From Blaine: These images were captured by software trained to detect intestinal protozoa in trichrome-stained stool specimens. The final report was Entamoeba histolytica/E. dispar and Blastocystis sp. It is difficult to determine the species of Entamoeba from only four examplars, but the even peripheral chromatin and discrete karyosome are supportive of E. histolytica/E. dispar which is how this case was signed out from both the slides and the whole-slide scanned image. I see a lot of respondents placing weight on the cytoplasm, which I always take as a ‘soft’ feature – I prefer peripheral chromatin.
The software we are training (still in validation stage) is designed to detect Entamoeba based on the nucleus (the amorphic forms of trophozoites created too many false positives when we tried to detect the entire organism). The software is not designed (at least at the moment) to identify the Entamoeba to the species level. All slides flagged as positive with a high score of being positive are to be manually backread for confirmation and species-level ID. The software is more of a screening tool, in hopes of cutting down manual read of 70% of our negative trichromes.
So to those who questions if artificial intelligence is going to make parasitologists obsolete - the answer is "no"! Dr. Marc Courturier assures us in his comment: "Have no fear. We will all have jobs as expert parasitologists. This technology that Blaine and I have been working on will make our work MUCH easier, faster, and more accurate. And fun...because lets face it...no one likes sitting for 8 hours at a scope.The future is bright for parasitology."
Thanks again to Blaine (and Marc) for donating this case!
EmoticonEmoticon