Evaluation of focus and deep learning methods for automated image grading and factors influencing image quality in adaptive optics ophthalmoscopy.

Adaptive optics flood illumination ophthalmoscopy (AO-FIO) is an established imaging tool in the investigation of retinal diseases. However, the clinical interpretation of AO-FIO images can be challenging due to varied image quality. Therefore, image quality assessment is essential before interpretation. An image assessment tool will also assist further work on improving the image quality, either during acquisition or post processing. In this paper, we describe, validate and compare two automated image quality assessment methods; the energy of Laplacian focus operator (LAPE; not commonly used but easily implemented) and convolutional neural network (CNN; effective but more complex approach). We also evaluate the effects of subject age, axial length, refractive error, fixation stability, disease status and retinal location on AO-FIO image quality. Based on analysis of 10,250 images of 50 × 50 µm size, at 41 retinal locations, from 50 subjects we demonstrate that CNN slightly outperforms LAPE in image quality assessment. CNN achieves accuracy of 89%, whereas LAPE metric achieves 73% and 80% (for a linear regression and random forest multiclass classifier methods, respectively) compared to ground truth. Furthermore, the retinal location, age and disease are factors that can influence the likelihood of poor image quality.

Retinal Differential Light Sensitivity Variation Across the Macula in Healthy Subjects: Importance of Cone Separation and Loci Eccentricity.

Purpose: Microperimetry measures differential light sensitivity (DLS) at specific retinal locations. The aim of this study is to examine the variation in DLS across the macula and the contribution to this variation of cone distribution metrics and retinal eccentricity. Methods: Forty healthy eyes of 40 subjects were examined by microperimetry (MAIA) and adaptive optics imaging (rtx1). Retinal DLS was measured using the grid patterns: foveal (2°-3°), macular (3°-7°), and meridional (2°-8° on horizontal and vertical meridians). Cone density (CD), distribution regularity, and intercone distance (ICD) were calculated at the respective test loci coordinates. Linear mixed-effects regression was used to examine the association between cone distribution metrics and loci eccentricity, and retinal DLS. Results: An eccentricity-dependent reduction in DLS was observed on all MAIA grids, which was greatest at the foveal-parafoveal junction (2°-3°) (-0.58 dB per degree, 95% confidence interval [CI]; -0.91 to -0.24 dB, P < 0.01). Retinal DLS across the meridional grid changed significantly with each 1000 cells/deg2 change in CD (0.85 dB, 95% CI; 0.10 to 1.61 dB, P = 0.03), but not with each arcmin change in ICD (1.36 dB, 95% CI; -2.93 to 0.20 dB, P = 0.09). Conclusions: We demonstrate significant variation in DLS across the macula. Topographical change in cone separation is an important determinant of the variation in DLS at the foveal-parafoveal junction. We caution the extrapolation of changes in DLS measurements to cone distribution because the relationship between these variables is complex. Translational Relevance: Cone density is an independent determinant of DLS in the foveal-parafoveal junction in healthy eyes.