Not seeing is not believing: improving the visibility of your fluorescence images

Abstract

The digital age has brought both technical advances and ethical quandaries regarding data acquisition and image presentation in the field of cell biology. Image manipulation has drawn considerable attention in the past decade, leading to general guidelines for ethical data processing. However, effective methods of image presentation have been discussed only cursorily and have been largely overlooked. Under standard viewing conditions, the human visual system imposes limitations for readers analyzing fluorescence images. In this paper, I discuss the advantages and limitations of image-manipulation techniques with respect to the human visual system, including contrast stretching, nonlinear grayscale transformations, and pseudocoloring. While online data viewing presents innovative ways to access image information, most images continue to be viewed in static publications, in which image presentation is critical for effective information transmission.

FIGURE 1:

Contrast stretch and nonlinear (power-law) transformations. Original (A) and scaled (B–F) images of GFP-v-SNARE in yeast cells responding to mating pheromone. Autoscale contrast stretch (B) redistributes pixel values across the whole display range, without losing spatial information. Contrast can be stretched further (C) by setting an intermediate gray value (B, arrowhead) as white, but gray values in the range above the upper limit are clipped (C, asterisks), losing spatial information in the brightest regions. The nonlinear power-law transformation (D and E) redistributes gray values according to a logarithmic formula with exponent γ, resulting in increased contrast for a subset of gray values. Inversion (F) of an autoscaled image (as in B) displays dark values as light and vice versa without altering the distribution of pixel intensities.

FIGURE 2:

Color-coded contrast and pseudocoloring. Color-coded contrast (A) increases visual sensitivity to shallow contrast by representing gray values as varying color hues, according to an arbitrary color LUT (B). Pseudocoloring (D–F) applies a single-hue LUT to a grayscale image (C), resulting in gray levels represented by color brightness. The original image is identical to the power-law transformation (γ = 3.0) from Figure 1.