Manuscript Tables 1–3

Evidence

Every number on this page is transcribed directly from the manuscript. Use the filters on the leaderboard to slice the 30-algorithm study by family.

Cohort size

408

Best CV AUC

0.891

Best Test AUC

0.892

Algorithms

Table 1 — Baseline characteristics

Stratified by outcome (≥13 vs <13 sections).

Manuscript §3.1

Characteristic	All n = 408	<13 sections n = 213	≥13 sections n = 195	p
Demographics
Age, years — mean ± SD	68.5 ± 12.9	66.1 ± 13.4	71.0 ± 11.8	<0.001
Male — n (%)	236 (57.8%)	115 (54.0%)	121 (62.1%)	0.081
Tumour characteristics
Tumour Size X, mm	17.5 ± 16.3	9.1 ± 7.6	26.5 ± 18.5	<0.001
Tumour Size Y, mm	16.4 ± 16.4	8.7 ± 7.8	24.8 ± 19.2	<0.001
Tumour Area, cm²	3.4 ± 6.3	0.75 ± 0.98	6.32 ± 8.09	<0.001
BCC — n (%)	367 (89.9%)	198 (93.0%)	169 (86.7%)	0.023
SCC — n (%)	41 (10.1%)	15 (7.0%)	26 (13.3%)	0.023
Recurrent — n (%)	127 (31.1%)	45 (21.1%)	82 (42.1%)	<0.001
Aggressive histopathology — n (%)	249 (61.0%)	113 (53.0%)	136 (69.7%)	<0.001
Anatomy
Head & Neck — n (%)	381 (93.4%)	205 (96.2%)	176 (90.3%)	0.002
H-zone — n (%)	291 (71.3%)	136 (63.8%)	155 (79.5%)	<0.001
M-zone — n (%)	100 (24.5%)	64 (30.0%)	36 (18.5%)	<0.001
L-zone — n (%)	17 (4.2%)	13 (6.1%)	4 (2.1%)	<0.001

From manuscript Table 1. Continuous variables: Mann–Whitney U; categorical: χ². Bold-coloured p-values indicate significance at 0.05.

Table 2 — Effect sizes

Cohen's d for continuous, Cramér's V for categorical variables.

Manuscript §3.2

Continuous variables — Cohen's d

Tumour Area (cm²)6.32 vs 0.75

0.982Large

Tumour Size X (mm)26.51 vs 9.12

1.237Large

Tumour Size Y (mm)24.79 vs 8.67

1.111Large

Age (years)71.03 vs 66.12

0.389Small

small 0.2medium 0.5large 0.8

Categorical variables — Cramér's V

Recurrentp = <0.001

0.284Medium

Body Zonep = <0.001

0.232Small–Medium

Aggressive Histopathologyp = <0.001

0.176Small

Body Sitep = 0.002

0.150Small

Tumour Typep = 0.023

0.113Small

Sexp = 0.081

0.087Negligible

Surgeon Experiencep = 0.283

0.053Negligible

small 0.1medium 0.3large 0.5

Table 3 — Model leaderboard

30 algorithms across 6 families. 3 ensemble · 8 neural networks.

Manuscript §3.3

Showing 30 of 30 algorithms evaluated in the manuscript. Lighter category fills in the right chart column.

#	Model	Category	CV AUC	Test AUC	F1	Brier	CV AUC
1	Stacking Ensemble (LR meta)Best	Ensemble	0.891±0.04	0.884	0.810	0.129	0.891
2	Random Forest	Tree	0.891±0.04	0.851	0.800	0.155	0.891
3	Soft Voting Classifier	Ensemble	0.888±0.04	0.873	0.795	0.140	0.888
4	Extra Trees	Tree	0.885±0.04	0.870	0.795	0.147	0.885
5	CatBoost	Gradient Boosting	0.885±0.04	0.881	0.779	0.134	0.885
6	MLP Wide 5-Layer (1024-512-256-128-64)	Neural Network	0.882±0.04	0.874	0.816	0.141	0.882
7	MLP 7-Layer (512-256-128-64-32-16-8)	Neural Network	0.881±0.02	0.871	0.800	0.140	0.881
8	SVM-RBF	SVM	0.877±0.02	0.887	0.785	0.137	0.877
9	SVM-Linear	SVM	0.875±0.04	0.886	0.800	0.132	0.875
10	Stacking Ensemble (XGB meta)	Ensemble	0.874±0.05	0.875	0.795	0.142	0.874
11	AdaBoost	Gradient Boosting	0.872±0.05	0.882	0.779	0.207	0.872
12	LightGBM	Gradient Boosting	0.870±0.05	0.853	0.760	0.177	0.870
13	MLP Wide 3-Layer (512-256-128)	Neural Network	0.869±0.02	0.892	0.825	0.126	0.869
14	Linear Discriminant Analysis	Traditional	0.869±0.03	0.887	0.795	0.130	0.869
15	Logistic Regression	Traditional	0.867±0.03	0.886	0.789	0.133	0.867
16	Ridge Classifier	Traditional	0.864±0.03	0.883	0.785	0.138	0.864
17	Gradient Boosting (sklearn)	Gradient Boosting	0.861±0.04	0.848	0.755	0.173	0.861
18	XGBoost	Gradient Boosting	0.858±0.05	0.844	0.753	0.168	0.858
19	MLP-Large (256-128-64-32)	Neural Network	0.856±0.03	0.850	0.765	0.152	0.856
20	Bagging Classifier	Tree	0.854±0.04	0.842	0.760	0.162	0.854
21	MLP-Medium (128-64-32)	Neural Network	0.851±0.04	0.848	0.755	0.160	0.851
22	MLP 6-Layer (512-256-128-64-32-16)	Neural Network	0.848±0.03	0.853	0.762	0.151	0.848
23	MLP-Small (64-32)	Neural Network	0.842±0.04	0.838	0.748	0.169	0.842
24	MLP 5-Layer (256-128-64-32-16)	Neural Network	0.840±0.03	0.841	0.750	0.165	0.840
25	SVM-Polynomial	SVM	0.838±0.05	0.821	0.720	0.185	0.838
26	Quadratic Discriminant Analysis	Traditional	0.819±0.05	0.817	0.710	0.196	0.819
27	SVM-Sigmoid	SVM	0.806±0.06	0.780	0.692	0.212	0.806
28	Decision Tree	Tree	0.762±0.06	0.750	0.680	0.245	0.762
29	Naïve Bayes (Gaussian)	Traditional	0.755±0.05	0.735	0.660	0.268	0.755
30	k-Nearest Neighbours	Traditional	0.720±0.07	0.715	0.650	0.275	0.720