Journal: Nature methods

Article Title: Mesoscale volumetric light field (MesoLF) imaging of neuroactivity across cortical areas at 18 Hz

doi: 10.1038/s41592-023-01789-z

Figure Lengend Snippet: (a) Left column: Standard deviation (SD) images of simulated LFM movie containing temporally active neurons (top) and a single plane from reconstructed SD image (bottom). The brightness values (i.e., magnitude of SD of their activity) of neurons in the SD image is chosen such that dimmest (least active) neuron is 8 times dimmer than the brightest (most active) one (indicated by yellow double arrow). Right column: Same as left column with brightness scaled 5× for clarity. (b) Result of morphological segmentation (Supplementary Note 6) applied to simulated data shown in (a). Individual segments highlighted in different colors. Dimmest neuron (yellow arow) is not segmented successfully and missed. (c) Result of NMF-based segmentation (Supplementary Note 7) applied to simulated data shown in (a). Individual segments highlighted in different colors. Dimmest neuron (yellow arow) is segmented successfully and detected. (d) Runtime comparison of NMF- and morphology-based methods. The morphology-based method is faster by orders of magnitude as the input frame side length increases. (e) Comparison between “ground truth” CaImAn-based segmentation of two-photon microscopy data (first column) to segmentation results obtained from morphological segmentation (second column), NMF-based segmentation (third column), and the combination of both (fourth column), for two different depths (170 μm, top row. 300 μm, bottom row). The purely morphological segmentation approach performs well in segmenting neurons but sometimes misses very dim neurons in the temporal summary image (highlighted with pink arrows). The NMF-based approach on the other hand tends to detect dim neurons more reliably (pink arrows) but comes with the risk of mis-segmenting close-by neurons and assigning them to one component only (yellow arrows). Combining both methods lead to the best results but incurs the largest computational cost of the three approaches since NMF has to be run on the full frames. It should therefore only be used if maximal detection sensitivity is required or if only a subset of the full MesoLF FOV is being used.

Article Snippet: Two-photon excitation pulses (920 nm, 140 fs pulse duration, 80 MHz pulse rate, Coherent Chameleon) were focused into mouse cortex and scanned in planes parallel to the cranial window at a series of depths, via the Slicescope’s galvo-galvo scan path and a Nikon 16×/0.8NA objective mounted on a motorized stage that allowed for axial translation.

Techniques: Standard Deviation, Activity Assay, Comparison, Microscopy