Automatic analysis of corneal nerves imaged using in vivo confocal microscopy

Abstract

Interest has grown over the past decade in using in vivo confocal microscopy to analyse the morphology of corneal nerves and their changes over time. Advances in computational modelling techniques have been applied to automate the estimation of sub-basal nerve structure. These objective methods have the potential to quantify nerve density (and length), tortuosity, variations in nerve thickness, as well as temporal changes in nerve fibres such as migration patterns. Different approaches to automated nerve analysis, methods proposed and how they were validated in previous literature are reviewed. Improved understanding of these approaches and their limitations will help improve the diagnostic leverage of emerging developments for monitoring the onset and progression of a broad class of systemic diseases, including diabetes.

An artificial intelligence-based deep learning algorithm for the diagnosis of diabetic neuropathy using corneal confocal microscopy: a development and validation study

“…The strength of deep learning is echoed by Oakley et al [44] who used corneal nerve segmentation in CCM images from macaques. Deep-learning-based approaches make the segmentation task relatively easier for the end user compared with the conventional approaches employing various filters and graphs [23]. In particular, compared with conventional machine-learning methods, such as support vector machines (SVM), deep learning reduces the need and additional complexity of feature selection and extraction, allowing the computer to learn features alongside the segmentation.…”

“…Kim and Markoulli [23] developed a nerve segmentation technique to delineate corneal nerve morphology in CCM. This involved processes ranging from filtering methods (with rapid implementation but low-contrast and imprecise focus) to more complex support vector machine approaches (which rely on features defined by the user).…”

Automated Tortuosity Analysis of Nerve Fibers in Corneal Confocal Microscopy

“…Existing clinical research [1], [2] suggests that the morphological changes observable in numerous anatomical curvilinear structures -e.g., retinal blood vessels, coronary blood vessel, carotid or corneal nerve fibers -are closely related to the presence of many diseases. For example, several studies [1]- [6] have confirmed that some quantitative properties of the corneal nerves such as nerve fiber branching, density, length, and tortuosity are linked to eye and systemic diseases.…”

“…Corneal nerve damage and repair from surgical interventions may be reflected in nerve branching, while fungal keratitis patients usually exhibit lower nerve density [7]. Tortuosity is one of the most significant biomarkers reflecting variations in corneal nerve fibers, as larger tortuosity implies process of nerve degeneration and subsequent regeneration, leading to active neural growth [2]. In addition, ophthalmologists often use the tortuosity of corneal nerve fibers as an important clinical parameter to assess hypertensive retinopathy [8] and diabetic neuropathy [9]- [12], respectively.…”

“…1 illustrates three examples of CCM images with increasing tortuosity. Clinicians have suggested that any measure of tortuosity should be invariant to translation, rotation and scaling [2]. An accurate measurement of tortuosity should be able to extract the key tortuosity characteristics -e.g., amplitude and number of inflection points (frequency) -and to overcome time constraints and inaccurate measurements.…”