Digital Pathology: An Overview


Digital pathology shows substantial promise. While the immediate mileau primarily supports the use of digital slides to replace conventional glass slides, the potential benefits of digital pathology extend far beyond the use of slides for diagnostics.

First, a primer on digital slides. Essentially, a conventional slide is inserted into a machine that contains microscope objectives and at least one camera. This slide is repeatedly passed under the objective and camera to generate a huge quantity of images, which can be stitched together by the scanning software to make one larger image. Depending on the hardware, scanning a slide at "20x" can take as little as one to three minutes, while scanning one at "40x" can take more like five minutes. The files range in size from 10MB for small biopsies scanned at 20x (and roughly 1GB for specimens that comprise the majority of the slide) to 1.5GB to 20GB for slides scanned at a higher resolution.

Since Philips was cleared by the FDA for the use of their slide scanners for diagnostic purposes, whole slide imaging can now be integrated into hospital workflows. Studies that assess the accuracy and speed at which pathologists diagnose disease based on digital slides suggest that digital slides can allow for similar rates of concordance as conventional slides (even at 20x), but may be somewhat slower to diagnose. This difference in speed is largely explained by the fact that digital slides offer a smaller field of view (~20% of an eyepiece) when only one average monitor is used. Diagnosis speed is roughly equivalent when multiple monitors are used, or when a single high-resolution monitor is used.

Currently, the major perceived advantages of digital slides are that digital slides don't degrade or break like conventional slides, and they can be easily shared, rather than getting additional cuts of precious tissue to send to other pathologists or for self-study. As a result, digital study sets are becoming more prominent. The process of archiving and creating a slide viewer is not challenging; one can learn how to program and use already existing mapping tools like Google Maps, Leaflet, and OpenSeaDragon to display converted slide files. Some examples of my work on this project can be found embedded on this page. I do also have a large archive that integrates associated media with the slides (such as de-identified demographic information, clinical photographs, gross photographs, and radiology), but it is cost-prohibitive (at least, on a resident's salary) to independently host and share that information.

The more exciting aspects of digital pathology come in what we can do with digital slides besides diagnostics and simple study patterns. After all, digital slides are collections of data, and data can be transformed, manipulated, or quantitatively analyzed. Machine learning (a form of artificial intelligence) is a method to analyze slide files and has, in my opinion, the most potential to transform the way pathologists practice and offer new opportunities for research. I have a page dedicated to it here.