Home▸Medicine & Biophysics▸AI Diabetic Retinopathy Screening: Retinal Photo to Risk Grade

👁️ AI Diabetic Retinopathy Screening: Retinal Photo to Risk Grade

How transfer learning, specialised image preprocessing and explainability tools like Grad-CAM combine to grade diabetic retinopathy severity from retinal photographs, and what it would take to support the NHS Diabetic Eye Screening Programme.

Medicine & Biophysics3DModerate60 FPS

Details

CategoryMedicine & Biophysics

RenderingThree.js · WebGL

DifficultyModerate

Target FPS60

ControlsGround-truth DR stage slider (0-4) sets how many microaneurysms, haemorrhages, hard exudates and neovascular tufts appear on the retina; Raw / Ben Graham+CLAHE buttons toggle preprocessing contrast and detection rate; Grad-CAM overlay button highlights the lesions the model actually detected; Auto-rotate button spins the camera around the fundus.

Three.jsmedical imagingdiabetic retinopathyGrad-CAMInstancedMesh

📖 Read the articleThe maths behind AI Diabetic Retinopathy Screening: Retinal Photo to Risk Grade

Related simulations

Cardiac Action PotentialA FitzHugh-Nagumo model of cardiac tissue: click to stimulate regions and watch…

Drug Diffusion & PharmacokineticsA two-compartment PK model: compare IV bolus, oral and infusion dosing, plot plasma/tissue…

Blood Flow & Vessel MechanicsA pulsatile Poiseuille flow model: adjust vessel radius, stenosis, pressure gradient and… ⊞ All 274 in Medicine & Biophysics →

⚙ Under the hood

Grade a simulated retinal photograph for diabetic retinopathy: pick the true DR stage, toggle Ben Graham + CLAHE preprocessing, and watch a Grad-CAM heatmap show which microaneurysms, haemorrhages and exudates an AI model actually detects.

Three.jsmedical imagingdiabetic retinopathyGrad-CAMInstancedMesh

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install