Microscopy figure maker: merge and split channels

$$d_c=\frac{v_c-{\text{min}}_c}{{\text{max}}_c-{\text{min}}_c},\text{ clipped to }[0,1]$$$${\text{RGB}}_{\text{merge}}=min(255,\sum _c{d}_c{\text{RGB}}_c)$$$$\begin{array}{rl}& \text{bar on the page}\\ & =\frac{\text{bar (µm)}}{\text{field width (µm)}}\times \text{panel width}\end{array}$$$$\text{zoom}=\frac{\text{inset width}}{\text{box width on the panel}}$$

Display range. Each channel keeps its raw values (8, 16 or 32 bit). For the figure, a value $v$ is mapped linearly between the channel’s minimum (black) and maximum (full colour): $d=(v-\text{min})/(\text{max}-\text{min})$, clipped to 0–1. The same mapping is applied to every pixel of the image, which is what image-integrity guidelines allow: brightness and contrast may be adjusted if the change is applied to the whole image and does not hide or create features (Rossner & Yamada 2004; Cromey 2010). The automatic range sets the minimum and maximum at the 0.1st and 99.9th percentiles of the channel’s pixels; the tool shows how many pixels end up at black and at full colour, so saturated structures are visible. Use the same range for images you compare. Gamma ($d^{\gamma }$) is a non-linear change: it is hidden under its own heading, and when used the legend text says so, as journals require.

Colours and the merge. Each channel is shown in one pseudocolour, and the merge adds the channels’ colours pixel by pixel, clipped at 255 (an additive overlay, as in ImageJ’s composite). Red and green are hard to tell apart for about 8 % of men; magenta and green, which add to white where they overlap, stay distinct (Wong 2011). Single channels can be shown in grey, which the eye reads with more contrast than a dark pure colour (Jambor et al. 2021). Channel names are written in their channel’s colour, so the labels double as a colour key; pure blue is lightened for the text only, because blue on black is hard to read.

Layout and size. The figure is laid out in points (1/72 inch) across the width you pick: panels are $(W-(n-1)\times \text{gap})/n$ wide in a row, the image’s shape is kept, and the panel letters, names and scale bar labels keep their sizes in points whatever the image’s pixel count. Nature asks for text between 5 and 7 pt in final figures; the defaults are 7 pt labels and 8 pt bold letters. The PNG is drawn at the DPI you choose (image pixels averaged when the panel is smaller than the image, interpolated when larger) and records that DPI; the SVG embeds each panel as an image at that resolution and writes letters, names, scale bar labels and text as text, the bars, boxes, lines and arrows as shapes.

Scale bar and zooms. With the pixel size $p$ (µm per pixel) the field is $W_{\text{px}}\times p$ µm wide and the automatic bar is the $1,2\text{ or }5\times {10}^n$ µm length covering 15–25 % of it (the same rule as the scale bar adder). A zoom inset shows the boxed region enlarged by inset width ÷ box width on the panel; its pixels are enlarged by nearest neighbour, so they stay visible as pixels rather than being smoothed, and it gets its own scale bar from the box’s width in µm. Connector lines join the matching corners of the box and the inset, the two that pass outside both.

Sources: Jambor H, Antonietti A, Alicea B, et al. Creating clear and informative image-based figures for scientific publications. PLoS Biol 2021;19:e3001161, doi:10.1371/journal.pbio.3001161. Schmied C, Nelson MS, Avilov S, et al. Community-developed checklists for publishing images and image analyses. Nat Methods 2024;21:170–181, doi:10.1038/s41592-023-01987-9. Wong B. Points of view: Color blindness. Nat Methods 2011;8:441, doi:10.1038/nmeth.1618. Rossner M, Yamada KM. What’s in a picture? The temptation of image manipulation. J Cell Biol 2004;166:11–15, doi:10.1083/jcb.200406019. Cromey DW. Avoiding twisted pixels: ethical guidelines for the appropriate use and manipulation of scientific digital images. Sci Eng Ethics 2010;16:639–667, doi:10.1007/s11948-010-9201-y.

Three 16-bit channels, 800 × 800 pixels each, taken with a 6.5 µm camera pixel behind a 40× objective: $6.5÷40=0.1625$ µm per pixel, so the field is $800\times 0.1625=130$ µm wide. The files are named cells_DAPI.tif, cells_WGA.tif and cells_LC3.tif.

1. Names and colours. The common part “cells_” is dropped, leaving DAPI, WGA and LC3. DAPI is recognised as a nuclear stain and gets cyan; the other two get magenta and green in turn.
2. Display range. With DAPI’s range at 180–2400, a pixel of 1290 is shown at $(1290-180)/(2400-180)=0.5$ of full cyan: RGB (0, 128, 128). Pixels at or below 180 are black, at or above 2400 full cyan.
3. Layout. 1 × 3 + merge across 183 mm with 1 mm gaps: $(183-3\times 1)/4=45$ mm per panel, square like the image; with room for the 8 pt letters above, the figure is 48.5 mm tall. At 300 dpi each panel is 531 pixels wide, so the 800-pixel images are averaged down.
4. Scale bar. 15–25 % of 130 µm is 19.5–32.5 µm; the only $1,2,5\times {10}^n$ value in that range is 20 µm, drawn $20/130\times 45=6.92$ mm long in the merge’s bottom-right corner.
5. Zoom. A 120 × 120 pixel box ($120\times 0.1625=19.5$ µm) on the merge, shown at 42 % of the panel width ($0.42\times 45=18.9$ mm) in the top-right corner: on the panel the box is $120/800\times 45=6.75$ mm, so it is enlarged $18.9/6.75=2.8$×. Its own scale bar is 5 µm (the 1–2–5 length nearest 20 % of 19.5 µm).

The legend text the tool writes for this figure: “DAPI (cyan), WGA (magenta), LC3 (green). Scale bar, 20 µm; zoom, 5 µm. Boxed region shown enlarged (2.8×). Display ranges were set linearly per channel and applied to the whole image (…).”