Lumen

Our project was inspired by a staggering reality: over 537 million people globally live with diabetes, and a significant portion will face diabetic retinopathy, a condition leading to irreversible blindness. We realized that the issue is not the lack of treatment, but the lack of accessible diagnostics. Patients in remote regions often wait months for an ophthalmologist appointment, while every day of delay risks their vision. We set out to shift the paradigm from "clinic-based treatment" to "universal, on-demand screening." We designed the Lumen system to act as a "specialist in a box." Our workflow relies on three core pillars: Hardware, utilizing a specialized macro-lens attachment that turns a standard smartphone into a functional ophthalmoscope; an AI Core, where we developed a Convolutional Neural Network (CNN) architecture optimized for medical imaging. The model classifies retinal images by identifying pathological markers such as exudates and microaneurysms. Our current validation phase has achieved an accuracy rate exceeding 92%. The system is integrated with a cloud-based infrastructure, enabling real-time diagnostic feedback. Our primary challenge was cleaning our dataset from low-quality images (artifacts and lens glare), which we solved by implementing image pre-processing techniques, specifically CLAHE (Contrast Limited Adaptive Histogram Equalization), to enhance the visibility of retinal vessels. Additionally, we recognized that automated medical screening carries ethical risks, so we introduced a "Confidence Score." If the model’s certainty falls below a specific threshold, it triggers a referral to a human specialist, ensuring a robust Human-in-the-loop approach. We learned that AI in medicine is not a replacement for doctors, but a tool for Augmented Intelligence. We gained deep experience in working with sensitive medical datasets and realized how critical "explainability" is for end-user trust. Lumen is our contribution to ensuring that access to healthcare is no longer determined by geographical location or socioeconomic status. Lumen is the result of a diverse, remote-first team committed to global health equity.

Built With

• figma
• google-colab
• keras
• python
• tensorflow