Colour blindness (CB) comprises a number of conditions involving one or more retinal visual pigments which are missing or defective. The prevalence of CB has been reported to be in the range 6-10% of males. Among these conditions, anomalous trichromatic vision occurs when one of the three colour receptors in the retina has altered spectral sensitivity. In this context, protanomaly, deuteranomaly and tritanomaly imply reduced sensitivity to red, green and blue respectively. Individuals with dichromatic vision instead have only 2 out of the 3 pigments functional and the conditions are known as protanopia, deuteranopia and tritanopia depending on whether the red, green or blue pigments are affected. Protanope and deuteranope individuals (also known as red-green CB) confuse red and green hues, while tritanopes confuse hues in the yellow-blue range. In the realm of bioimaging, the perception of some types of imagery can be difficult for CB individuals. It therefore seems important to devise and implement remedial visualisation procedures to facilitate perception of CB and enable access (or facilitate education) to otherwise problematic imagery without detriment to trichromatic (normal) observers. It is well known that two-channel confocal microscopy images, are better perceived by red-green CB if channel data is encoded as magenta-green images rather than red-green pairs, however bright field microscopy images of stained samples cannot be subjected to that type of encoding. The most common stain used in routine histopathology and histology teaching is Haematoxylin & Eosin (H&E) and for CB observers it can be difficult or impossible to discriminate structures stained with one or another dye, or to detect areas where the dyes colocalise. To resolve this, we exploit the fact that such images contain mainly two dye hues, so they can be processed to create artificially coloured images that can be perceived by the available retinal sensors of the observer. We identified two methods of enhancing images of H&E stained sections for red-green CBs that in addition remain visible and acceptable to people with normal vision. The first method consists of encoding the image RGB data into hue, saturation and brightness colour space, identify the range of hues of the H&E dyes and rotate and stretch the hue range to span the perceptual range of the CB observers. The second method utilises digitally unmixing of the original dyes by means of colour deconvolution, to then reposition the dye information separately into hues that are more distinctly perceived. Images created using these methods were tested in a number of volunteers with normal vision and with red-green colour blindness. The CB subjects reported they could perceive better the differently stained structures in the enhanced versions compared to the original H&E stained images. Interestingly many subjects with normal vision also preferred the enhanced images to the original H&E. In addition, the analysis of the perceived hue ranges in red-green colour blinds suggested construction principles for the design of sequential, diverging and qualitative look up tables for monochromatic images. This in turn enables designing false colour images safely without ambiguous colour mappings for dichromatic observers.
colour blindness, microscopy, haematoxylin & eosin, look up tables
Presenting author: Gabriel Landini
Organisation: School of Dentistry, University of Birmingham, UK
co-authors: D. Giles Perryer