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.

Electrophysiological priming effects demonstrate independence and overlap of visual regularity representations in the extrastriate cortex

An Event Related Potential (ERP) component called the Sustained Posterior Negativity (SPN) is generated by regular visual patterns (e.g. vertical reflectional symmetry, horizontal reflectional symmetry or rotational symmetry). Behavioural studies suggest symmetry becomes increasingly salient when the exemplars update rapidly. In line with this, Experiment 1 (N = 48) found that SPN amplitude increased when three different reflectional symmetry patterns were presented sequentially. We call this effect ‘SPN priming’. We then exploited SPN priming to investigate independence of different symmetry representations. SPN priming did not survive changes in retinal location (Experiment 2, N = 48) or non-orthogonal changes in axis orientation (Experiment 3, N = 48). However, SPN priming transferred between vertical and horizontal axis orientations (Experiment 4, N = 48) and between reflectional and rotational symmetry (Experiment 5, N = 48). SPN priming is interesting in itself, and a useful new method for identifying functional boundaries of the symmetry response. We conclude that visual regularities at different retinal locations are coded independently. However, there is some overlap between different regularities presented at the same retinal location.

Spatial Heterogeneity in Bistable Figure-Ground Perception

The appearance of visual objects varies substantially across the visual field. Could such spatial heterogeneity be due to undersampling of the visual field by neurons selective for stimulus categories? Here, we show that which parts of a bistable vase-face image observers perceive as figure and ground depends on the retinal location where the image appears. The spatial patterns of these perceptual biases were similar regardless of whether the images were upright or inverted. Undersampling by neurons tuned to an object class (e.g., faces) or variability in general local versus global processing cannot readily explain this spatial heterogeneity. Rather, these biases could result from idiosyncrasies in low-level sensitivity across the visual field.

A new ERP repetition paradigm to assess independence of regularity representations in the extrastriate cortex.

An ERP component called the Sustained Posterior Negativity (SPN) indexes the brain response to regularity in visual patterns (e.g. vertical reflectional symmetry, horizontal reflectional symmetry or rotational symmetry). However it is unclear if different regularities are coded by independent or overlapping neural populations. Previous work has successfully exploited repetition effects to assess representational independence, and we adapted this approach to investigate visual regularity. Experiment 1 (N = 48) found that SPN amplitude increased when three reflectional symmetry patterns were presented sequentially. This SPN priming effect did not survive changes in retinal location (Experiment 2, N = 48) or non-orthogonal changes in axis orientation (Experiment 3, N= 48). However, SPN priming transferred between vertical and horizontal axis orientation (Experiment 4, N= 48) and between reflectional and rotational symmetry (Experiment 5, N= 48). We conclude that visual regularities at different retinal locations are coded independently. However, there is some overlap between different regularities presented at the same retinal location.

Bistable Perception Is Biased by Search Items but Not by Search Priming

During visual search, selecting a target facilitates search for similar targets in the future, known as search priming. During bistable perception, in turn, perceiving one interpretation facilitates perception of the same interpretation in the future, a form of sensory memory. Previously, we investigated the relation between these history effects by asking: can visual search influence perception of a subsequent ambiguous display and can perception of an ambiguous display influence subsequent visual search? We found no evidence for such influences, however. Here, we investigated one potential factor that might have prevented such influences from arising: lack of retinal overlap between the ambiguous stimulus and the search array items. In the present work, we therefore interleaved presentations of an ambiguous stimulus with search trials in which the target or distractor occupied the same retinal location as the ambiguous stimulus. Nevertheless, we again found no evidence for influences of visual search on bistable perception, thus demonstrating no close relation between search priming and sensory memory. We did, however, find that visual search items primed perception of a subsequent ambiguous stimulus at the same retinal location, regardless of whether they were a target or a distractor item: a form of perceptual priming. Interestingly, the strengths of search priming and this perceptual priming were correlated on a trial-to-trial basis, suggesting that a common underlying factor influences both.