Deep learning single-cell analysis for cytologic evaluation of oral potentially malignant disorders.
An artificial intelligence tool analyzing a simple oral brush swab detects early cancer signs with near-perfect accuracy, bringing high-tech screening to everyday clinical settings.
This study applies a deep learning model to single-cell oral brush cytology from 692 subjects to classify four cell phenotypes (differentiated squamous epithelial cells, small round cells, leukocytes, lone nuclei) without manual feature extraction, generating an objective cytology-derived oral cancer numerical index (OCNI). OCNI achieved AUROC up to 0.99 for malignant versus healthy lesions with high reproducibility (ICC >=0.96), providing a clinically feasible platform for early detection and surveillance of oral potentially malignant disorders at point-of-care settings.
What the study was
- Study design
- retrospective_cohort
- Population
- Subjects with OPMDs, OSCC, and healthy controls (multi-institutional US/UK)
- Sample size
- 692
- Category
- Early Detection
- Maturity
- Exploratory
- Journal
- Sci Rep
Why it surfaced
Novel DL cytology platform with exceptional diagnostic accuracy (AUROC 0.99) for early oral cancer detection via minimally invasive brush biopsy; validated in 692 subjects; relevant to both AI/ML diagnostics and early cancer detection watchlists; multi-institutional US/UK team.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.