From Digital pathology

Primary digital reporting in gastrointestinal pathology

The technology for producing digital scans of whole glass slides at high resolution has matured to a point where the user experience is at a level that could be compatible with using digital images as the primary reporting medium in histopathology. The quality of the scans is excellent, the user interface is getting better and the speed of transmission is good with virtually no lag on most systems. However there are two areas which need a lot more investigation before there is a wholesale transition to digital pathology – validation of the digital diagnoses against glass slide diagnosis and ‘time and motion’ studies to investigate how digital reporting affects workflow. A new study by Loughrey et al. in Virchovs Archiv provides some validation of digital reporting for gastrointestinal specimens. They took a sample of 100 consecutive gastrointestinal specimens in routine practice and three pathologists reported them as digital images and later reviewed them on glass slides. They found a concordance rate of 95.3% which is comparable with inter- and intra-observer agreement on pure glass slide studies. All the discordant cases were judged to be of minor clinical significance. In 10 of the discordant cases the glass slide diagnosis was favoured, in the other 4 the digital diagnosis was favoured. This study is a good start to producing enough validation of digital images to enable this to be the primary reporting medium but it should be recognised that 100 consecutive gastrointestinal specimens are unlikely to have many subtle diagnoses in them, they could be nearly all normal duodenal and colonic biopsies, so a high level of concordance would be expected using digital or glass slides.

Submucosal invasion in colorectal cancer

smHistopathology produces images which are basically two-dimensional sections across a three-dimensional object in a particular plane (and that plane is often fairly random). There is a long history of trying to reconstruct three-dimensional objects from histology images, I remember looking at a three-dimensional reconstruction of a bronchial mucus gland produced from polystyrene ceiling tiles by the late Prof. Bill Whimster – an early pioneer of these techniques. Advances in digital imaging and computing power have made three-dimensional reconstruction a more feasible proposition but so far it has rarely demonstrated anything that has relevance to diagnostic histopathology. In this month’s Histopathology Prof. Phil Quirke’s group at Leeds have produced results from three-dimensional reconstruction that could change our practice. The study used immunohistochemical staining of colon specimens to define the lymphovascular structures in the submucosa and they found that the density of lymphovascular vessels does not increase with increasing depth in the submucosa. This means that risk of metastases does not necessarily increase with deeper invasion (as embedded in some staging systems such Kikuchi’s) but it is more likely that total volume of submucosal invasion would be the best predictor of metastatic potential.

How do histopathologists look at slides?

When a histopathologist looks down a light microscopy we have very little idea of what different patterns of observing might happen. As trainees we probably did a lot of work on double-headed microscopes with experienced consultants so we would get some idea of the methods they used, I was always very impressed by how very good diagnostic pathologists could zoom in on a diagnostic Reed-Sternberg cell or a tongue of microinvasion after a very quick scan over the slide at low power. With the introduction of digital images in diagnostic histopathology it is possible to record the actions of pathologists as they observe slides and to look for any recurrent patterns in those actions. Molin et al. in this month’s Histopathology look at the patterns of observation of five pathologists looking at four cases (some of which had multiple slides with immunohistochemistry). They found, and elegantly illustrate, a number of different patterns including directed panning, cover panning, sporadic panning and dip zooming – giving us a whole new vocabulary for the way we look at slides. They show that pathologists do use methods of looking at digital images that cannot be used down the light microscopy and so are showing adaptation to the new technology. There is probably scope for a lot more related studies as histopathology moves towards full digital reporting.