Visual Impairment Screening using Images from Ophthalmology and Novel pathways for Structural Analysis and Fast Evaluation

Clinical eye images could be explored in an entirely new way to provide early detection of serious eye diseases that currently affect thousands worldwide yet could be preventable.

The objective of VISIONSAFE is to extract macro- and microstructure biomarkers from real-world clinical images depicting the retina, lens, and cornea, thereby facilitating early disease detection and personalised management.

While macrostructural biomarkers related to ocular shape are well-established, microstructural biomarkers related to tissue remain largely unexplored.

The project's foundation rests on the preliminary work of the PI, which demonstrated the feasibility of deriving disease-related microstructure biomarkers from low-resolution clinical eye images.

The approach combines data science, advanced image processing, and rigorous experimental validation to better understand the interplay between macro- and microstructural biomarkers in prevalent sight-threatening eye diseases, including age-related macular degeneration (AMD), glaucoma, cataracts, and corneal ectasia, using efficient, cost-effective, and accessible tools.

VISIONSAFE covers several ground-breaking aspects: clinical eye data is already available but underused (can we use data science for vision preservation?); the acquisition and combination of macro- and microstructural biomarkers from clinical images have not yet been considered (is this combination a key for the early diagnosis of eye diseases?

Are we missing out essential information from already available data?). The project's potential gains are significant.

It enhances knowledge on eye disease biomarkers accessible through routine eye exams, shifts the paradigm toward macro- and microstructural biomarker synergy, and represents an important step towards a future where eye diseases are identified and managed earlier, offering hope and improved quality of life to countless individuals.