Extended Data Fig. 7: Image restoration and segmentation inaccuracy in relation to biological motion.

Overlaid manual segmentations of individual dendritic spines from paired LIONESS measurements in neuropil of living organotypic hippocampal slice cultures recorded across n = 2 independent samples, excluding or allowing biological motion between measurements. We simultaneously generated two LIONESS measurements of the same imaging volume at time point T_{0 min} by dialing in 20 µs dwell time at each voxel and assigning photons from the first 10 µs to measurement 1 (green 3D reconstructions of individual spines) and those from the second 10 µs to measurement 2 (blue reconstructions of the same spines), such that biological motion was excluded for these two simultaneous measurements. Image restoration was applied to the raw data, selected spines were manually segmented and their 3D reconstructions were overlaid. Asterisks are placed near the attachment site of spine necks to the respective parent dendrites. We then repeated this dual measurement 10 minutes later at time point T_{10 min}, again generating measurements 1 (orange reconstructions) and 2 (magenta reconstructions) for which biological motion was excluded. Finally, reconstructions corresponding to measurement 1 at the two different time points (T_{0 min}/T_{10 min}; green, orange) were overlaid by 3D-aligning spine necks to each other. Differences between the respective voxel exactly matched measurements 1 and 2 at single time points are attributable to measurement noise, inaccuracies in image restoration, and manual segmentation. Differences between the two time points (T_{0 min}/T_{10 min}) highlight changes in the biological structures. The scale cube refers to 500 nm in the center of the respective renderings.