Table 5. Analysis of the IQ-OTH/NCCD^a dataset using different deep learning architectures and Random Pixel Swap data augmentation with single-value and multivalue swap area factors (224×224 image resolution).

Swap factor	ResNet-34		MobileNetV3 (small)		Vision Transformer (base-16)		Swin Transformer (tiny)
	Accuracy, %	AUROCb, %	Accuracy, %	AUROC, %	Accuracy, %	AUROC, %	Accuracy, %	AUROC, %
Single value
0.1	89.02	92.26	93.90	95.10	64.02	74.97c	86.28	92.02
0.2	91.16	94.62	92.99	95.04	60.98	63.20	83.23	93.59
0.3	90.85	93.48	92.07	95.38	64.02	67.95	89.02	91.63
0.4	89.63	92.66	92.99	95.24	69.21	72.59	84.76	92.81
0.5	90.55	90.96	92.68	95.01	68.60	72.47	87.80	84.13
0.6	90.55	94.76	92.99	95.24	69.21	72.36	84.76	89.63
0.7	91.46	95.23	92.68	95.12	67.99	66.48	83.54	90.65
0.8	89.63	92.22	93.60	95.60	67.99	69.83	89.94c	95.95c
0.9	90.85	94.42	94.51c	95.77c	71.65	72.79	83.84	90.99
1.0	92.07c	95.02c	94.21	95.67	73.17c	74.64	89.94c	94.79
Multivalue
0.1‐0.2	90.55	93.80	93.90c	94.83	66.16	69.14	85.98	94.29
0.1‐0.3	89.33	90.53	93.29	95.18	72.56c	77.93c	84.76	91.84
0.1‐0.4	89.33	90.18	93.60	95.03	60.98	73.15	87.50	94.30
0.1‐0.5	91.16	95.93c	93.60	94.84	59.15	68.80	88.11c	94.94c
0.1‐0.6	90.55	93.26	92.38	94.60	62.20	59.86	87.20	93.45
0.1‐0.7	89.94	90.99	93.90c	95.33	61.28	70.73	88.11c	92.53
0.1‐0.8	87.80	93.13	93.29	95.00	66.77	76.81	86.28	85.93
0.1‐0.9	92.68c	95.29	93.60	95.29	68.29	64.98	86.59	92.69
0.1‐1.0	89.02	92.53	93.60	95.71c	62.80	69.05	83.54	94.35

^aIQ-OTH/NCCD: Iraq-Oncology Teaching Hospital/National Center for Cancer Diseases.

^bAUROC: area under the receiver operating characteristic curve.

^cHighest value in the column.