Early NTD-Related Cataract Detection

Problem

Cataract and related ocular presentations associated with neglected tropical disease are often caught late, especially where ophthalmology coverage is thin.

Context

Deep-learning computer vision for early detection, recognised in a UN award listed on the professional profile.

Dataset

Ocular imagery for supervised visual screening. Clinical dataset governance documented before public claims about data sources.

Methodology

Supervised visual screening models intended as a triage aid, not a stand-alone diagnosis.

Model / architecture

Deep convolutional models for early presentation detection.

Results

A screening hypothesis: that computer vision can extend scarce specialist attention rather than replace it.

Evaluation metrics

Domain

Ocular screening

Recognition · unverified

UN award

Challenges

Clinical image quality, demographic bias in training sets, and the ethical bar for any model that might influence a referral.

Impact

Extending scarce specialist attention through computer-vision triage in low-resource settings.

Tools / technology

• Python
• Deep learning
• Computer vision