.5 minutes manually. CONCLUSIONS The automatic algorithm accurately and reproducibly segmented three retinal boundaries in images containing drusen and GA. This automatic approach can reduce time and labor costs and yield objective measurements that potentially reveal quantitative RPE changes in longitudinal clinical AMD studies. (ClinicalTrials.gov number, NCT00734487.).

Dynamics of Human Foveal Development after Premature Birth

PURPOSE To determine the dynamic morphologic development of the human fovea in vivo using portable spectral domain-optical coherence tomography (SD-OCT). DESIGN Prospective, observational case series. PARTICIPANTS Thirty-one prematurely born neonates, 9 children, and 9 adults. METHODS Sixty-two neonates were enrolled in this study. After examination for retinopathy of prematurity (ROP), SD-OCT imaging was performed at the bedside in nonsedated infants aged 31 to 41 weeks postmenstrual age (PMA) (= gestational age in weeks + chronologic age) and at outpatient follow-up ophthalmic examinations. Thirty-one neonates met eligibility criteria. Nine children and nine adults without ocular pathology served as control groups. Semiautomatic retinal layer segmentation was performed. Central foveal thickness, foveal to parafoveal (FP) ratio (central foveal thickness divided by thickness 1000 μm from the foveal center), and 3-dimensional thickness maps were analyzed. MAIN OUTCOME MEASURES In vivo determination of foveal morphology, layer segmentation, analysis of subcellular changes, and spatiotemporal layer shifting. RESULTS In contrast with the adult fovea, several signs of immaturity were observed in the neonates: a shallow foveal pit, persistence of inner retinal layers (IRLs), and a thin photoreceptor layer (PRL) that was thinnest at the foveal center. Three-dimensional mapping showed displacement of retinal layers out of the foveal center as the fovea matured and the progressive formation of the inner/outer segment band in the opposite direction. The FP-IRL ratios decreased as IRL migrated before term and minimally after that, whereas FP-PRL ratios increased as PRL subcellular elements formed closer to term and into childhood. A surprising finding was the presence of cystoid macular edema in 58% of premature neonates that appeared to affect inner foveal maturation. CONCLUSIONS This study provides the first view into the development of living cellular layers of the human retina and of subcellular specialization at the fovea in premature infant eyes using portable SD-OCT. Our work establishes a framework of the timeline of human foveal development, allowing us to identify unexpected retinal abnormalities that may provide new keys to disease activity and a method for mapping foveal structures from infancy to adulthood that may be integral in future studies of vision and visual cortex development. FINANCIAL DISCLOSURE(S) Proprietary or commercial disclosure may be found after the references.

Maturation of the Human Fovea: Correlation of Spectral-Domain Optical Coherence Tomography Findings With Histology

PURPOSE To correlate human foveal development visualized by spectral-domain optical coherence tomography (SDOCT) with histologic specimens. DESIGN Retrospective, observational case series. METHODS Morphology and layer thickness of retinal SDOCT images from 1 eye each of 22 premature infants, 30 term infants, 16 children, and 1 adult without macular disease were compared to light microscopic histology from comparable ages. RESULTS SDOCT images correlate with major histologic findings at all time points. With both methods, preterm infants demonstrate a shallow foveal pit indenting inner retinal layers (IRL) and short, undeveloped foveal photoreceptors. At term, further IRL displacement forms the pit and peripheral photoreceptors lengthen; the elongation of inner and outer segments (IS and OS, histology) separates the IS band from retinal pigment epithelium. Foveal IS and OS are shorter than peripheral for weeks after birth (both methods). By 13 months, foveal cone cell bodies stack >6 deep, Henle fiber layer (HFL) thickens, and IS/OS length equals peripheral; on SDOCT, foveal outer nuclear layer (which includes HFL) and IS/OS thickens. At 13 to 16 years, the fovea is fully developed with a full complement of SDOCT bands; cone cell bodies >10 deep have thin, elongated, and tightly packed IS/OS. CONCLUSIONS We define anatomic correlates to SDOCT images from normal prenatal and postnatal human fovea. OCT bands typical of photoreceptors of the adult fovea are absent near birth because of the immaturity of foveal cones, develop by 24 months, and mature into childhood. This validates the source of SDOCT signal and provides a framework to assess foveal development and disease.

Automated non-rigid registration and mosaicing for robust imaging of distinct retinal capillary beds using speckle variance optical coherence tomography

Variance processing methods in Fourier domain optical coherence tomography (FD-OCT) have enabled depth-resolved visualization of the capillary beds in the retina due to the development of imaging systems capable of acquiring A-scan data in the 100 kHz regime. However, acquisition of volumetric variance data sets still requires several seconds of acquisition time, even with high speed systems. Movement of the subject during this time span is sufficient to corrupt visualization of the vasculature. We demonstrate a method to eliminate motion artifacts in speckle variance FD-OCT images of the retinal vasculature by creating a composite image from multiple volumes of data acquired sequentially. Slight changes in the orientation of the subject’s eye relative to the optical system between acquired volumes may result in non-rigid warping of the image. Thus, we use a B-spline based free form deformation method to automatically register variance images from multiple volumes to obtain a motion-free composite image of the retinal vessels. We extend this technique to automatically mosaic individual vascular images into a widefield image of the retinal vasculature.

Progression of Intermediate Age-related Macular Degeneration with Proliferation and Inner Retinal Migration of Hyperreflective Foci

PURPOSE Drusen and migrating retinal pigment epithelium have been associated with hyperreflective foci (HF) detected by spectral-domain optical coherence tomography (SD-OCT). This study sought to quantify the change in intraretinal HF distribution and its correlation with age-related macular degeneration (AMD) disease progression. DESIGN Prospective observational study from the multicenter Age-Related Eye Disease Study 2 (AREDS2) Ancillary SD-OCT Study. PARTICIPANTS Patients (n=299) with 1 enrolled eye with intermediate AMD and baseline SD-OCT, followed by SD-OCT imaging at 1-year and 2-year visits. METHODS The number and location of HF were scored in SD-OCT scans of all 299 eyes. The change in transverse (horizontal) and axial (vertical) distribution of HF in the macula were evaluated with pairwise signed-rank tests. Two-year inner retinal HF migration was determined by the change in HF-weighted axial distribution (AxD) score calculated for each eye. The correlation of HF with SD-OCT features of AMD progression was evaluated with logistic regression analysis. MAIN OUTCOME MEASURES The mean change in number of HF, transverse and axial distribution of HF in the macula, and AxD per eye. RESULTS In 299 study eyes, the 2-year increase in the number of HF (P<0.001) and the AxD (P<0.001) per eye represented longitudinal proliferation and shift to inner retinal layers, respectively. Eyes with geographic atrophy (GA) at 2 years were correlated with the presence of baseline HF (P<0.001; odds ratio [OR], 4.72; 95% confidence interval [CI], 2.43-9.80), greater number of baseline HF (P<0.001; OR, 1.61 per HF; 95% CI, 1.32-2.00), and greater baseline AxD (P<0.001; OR, 1.58 per AxD point; 95% CI, 1.29-1.95). CONCLUSIONS Proliferation and inner retinal migration of SD-OCT HF occurred during follow-up in eyes with intermediate AMD. These characteristics were associated with greater incidence of GA at year 2; therefore, SD-OCT HF proliferation and migration may serve as biomarkers for AMD progression. FINANCIAL DISCLOSURE(S) Proprietary or commercial disclosure may be found after the references.

Robust automatic segmentation of corneal layer boundaries in SDOCT images using graph theory and dynamic programming.

Segmentation of anatomical structures in corneal images is crucial for the diagnosis and study of anterior segment diseases. However, manual segmentation is a time-consuming and subjective process. This paper presents an automatic approach for segmenting corneal layer boundaries in Spectral Domain Optical Coherence Tomography images using graph theory and dynamic programming. Our approach is robust to the low-SNR and different artifact types that can appear in clinical corneal images. We show that our method segments three corneal layer boundaries in normal adult eyes more accurately compared to an expert grader than a second grader—even in the presence of significant imaging outliers.

Kernel regression based segmentation of optical coherence tomography images with diabetic macular edema

We present a fully automatic algorithm to identify fluid-filled regions and seven retinal layers on spectral domain optical coherence tomography images of eyes with diabetic macular edema (DME). To achieve this, we developed a kernel regression (KR)-based classification method to estimate fluid and retinal layer positions. We then used these classification estimates as a guide to more accurately segment the retinal layer boundaries using our previously described graph theory and dynamic programming (GTDP) framework. We validated our algorithm on 110 B-scans from ten patients with severe DME pathology, showing an overall mean Dice coefficient of 0.78 when comparing our KR + GTDP algorithm to an expert grader. This is comparable to the inter-observer Dice coefficient of 0.79. The entire data set is available online, including our automatic and manual segmentation results. To the best of our knowledge, this is the first validated, fully-automated, seven-layer and fluid segmentation method which has been applied to real-world images containing severe DME.

Characterization of the Choroid-Scleral Junction and Suprachoroidal Layer in Healthy Individuals on Enhanced-Depth Imaging Optical Coherence Tomography

IMPORTANCE Accurate measurements of choroidal thickness (CT) using enhanced-depth imaging optical coherence tomography (EDI-OCT) require a well-defined choroid-scleral junction (CSJ), which may appear in some eyes as a hyporeflective band corresponding to the suprachoroidal layer (SCL). OBJECTIVE To identify factors associated with the presence and thickness of the SCL in healthy participants and determine how different CSJ boundary definitions impact CT measurements. DESIGN, SETTING, AND PARTICIPANTS Secondary analysis of EDI-OCT images obtained prospectively from 74 eyes of 74 controls (mean age, 68.6 years) from the Age-Related Eye Disease Study 2 Ancillary SDOCT Study. MAIN OUTCOMES AND MEASURES The CSJ appearances were categorized as either having no visible SCL or a hyporeflective band corresponding to the SCL. Ocular parameters associated with the presence and thickness of the SCL were identified. Subfoveal CT was measured using 3 different posterior boundaries: (1) the posterior vessel border (vascular CT [VCT]), (2) inner border of the SCL (stromal CT [StCT]), and (3) inner border of the sclera (total CT [TCT]). Manual segmentation using custom software was used to compare VCT, StCT, and TCT across the macula. RESULTS The SCL was visible in 33 eyes (44.6%). Factors associated with SCL presence and thickness included hyperopic refractive error (R2 = 0.123; P = .045) and increased TCT (R2 = 0.215; P = .004), but not age, visual acuity, intraocular pressure, retinal foveal thickness, VCT, or StCT. In eyes where the SCL was not visible, mean [SD] subfoveal VCT was 222.3 [101.5] μm and StCT and TCT were 240.0 [99.0] μm, with a difference of 17.7 [16.0] μm (P < .001). In eyes where the SCL was visible, mean [SD] subfoveal VCT, StCT, and TCT were 221.9 [83.1] μm, 257.7 [97.3] μm, and 294.1 [104.8] μm, respectively, with the greatest difference of 72.2 [30.4] μm between VCT and TCT (P < .001). All 3 CT measurements were significantly different along all points up to 3.0 mm nasal and temporal to the fovea. CONCLUSIONS AND RELEVANCE A hyporeflective SCL is visible at the CSJ on EDI-OCT in nearly half of healthy individuals, and its presence correlates with hyperopia. Different posterior boundary definitions may result in significant differences in CT measurements and should be explicitly identified in future choroidal studies and segmentation algorithms.

Pilocarpine-Induced Dilation of Schlemm's Canal and Prevention of Lumen Collapse at Elevated Intraocular Pressures in Living Mice Visualized by OCT

PURPOSE The goal was to assess effects of IOP and pilocarpine-induced ciliary muscle contraction on conventional outflow pathway tissues in living anesthetized mice. METHODS Intraocular pressure was controlled by intracameral cannulation of mouse eyes while imaging using spectral-domain optical coherence tomography (SD-OCT). Time-lapse sagittal SD-OCT sections through Schlemms canal (SC) were acquired while changing IOP stepwise between 10 and 45 mm Hg. After topical application of 1% pilocarpine, the series of IOP steps and imaging were repeated. Effects of pilocarpine on IOP and outflow facility in living mice were verified by rebound tonometry and flow measurements at three different IOPs, respectively. In vivo OCT images were compared with eyes analyzed by standard histology. RESULTS In living mice imaged by SD-OCT, the lumen of SC progressively collapsed with increasing IOP, reaching near complete closure at 20 mm Hg. Schlemms canal collapse was reversible, with the lumen opening within minutes after returning IOP from 45 to 10 mm Hg. Pilocarpine-induced ciliary muscle contraction changed SC lumen area by 131.6% ± 21.0% compared with untreated controls at 10 mm Hg, opened the trabecular meshwork, and prevented complete collapse of the SC lumen at higher pressures. Similar results were observed by standard histology. Pilocarpine increased outflow facility 4-fold (P = 0.02) and lowered IOP (16.46 ± 2.23 vs. 11.08 ± 2.28 mm Hg, P = 0.03). CONCLUSIONS Spectral-domain OCT was effective at visualizing changes in SC lumen in living mice. Results with pilocarpine are consistent with the concept that a primary role for the ciliary muscle is to prevent collapse of SC.