Lesion library

The 36 lesions Oranosis ranks, with their clinical summary, how to tell them apart and references.

Search lesions

Evidence

Nested cross-validation on 255 biopsy-confirmed lesions (21 malignant) from one centre.

Measured for the full research model. This page runs the public model, which is about 1.6 points lower in top-1 accuracy (sensitivity analysis).

Measure	Value	What it means
Top-3 accuracy (differential)	83.8%	The biopsy diagnosis is among the first three ranked lesions.
Malignancy AUC	0.953	How well the risk score separates malignant from benign lesions (1 = perfect).
Cancer sensitivity at the borderline tier or above	95.2%	Share of cancers rated borderline or higher. Specificity at this cut-off: 66.1%.
Specificity at the high tier	96.4%	Share of benign lesions not rated high. Sensitivity at this cut-off: 76.2%.

Loading the comparison…

Loading the cut-offs…

Loading…

For: dentists and oral medicine/surgery clinicians, as a second opinion on peripheral (exophytic) soft-tissue lesions.
Not for: ruling out cancer, intraosseous lesions, or use without a clinician.
Evidence: validated at one centre (255 lesions, 21 malignant). External validation pending. Figures refer to the research model; this page runs the public model (about 1.6 points lower top-1 in the sensitivity analysis).
Privacy: everything runs in your browser; nothing you enter is sent anywhere or stored. Do not enter names or IDs anyway.
Same engine as the Python version: this page contains a JavaScript port of the Oranosis engine. It was tested against the Python reference on 45,900 inputs with 0 differences (see reports/js_equivalence.md in the repository).

Zeinali MR, Mirzaei H, Khalighi HR. Introducing Oranosis v1 (manuscript in preparation).