Full write-up coming soon — detailed content for this project is being finalized.
Overview
Silicosis is a preventable lung disease common among miners and quarry workers, but early-stage cases are easy to miss without a specialist radiologist on hand. This project screens chest X-rays for early nodular patterns directly on portable imaging hardware, so a result is available at the point of care rather than after a referral.
Approach
The model is a compact convolutional network distilled from a larger radiology model, trained on a mix of public chest X-ray datasets and site-specific imaging to account for hardware variation across portable X-ray units.
Results
In early field testing across partner clinics, the model flagged early-stage cases later confirmed by specialist review, while keeping the false-positive rate low enough to avoid overwhelming referral capacity.