Fig. 4: Spatial resolution identified from neuronal responses supports optical modelling.

a Average spike counts per trial (background subtracted) for all tested μLEDs; mean of n = 13, n = 16 time-locked neurons; cortical surface irradiances given as pooled mean ± pooled standard deviation. b Neuronal response to μLEDs illuminated across individual rows, with Weighted Gaussian fits; error bars show pooled mean ± pooled SEM of average spike counts per trial (n = 13, n = 16 time-locked neurons). Cortical surface irradiances given as pooled mean ± pooled standard deviation for μLEDs across each row. c Cross-section from Monte Carlo simulation showing light propagation from μLED into brain tissue (acute device model with 7 μm parylene-C); contours show the 1 mW mm⁻² threshold for modelled cortical surface irradiances of 1.0 ± 0.3, 2.9 ± 0.5, 5.3 ± 0.7 mW mm⁻². Equivalent optical power delivered into the brain tissue was determined from optical modelling. d Comparison of spatial resolutions determined from the FWHM of weighted Gaussian fits in (b) (pooled mean ± pooled SEM of n = 29 time-locked neurons) and optical model in (c) for the highest tested power; taken as the width of the 1 mW mm⁻² threshold contour at a depth of 150 μm. e Same as (c) but μLED array modelled directly on brain surface (chronic device model with 15 μm parylene-C); cortical surface irradiances of 4.0 ± 1.1, 11.3 ± 2.0, 20.4 ± 2.8 mW mm⁻² with corresponding optical power/drive current same as (c). f Modelled comparison of spatial resolution (diamonds) and depth of penetration (squares) of 1 mW mm⁻² contour with and without a cortical window. Error bars arise from shaded regions in (c) and (e). g Summation of Monte Carlo simulation results for multiple simultaneously illuminated μLEDs, delivering the same optical power/drive current per μLED as in (c) and (e).