ABSTRACT Coproparasitological stool analysis based on microscopic examination is the reference diagnostic technique routinely performed in clinical microbiology laboratories. Automated image analysis systems could offer a suitable solution, reducing technologist workload and improving screening efficiency. The Human Fecal Ova & Parasite (O&P) Detection Solution (Techcyte) is an artificial intelligence (AI) software employing imaging-based algorithms to provide screening presumptive diagnostic results. This study aimed to validate the Wet Mount Iodine Solution software for detecting and diagnosing protozoan and helminthic infections in stool specimens. A total of 178 clinical stool specimens were retrospectively analyzed between July and October 2024 for comparative evaluation of the AI-based system. Positive specimens with confirmed mono ( n = 116) and mixed ( n = 31) protozoa and helminth infections, as well as negative specimens ( n = 31), were included. Conventional microscopy results were compared with AI presumptive detection and identification outcomes. In addition, operator-reviewed AI results were evaluated against the reference diagnosis. The system achieved 96.62% positive percent agreement (PPA) and 93.33% negative percent agreement (NPA) for overall detection considering operator-reviewed interpretation. The kappa coefficient was 0.865, indicating strong agreement with conventional microscopy. The AI system detected additional parasites in 74/178 (41.57%) specimens, providing a diagnostic improvement compared to traditional microscopy. AI-assisted microscopy enhances intestinal parasite diagnosis, providing reliable results when combined with operator review. Despite limitations such as specimen low volume analysis and artifact detection, the system shows strong potential for improving diagnostic workflows in clinical microbiology laboratories. IMPORTANCE Microscopic examination of stool specimens remains the reference technique for diagnosing intestinal parasitic infections although it is labor-intensive and operator-dependent methodology. The implementation of artificial intelligence-assisted microscopy offers a promising approach to streamline diagnostic workflows, improve consistency, and enhance traditional methods. Our evaluation study provides valuable clinical evaluation data of the Human Fecal Ova & Parasite (O&P) Detection Wet Mount Iodine Solution (Techcyte), demonstrating high sensitivity, specificity, and agreement with conventional microscopy. Importantly, AI-assisted review yielded diagnostic gains in a substantial proportion of specimens, underscoring its value as a screening and complementary tool for routine parasitological diagnostics. These findings highlight the potential of AI-powered image analysis to improve the accuracy and efficiency of coproparasitological testing, thereby supporting clinical decision-making and optimizing resource utilization in microbiology laboratories.
The system achieved 96.62% positive percent agreement (PPA) and 93.33% negative percent agreement (NPA) for overall detection considering operator-reviewed interpretation.
retrospective positive-enriched sampling limits generalization despite table-consistent agreement estimates
The kappa coefficient was 0.865, indicating strong agreement with conventional microscopy.
the reported contingency table supports strong agreement within this selected clinical sample
The AI system detected additional parasites in 74/178 (41.57%) specimens, providing a diagnostic improvement compared to traditional microscopy.
AI-only findings lacked independent adjudication and predominantly involved organisms of uncertain or limited clinical significance
AI-assisted microscopy enhances intestinal parasite diagnosis, providing reliable results when combined with operator review.
operator-reviewed agreement supports screening assistance but does not establish improved diagnosis in routine populations
Derived from the full evaluation — not a separate score.
Strengths
Table 2 separates raw AI performance from operator-reviewed results, while the Results disclose that artifacts occurred in 162 of 178 specimens and made most negative specimens initially appear positive.
↳ Results, Diagnostic performance; Table 2
Eight low-parasite-burden helminth specimens were assessed across replicate slides, revealing substantial within-specimen variability and practical sensitivity limits associated with the 15 µL volume.
↳ Results, Reproducibility study; Table 6
The Discussion connects slide drying, low specimen volume, and unequal microscopy replication to image quality, false negatives, reproducibility, and comparability.
↳ Discussion, limitations paragraphs
Limitations
The Abstract, Methods, and Results describe 147 positive and 31 negative specimens, while Table 2 calculates PPA and NPA from 148 positive and 30 negative specimens. This unresolved discrepancy affects the basis of the headline estimates.
↳ Abstract; Materials and Methods, Study design; Results, Diagnostic performance; Table 2
Additional AI findings were defined as gains without molecular testing or another adjudication procedure to distinguish true microscopy misses from false-positive AI detections.
↳ Materials and Methods, Data analysis; Results and Table 5
The retrospective convenience collection is heavily enriched for known positives and includes only 31 stated negatives from one regional clinical network, limiting inference to routine screening populations.
↳ Materials and Methods, Study design
The study earns credit for separating autonomous AI output from the expert-reviewed workflow and for documenting artifacts, discordances, and low-burden replicate performance. Its contribution remains incremental because the Discussion identifies a recent validation of the same platform with similar results. The strongest downward pressure comes from convenience selection, positive enrichment, and the lack of independent adjudication for AI-only findings labeled as diagnostic gains. Table 2 also uses a positive-negative distribution that conflicts with the Abstract, Methods, and Results, leaving the basis of the headline PPA and NPA unresolved.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Sound3.1
Confidence mediumThe study adds a new clinical cohort, detailed pathogen-level failure patterns, and a low-burden replicate analysis, but it follows a recent validation of the same system with similar results. Its advance is therefore useful and incremental rather than field-defining.
The comparison separates raw AI output from operator-reviewed performance and includes failure and reproducibility analyses. Convenience selection, positive enrichment, different physical slides, unequal microscopy replication, and no independent adjudication of AI-only findings introduce meaningful uncertainty.
The workflow and dependence on expert review are clearly explained, and the Discussion ultimately frames the system as assistive. The Abstract's description of unadjudicated additional detections as a diagnostic improvement materially overstates what the comparison establishes.
“providing a diagnostic improvement compared to traditional microscopy”
The Discussion compares the system with closely related automated microscopy platforms and traces drying, specimen volume, and artifact burden to performance. It does not adequately discuss spectrum restriction or the lack of independent confirmation for the claimed diagnostic gains.
Concerns4 of 4 checks
The reported reference distribution changes between the narrative specimen accounting and the aggregate performance table. The discrepancy directly concerns the denominators underlying the headline agreement estimates.
Ethics approval, funding, privacy protections, and qualified data availability are declared. The article does not clarify any relationship between Grifols and Techcyte or the vendor's role beyond server-based image analysis.
Flags: 3 declared / 5 total
24 of 24 checkable references verified
29 references in manuscript 5 have no canonical index record — counted, but not index-checkable 1 reference confirmed by manual review
References not verified automatically are resolved manually within 24 hours.
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Sources: Crossref (incl. Retraction Watch) ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The commercial workflow was evaluated on clinical specimens under laboratory-like conditions, placing it beyond a technical proof of concept. The paper states that routine-care workflow and response-time evaluation is still required.
AI-generated, human-governed. Something look off? Contact us to request a review.
Keep reading
Other papers Nabu has evaluated blind, starting with the same field.
Antibiotic prescription rate was reduced by 14.6% [8.7%, 20.6%] in 1-59 days and by 22.6% [18.3%, 26.9%] in 2-59 months in the post-intervention period
SupportedThe long ZAP isoform preferentially binds viral RNA and has distinct binding motifs compared to the short isoform.
SupportedMannes, Zelek, Trouw · 2026 · Health biotechnology