One portable workflow. Three deliberate steps.
The hardware and app remain in development. This is the intended sequence, not a claim of clinical readiness.
- 01
Attach
Align a 3D-printed clip and 90D aspheric lens with a compatible smartphone camera.
- 02
Capture
Use guided image capture to obtain a retinal image suitable for model evaluation.
- 03
Interpret
Review a proposed EfficientNet-B3 severity output with a SHAP explanation and a path to professional follow-up.
See what explainable AI is meant to show.
Drag the control to compare a synthetic retinal illustration with an example attention overlay. It demonstrates the interface idea; it is not a scan, model output, lesion localization, or diagnostic result.
Explore the overlay
SHAP heatmaps can indicate image regions that influenced a model’s output. They do not establish that a highlighted region is a lesion or explain a clinical diagnosis on their own.
See the full walkthrough ↗Help shape a clearer first step.
We welcome conversations with clinicians, researchers, clinics, NGOs, and people who want to follow the project.
Both actions open a contact form that prepares an email draft; this site does not store submissions or operate an automated waitlist.
