Analysis of Autofluorescent retinal images and measurement of atrophic lesion growth in Stargardt disease

https://doi.org/10.1016/j.exer.2010.03.021Get rights and content

Abstract

Current retinal imaging techniques using scanning laser ophthalmoscopy (SLO) provide a powerful mechanism for characterizing the topographical distribution of lipofuscin fluorophores and atrophic lesions (ALs) in retinal disease. In this paper we describe a novel Edge-Flow-Driven Variational Image Segmentation analysis to measure and evaluate progressive change in the area of ALs as well as regions of hyperfluorescence (HF). The algorithm is embedded in a series of almost completely automated image processing steps that allow rapid comparison of serial images. The sensitivity of the methodology to detect change was evaluated by measuring progression of AF lesion size in a cohort of Stargardt Macular Dystrophy (STGD) patients. Fifty-two STGD subjects (mean age = 41.0 ± 16.6 years, range 9–78 yrs) at varying stages of disease participated in this prospective study. Twenty-four of the 52 subjects presented with atrophic lesions in one or both eyes on first evaluation. For this subgroup of subjects, the mean (±1 sd) follow-up time was 2.92 (+0.26) years (range 0.57–3.26 years) and the mean (±1 sd) rate of change was found to be approximately 0.94 (±0.87) mm2/year (range 0.2–2.13 mm2/yr). With this methodology, progressive enlargement of AL area was detectable in as little as one year, while regions of HF generally decreased, although there was considerable variability in the appearnce of HF, presumably reflecting the combined effects of the creation or expansion of lipofuscin deposits and resorption and loss associated with retinal cell death. Our findings suggest that this methodology is sufficiently sensitive to detect change and provides a clinically relevant tool to monitor progression not only with regards to natural history, but also to evaluate the efficacy of potential therapeutic interventions in STGD. Finally, we evaluated the association between AL area and measures of rod- and cone-mediated retinal function, as assessed with electroretinography (ERG). In general, the larger the AL, the poorer the ERG response, with a greater impact of lesion size on cone- rather than rod-mediated retinal function, a finding that was expected on the basis of the location and size of the AL and the distribution of rod- and cone-photoreceptors.

Introduction

The accumulation of toxic lipofuscin fluorophores in the retinal pigment epithelium (RPE) represents a common pathological pathway leading to photoreceptor cell loss in a number of disease entities (Bui et al., 2006, Dorey et al., 1989, Eldred and Lasky, 1993, Feeney-Burns et al., 1980, Okubo et al., 1999, Sparrow and Boulton, 2005, Weiter et al., 1986) (for a review, see Travis et al., 2007). For example, the lipofuscin fluorophores, A2E and A2PE-H2, are dramatically elevated in the RPE of postmortem samples taken from patients with recessive Stargardt Disease (STGD) and in the abcr−/− mouse model of STGD (Mata et al., 2000, Weng et al., 1999). The cytotoxicity of A2E in RPE cells is well known and is assumed to occur through a series of well-defined biochemical events that ultimately lead to RPE and overlying photoreceptor cell death (De and Sakmar, 2002, Finnemann et al., 2002, Sparrow et al., 1999, Sparrow et al., 2003, Suter et al., 2000, Travis et al., 2007).

Scanning laser ophthalmoscopy (SLO) provides an effective means for characterizing the topographical distribution of lipofuscin fluorophores noninvasively in human subjects. This imaging technique capitalizes on a fundamental property of retinal fluorophores in that they emit light in the spectral range 500–700 nm when excited by short-wavelength light (<490 nm). The topographical distribution and density of the emitted light define regions where the lipofuscin fluorophores have accumulated above background levels (hyperfluorescence, HF). More importantly, intact RPE and overlying photoreceptor cells are required for the production of lipofuscin fluorophores. Thus, in the absence of other factors that might reduce fundus autofluorescence (e.g. hemorrhages), regions with fluorescent signal well below background levels (hypofluorescence) identify regions where RPE and photoreceptor cells are presumed to have been irretrievably lost. Thus, AF imagery has become a powerful tool to identify regions of retinal atrophy. Further, the abnormal patterns of increased fundus fluorescence that surround atrophic lesions (ALs) may serve as a prognostic marker for the future progression and expansion of the AL (Holz et al., 2001, Holz et al., 2007, Schmitz-Valckenberg et al., 2006, Schmitz-Valckenberg et al., 2009). However, the validity of this sequence of events has recently been challenged (Hwang et al., 2006, Smith et al., 2009).

In this study, a novel Edge-Flow-Driven image segmentation algorithm (UCSB Vision Research Lab, vision.ece.ucsb.edu) is described and used to measure AL area in STGD. The algorithm is embedded in a series of automated image processing steps that allow rapid analysis of serial images. We test the hypothesis that this methodology is sufficiently sensitive to monitor progressive change of AL size in STGD so that it might be used as an outcome measure in a clinical trial. In addition, we evaluated the association of regional loss of RPE and overlying photoreceptors, particularly in areas where rod and cone photoreceptor densities are highest, with measures of rod- and cone-mediated retinal function as determined by standardized electroretinography (ERG). We test the hypothesis that large central ALs that are accompanied by subnormal ERG responses may be interpreted incorrectly as indicating panretinal dysfunction.

Section snippets

Subjects

Fifty-two STGD subjects (mean age = 41.0 ± 16.6 years, range 9–78 yrs) at varying stages of disease participated in a prospective study. All affected individuals were confirmed as having recessive (or sporadic) STGD or STGD-like disease based on a clinical exam by at least two retina specialists based on the presence of bilateral macular or posterior pole disease consisting of subretinal flecks, RPE disturbances and/or geographic atrophy and a careful family history. Based on fundus appearance

Measurement variability

A series of 5–7 AF images were obtained from each of five STGD patients, all recorded on the same day but with different camera sensitivities to vary edge gradients as described in the Methods section. In addition, the patient’s head was repositioned between trials to produce changes in image orientation and magnification. Across all tested eyes, the median variation (data not shown) for the series of images was 2.9% (range = 2.2%–7.7%). (A t-test comparing the variability between eyes was not

Discussion

The main goal of this study was to evaluate progressive change in AL size in patients with recessive STGD. To this end, a novel automated methodology was developed to process AF retinal images obtained with scanning laser ophthalmoscopy. The methodology required linking together several image analysis programs that ran sequentially but which automatically accomplish the task of registering serial images, maximizing signal-to-noise ratios, and segmenting the AL using an Edge-Flow-Driven

Commercial relationships

None.

Acknowledgements

We are grateful to the Foundation Fighting Blindness and the Sarkaria Family Fund for their generous support of this research. We also thank Kristin Lipka and David Le Beck for their assistance with recording the Autofluorescent images, and Dr. Carolina M. Ortube, Ariadna Martinez, and Arturo Garcia for their help with recruiting patients.

Grant Support: Foundation Fighting Blindness and the Sarkaria Family Fund for Macular Dystrophy Research.

References (43)

  • R.O. Duda et al.

    Pattern Classification and Scene Analysis. Pattern Classification and Scene Analysis

    (1973)
  • G.E. Eldred et al.

    Retinal age pigments generated by self-assembling lysosomotropic detergents

    Nature

    (1993)
  • S.C. Finnemann et al.

    The lipofuscin component A2E selectively inhibits phagolysosomal degradation of photoreceptor phospholipid by the retinal pigment epithelium

    Proc. Natl. Acad. Sci. U.S.A.

    (2002)
  • A.F. Frangi et al.

    Multiscale vessel enhancement filtering

  • G.A. Hewer et al.

    Variational image segmentation using boundary functions

    IEEE Trans. Image Process

    (1998)
  • F.G. Holz et al.

    Fundus autofluorescence and development of geographic atrophy in age-related macular degeneration

    Invest. Ophthalmol. Vis. Sci.

    (2001)
  • F.G. Holz et al.

    Progression of geographic atrophy and impact of fundus autofluorescence patterns in age-related macular degeneration

    Am. J. Ophthalmol.

    (2007)
  • J.C. Hwang et al.

    Predictive value of fundus autofluorescence for development of geographic atrophy in age-related macular degeneration

    Invest. Ophthalmol. Vis. Sci.

    (2006)
  • Keilhauer, C.N., Delori, F.C., 2008. Mapping the junctional zone of geographic atrophy with short wavelength and...
  • H. Kolb

    Anatomical pathways for color vision in the human retina

    Vis. Neurosci.

    (1991)
  • J. Kong et al.

    Correction of the disease phenotype in the mouse model of Stargardt disease by lentiviral gene therapy

    Gene Ther.

    (2008)
  • Cited by (51)

    • Quantifying the Rate of Ellipsoid Zone Loss in Stargardt Disease

      2018, American Journal of Ophthalmology
      Citation Excerpt :

      The high intragrader and intergrader ICC suggests that highly precise measurements can be obtained using the area of EZ loss, and indicates that this method can provide a high degree of confidence in defining disease progression. Prior studies of STGD progression based on area of RPE atrophy range from 0.28 mm2/year to 1.58 mm2/year.27,28,30,31,44 Testa and associates estimated the rate of RPE loss as seen on OCT in a heterogeneous group of STGD patients using an automated segmentation program to be around 0.28 mm2/year.44

    View all citing articles on Scopus
    View full text