Live ensemble · runs in your browser

Predictor

Calibrated logistic regression shipped from the 30-model study. Every change to an input recomputes the probability, confidence, and factor breakdown — there is no server round-trip.

CV AUC

0.891

High-conf acc

91.4%

Patient input

All 12 pre-operative variables. Prediction updates instantly.

Patient

Demographics

Age (years)68

Cohort mean 68.5

Sex

Smoking

Laterality

Tumour

Strongest predictors

Width / size X (mm)10

Cohort median ~10 mm

Height / size Y (mm)8

Cohort median ~8 mm

Type

Recurrent

Aggressive histo

Biopsy method

Location

Anatomy

Body site

Body zone

Anatomical unit

Surgeon

Operator factors

Experience

See & do planning

17.0%P(≥13 sections)95% CI 8.8–25.3%

<13 sections likelyHigh confidence ·0.33

Tumour area (ellipse)0.63 cm²

01.5 cm² SHAP threshold8+ cm²

Estimated MBS

31000

Mohs 1–6 sections

Fee A$677.70 · 75% rebate A$508.28

Code is a model-derived estimate. Final billing must reflect actual section count.

Confidence

high

Distance to 50% decision boundary: 0.330

Paper: high-confidence cases (|p−0.5| ≥ 0.15) reached 91.4% accuracy on held-out test.

Ensemble breakdownspread 38%

LR Logistic Regression (regularised)11.0%

LinearCV-AUC 0.894

HGB Histogram Gradient Boosting1.0%

Gradient BoostingCV-AUC 0.870

MLP Multilayer Perceptron39.1%

Neural NetworkCV-AUC 0.862

Soft-vote ensemble17.0%

What's driving this prediction← lowers · raises →

Recurrent tumour0

Tumour height8 (z -0.508)

Anatomical unitNOSE

Aggressive histopathology0

Tumour width10 (z -0.451)

Tumour area (ellipse)0.63 (z -0.446)

Biopsy method1

Smoking0

Not a substitute for clinical judgement. External validation in independent cohorts is pending.