Prospective evaluation of automated KU F40 fecal analysis versus microscopy for intestinal protozoa and illustrative labor costs

Abstract Stool ova-and-parasite (O&P) microscopy remains widely used but is labor-intensive and constrained by expert availability. AI-assisted automated fecal analyzers can digitize stool microscopy workflows and generate images that may support remote expert review; however, evaluations should distinguish algorithm-only output from clinically deployable workflows that incorporate human verification. In a prospective diagnostic study at a tertiary-care parasitology laboratory in Izmir, Türkiye (November 2024–October 2025), 1000 consecutive, unselected adult human stool specimens submitted for routine testing were assessed in parallel by routine microscopy and the KU-F40 automated fecal analyzer. All 1,000 specimens contributed paired entries to the analysis tables. KU-F40 was evaluated as automated diagnosis (AD; algorithm call before human review) and AD plus expert user verification (AD + UV; expert review of KU-F40-generated images). The diagnostic AD + UV analysis was kept distinct from the labor-cost model, which evaluated illustrative 30%-verification and universal-verification scenarios. Relative to routine microscopy, KU-F40 AD showed positive/negative percent agreement (PPA/NPA) of 100%/99.8% for Entamoeba histolytica/dispar , 100%/98.6% for Giardia spp ., and 67.3%/97.7% for Blastocystis spp. Expert verification removed the 14 microscopy-negative Giardia AD calls and reduced microscopy-negative Blastocystis calls from 22 to 12; the two microscopy-negative E. histolytica/dispar calls remained UV-supported but were not independently adjudicated. Coccidia ( Cryptosporidium spp., Cyclospora cayetanensis ) fell outside the KU-F40 classifier panel and were identified only by conventional microscopy. The illustrative labor model assumed 35 min/specimen for microscopy and 2 min/specimen for AD, with 10 min of specialist time per verified specimen; no formal time-and-motion study was performed. The resulting modeled direct labor costs (USD/specimen) for microscopy, AD and illustrative 30%-verification AD + UV were Türkiye 4.56/0.20/0.72, Germany 21.87/0.91/3.46 and UK 27.21/0.79/4.66. KU-F40 provided a digitized screening workflow. Under the specified assumptions, the illustrative labor model yielded lower estimated hands-on labor inputs and direct labor costs than conventional microscopy; these were model outputs, not observed operational performance. Expert verification improved agreement for selected organisms, but UV-supported discordant calls were not an independent confirmation of infection. KU-F40 did not cover Cryptosporidium or Cyclospora in this evaluation; conventional and dedicated methods therefore remain necessary whenever these organisms fall within the requested diagnostic scope. Multi-site studies with measured workflow times and independent discordance adjudication are required before implementation or cost-effectiveness claims.