Same accumulation-vs-clearance biology as the 3D drusen simulator, computed a completely different way: instead of tracking a list of discrete spheres, this version simulates a continuous concentration field c(x,y) — local deposit volume per unit area — on a grid under the macula, evolved by a reaction–diffusion PDE:
∂c/∂t = s(x,y) − k_clear(age)·c + D·∇²c
k_clear(age) = k_clear0 · exp(−(age−50)/30)
s(x,y) = k_dep / (cell area) at nucleated cells, else 0
Every grid cell obeys the identical linear ODE as a single drusen in the 3D model (deposition rate in, clearance proportional to local amount out); the reaction term is integrated with the exact solution c(t) = css + (c₀−css)·e−k·t each frame, so it stays stable at any simulation speed. The diffusion term D∇²c is new physics this representation adds: it lets a nucleus spread laterally into neighboring cells, so two nearby deposits merge into one continuous region purely because their fields overlap — no explicit "distance < sum of radii" merge test is needed, coalescence just falls out of the field.
- En-face heatmap (top) — a top-down concentration map, like the en-face view of an OCT scan, colored from the same size legend as the 3D model.
- Radial cross-section (bottom) — the field's thickness profile along one diameter through the macula's center, the 2D analogue of the 3D model's bulging RPE surface.
- Classification — every few frames the field is scanned with flood-fill connected-component labeling; each connected region's total volume (Σc·cellArea) is converted back to an equivalent-sphere diameter with the same r=(3V/4π)^(1/3) formula used in 3D, then graded small/intermediate/large per AREDS.