Read about the technical details for the AI-based algorithm and smartphone retinal camera system.
In collaboration with BRAC University, we have published a validation study of the performance of our neural network on Bangladeshi Eyes. Our paper is available on ArXiv here, and our code is available on Github here. The paper is also currently in review for a top journal.
We use a DenseNet-121 state-of-the-art image classification model to perform diagnosis based retinal scans. We train our algorithm on a publically available dataset and perform grade-based screening. A diagram of the DenseNet convolutional block with our retinal scans is shown below. Our paper goes in-depth into the architectural details and training procedures.
This mobile, on-the-go system is designed for clinics in Bangladesh to screen patients for Diabetic Retinopathy (DR) using a smartphone camera with a retinal attachment. The purpose of this rig is to allow precise positioning of the smartphone to any patient's left and right eye such that the images can be efficiently fed into Drishti's AI algorithms for DR diagnosis. The system is completely adjustable for all head sizes. It is made of readily available components that can be purchased at many local hardware stores, and is designed for low-cost fabrication. All aseembly tools are common household tools or easily purchasable/rentable from a local hardware store. The 3D-printed components can be printed on low-end machines and with cheap PLA filament. We designed this system to be completely collapsable, such that it can fit into a standard size backpack.
All the CAD files were designed in SOLIDWORKS and are provided both as .sldprt files and as .STL file types, which are viewable directly in github using the embedded 3D model viewer. To view the final assembled CAD model, view Assembly.STL. For specifics on which components to purchase, and which to 3D print, refer to the Assembly section.
Interactive CAD Visualization
Initial Prototype Sketches