Expert image review changes automated parasite findings in İzmir
A peer-reviewed evaluation in Türkiye compared an automated stool analyser with routine microscopy on 1,000 specimens. Expert review of the machine’s images removed some discordant parasite calls but could not recover organisms missing from those images. The findings distinguish the contribution of human review from a separate labour-cost calculation based on assumed working times.
Science··Midday
A thousand specimens compared side by side
Expert review changed some of the parasite calls made by the KU-F40, an automated stool-imaging analyser, in a peer-reviewed study from İzmir, Türkiye. A university hospital parasitology laboratory examined 1,000 consecutive adult specimens with the instrument and routine microscopy between November 2024 and October 2025. The researchers separately assessed the initial automated output and a specialist’s reading of the digital images. Specimens, rather than uniquely identified patients, were the units of comparison.[1]
Image review corrected some discordant calls
For the intestinal parasite Giardia, review rejected all fourteen automated positive calls that routine microscopy had classified as negative. For Blastocystis, another intestinal organism, that discordant group shrank from twenty-two specimens to twelve. Seventeen microscopy-positive Blastocystis specimens remained negative in the machine images after review. Because the specialist examined images from the same device, the review could neither recover organisms absent from those fields nor independently establish infection in disputed specimens.[1]
The classifier covered only part of the parasite panel
Cryptosporidium and Cyclospora fell outside the instrument’s classification scope; microscopy found such organisms in thirty specimens. Overall positive totals consequently covered different targets. Separately, a labour calculation used assumed task times and illustrative wages, excluding equipment, reagents and maintenance. Its lower estimated labour costs describe a scenario. The study did not measure actual workflow savings or the full expense of operating the system.[1]
