Journal: eLife

Article Title: Neural control of growth and size in the axolotl limb regenerate

doi: 10.7554/eLife.68584

Figure Lengend Snippet:

Article Snippet: Software, algorithm , ZenPro software , Zeiss , 410138-1104-260 , .

Techniques: In Situ, Software, Plasmid Preparation, Modification

Journal: Scientific Reports

Article Title: Circulating tumor DNA as a prognostic indicator in resectable pancreatic ductal adenocarcinoma: A systematic review and meta-analysis

doi: 10.1038/s41598-019-53271-6

Figure Lengend Snippet: Characteristics of studies selected for analysis.

Article Snippet: Hadano et al . , 2007–2013 , 105 , Baseline , 31.43% , 14–96mo (mean 54mo) , 13.6mo/27.6mo , Plasma , ddPCR , Bio-Rad QX100 , KRAS - G12D - G12V - G12R , 69 , 52.4 , Asian , 81.9% (86/105).

Techniques: Mutagenesis, Clinical Proteomics, Digital PCR, Amplification

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: “Temporally interleaved” 2pM–MesoLF functional ground truth generated by interleaving volumetric MesoLF frames and planar 2pM frames. Data in a–g is based on a total of 34 recordings from 3 mice. (a) Left column: Ground truth (blue circles) and MesoLF-extracted neuron positions (red circles) overlaid on a 2pM temporal standard deviation image from “temporally interleaved” 2pM–MesoLF recording, for two different depths (top: 150 μm; bottom: 300 μm). Full 2pM FOV of 280 × 280 μm is cropped for clarity. Middle column: MesoLF-extracted neuron positions (red circles) overlaid on a slice from the reconstructed MesoLF temporal summary volume corresponding to the depth of the 2pM plane shown in left column in the same hybrid 2pM–MesoLF recording as in left column. Right column: Neuronal activity traces corresponding to circles in left and middle column panels, as used for performance quantifications, in experimental functional ground truth (blue traces, corresponding to blue circles in middle column panel), recorded by standard 2pM, analyzed using CaImAn followed by human annotation, and simultaneously acquired LFM recordings, analyzed using MesoLF (red traces, corresponding to red circles in middle column panel), for same two depths as in left column. Solid lines: denoised, shaded lines: raw. (b) Neuron detection scores precision, sensitivity and F-score achieved by MesoLF on experimental “temporally interleaved” functional verification dataset as a function of depth. Shaded areas: mean ± SD; data from n = 34 recordings, each containing 45 ± 24 (mean ± SD) ground truth active neurons in 280 × 280 μm planar 2pM FOV. (c) Distributions of lateral neuron localization errors between MesoLF-extracted neuron positions and experimental functional “temporally interleaved” ground truth. White circle: median. Thick grey vertical line: Interquartile range. Thin vertical lines: Upper and lower proximal values. Transparent blue disks: data points. Transparent violin-shaped area: Kernel density estimate of data distribution. n = 1146 neuron pairs. (d) Distributions of temporal correlations between experimental “temporally interleaved” ground truth activity traces and matched MesoLF traces versus depth. Violin plot elements as in c. n = 1146 neuron pairs. (e) Mean pairwise correlation between all pairs of traces in “temporally interleaved” experimental functional ground truth (blue line) and mean pairwise correlation between corresponding pairs of MesoLF-extracted traces (red line) as a function of lateral distance between the neurons in the pairs. No substantial change in excess correlation, i.e., in the difference between MesoLF-extracted correlation and ground truth correlation, is observable across all pair distances. Shaded areas: Mean ± SD. (f) Distributions of excess correlation between pairs of neuronal traces in experimental “temporally interleaved” ground truth and corresponding pairs of MesoLF-extracted traces, as a function of depth. White circle: median. Thick grey vertical line: Interquartile range. Thin vertical lines: Upper and lower proximal values. Violin-shaped area: Kernel density estimate of data distribution. n = 50,904 neuron pairs. (g) Transient extraction scores precision, sensitivity and F-score achieved by MesoLF on experimental “temporally interleaved” functional verification dataset as a function of depth. Transients identified by human annotation of a random subset of n = 136 paired neuron traces (150 s), both in the MesoLF and the 2pM data. Random subset from a total population of 1100 paired neuron traces from 34 recordings across depths 50–400 μm. Individual recordings are 600 seconds long, contain 45 ± 24 (mean ± SD) ground truth active neurons and 291 ± 170 ground truth transients).

Article Snippet: Mesoscale high-speed volumetric functional imaging in mouse Here we demonstrate a modular Mesoscale Light Field (MesoLF) imaging hardware and software solution that combines mesoscale optical design and aberration correction with a scalable computational pipeline for neuronal localization and signal extraction.

Techniques: Functional Assay, Generated, Standard Deviation, Activity Assay, Extraction

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) Illustration of MesoLF light field phase space reconstruction with background peeling. (b) Comparison of volumetric light field reconstruction methods and ground truth. Top left: simulated ground truth volume containing neurons, blood vessels, and neuropil. Top right: Volumetric reconstruction of simulated LFM raw data using Richardson-Lucy deconvolution (Pixel space). Bottom left: Reconstruction using phase space deconvolution without background peeling (Phase space). Bottom right: Reconstruction using phase space deconvolution with background peeling (MesoLF). Red arrows: positions where artefacts are present in other methods but absent in MesoLF reconstruction. Yellow arrows: position where a ground truth neuron was falsely suppressed in MesoLF. All panels: large image is x-y slice at z = 60 μm. Smaller images are maximum intensity projections of the reconstructed volume along the x and y axes, respectively. Simulated depth of center of volume: 60 μm. Size of volumes: 600 × 600 × 200 μm3, depth range 0–200 μm (c) Structural similarity index between the simulated ground truth volume and the three different classes of reconstructed volumes shown in b, quantifying quality of reconstruction. n = 9 sets of reconstructions. Paired two-sided Wilcoxon signed rank test for equal median. p = 0.004 (pixel space vs. phase space), 0.004 (pixel space vs. MesoLF), 0.004 (phase space vs. MesoLF). ** p < 0.01. (d) Violin plot of 3D localization error, defined as minimum 3D distance between neurons in simulated ground truth and neurons found in the three different reconstructions shown in b. White circle: median. Thick grey vertical line: Interquartile range. Thin vertical lines: Upper and lower proximal values. Transparent blue disks: data points. Transparent violin-shaped area: Kernel density estimate of data distribution. n = 60, 79, 94 data points, respectively. Two-sided Wilcoxon rank sum test for equal medians, p = 0.567 (pixel space vs. phase space), 0.003 (pixel space vs. MesoLF), 0.019 (phase space vs. MesoLF). n.s., not significant. * p < 0.05, ** p < 0.01. (e) Violin plot of lateral localization error, defined as minimum lateral distance between neurons in simulated ground truth and neurons found in the three different reconstructions shown in b. Symbols as in d. n = 65, 87, 104 data points, respectively. Two-sided Wilcoxon rank sum test for equal medians, p = 0.766 (pixel space vs. phase space), 0.009 (pixel space vs. MesoLF), 0.029 (phase space vs. MesoLF). n.s., not significant. ** p < 0.01. Violin plot elements as in d. (f) Segmentation performance in MesoLF. Background: Slice from volume reconstruction of temporal summary image, SomaGCaMP7f, mouse cortex, depth 100 μm, simulated data. Colored circles indicate MesoLF segmentation results compared to manual segmentation. (g) Comparison of MesoLF segmentation performance versus PCA/ICA-based segmentation for four simulated neurons with highly correlated temporal activities (activity traces shown above segmented images). Ground truth neurons and corresponding time traces labelled with black digits. Individual segments shown as contour lines with different colors. Note the overlapping and under-segmented output from PCA/ICA. (h) Overall neuron detection scores for the MesoLF morphological segmentation compared to the CNMF-E initial segmentation phase (template matching and shape-based selection steps) (simulated, SomaGCaMP7f, mouse cortex, depth 100 μm). Height of bars: Mean. Error bars: SD. Black circles: n = 5 simulation runs. (i) Illustration of core-shell geometry for demixing neuropil activity from soma signals. Signals from segmented regions in f (cores, neurons) and a Gaussian shell region (extending from ~10 to ~20 μm diameter) surrounding the cores (background shell, neuropil) are identified and demixed. (j) Sets of representative example traces for core, shell, and demixing result, taken from a recording in mouse cortex at depths 200–400 μm. Arrows indicate crosstalk between shell and core that is removed in the demixed traces. Experimental data from SomaGCaMP7f-labelled mouse cortex, depth 100 μm. (k) Matrices of Pearson correlation coefficients between 400 pairs of neuronal activity traces extracted from a MesoLF recording in mouse cortex, before and after core-shell demixing. Average absolute correlation between signal pairs is reduced by 37% in MesoLF. (l) Illustration of convolutional neural network (CNN) architecture used for classification of candidate neural activity traces (m) Representative examples of 25 kept and 10 rejected traces by CNN. Experimental data, SomaGCaMP7f, mouse cortex, various depths. (n) Classification performance of two differently trained CNNs, one optimized for prioritizing precision (“precise mode”, blue bars) and one for prioritizing sensitivity (“sensitive mode”, violet bars), both while maintaining an overall high F-score. The CNN in “precise” mode achieves precision 0.98 ± 0.01, sensitivity 0.60 ± 0.03, F-score 0.75 ± 0.02; CNN “sensitive” mode achieves precision 0.90 ± 0.02, sensitivity 0.96 ± 0.01, F-score 0.93 ± 0.01. Classification performance evaluated on withheld data that was not used during training. Height of bars: Mean. Error bars: SD. Black circles: n = 5 held-out datasets in each bar. Boxed insets: Miniature representations of the pipeline schematic shown in Fig. 1, for orientation.

Article Snippet: Mesoscale high-speed volumetric functional imaging in mouse Here we demonstrate a modular Mesoscale Light Field (MesoLF) imaging hardware and software solution that combines mesoscale optical design and aberration correction with a scalable computational pipeline for neuronal localization and signal extraction.

Techniques: Comparison, Activity Assay, Selection

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) Comparison of segmentation performance of MesoLF versus CNMF-E (template matching and shape-based selection steps) in a 2D slice from a MesoLF recording in mouse cortex, depth 100 μm. Green circles: segments that strongly match with the ground truth. Blue circles: segments that only appear in the ground truth. Magenta circles: segments that are not consistent with ground truth. (b) Comparison of precision, sensitivity, and F1-scores for neuron detection performance in CNMF-E (template matching and shape-based selection steps) and MesoLF segmentation. Same data as in main Fig. 3h, reproduced here for convenience. Height of bars: Mean. Error bars: SD. Black circles: n = 5 simulation runs. (c) Top panel: Illustration of 3D volume containing neurons and exhibiting scattering, as used for volumetric segmentation comparisons in remainder of figure. Schematic illustration of segmentation pipelines in CNMF-E (middle box) and MesoLF (bottom box). (d) 3D rendering of segmentation results from CNMF-E (left) and MesoLF (right). Magenta: Ground truth neurons, green: segments. (e) Zooms into areas indicated by dashed rectangles in (d). (f) Comparison of the spatial similarity index of neurons paired between ground truth and output of CNMF-E (template matching and shape-based selection steps) versus MesoLF segmentation. p = 2.0e-9, paired one-sided Wilcoxon signed rank test. n = 63 neuron pairs. ** p < 0.01. (g) Histogram of spatial similarity indices of segmented neurons compared to ground truth by both methods (same data as in (f)).

Article Snippet: Mesoscale high-speed volumetric functional imaging in mouse Here we demonstrate a modular Mesoscale Light Field (MesoLF) imaging hardware and software solution that combines mesoscale optical design and aberration correction with a scalable computational pipeline for neuronal localization and signal extraction.

Techniques: Comparison, Selection