Table 3. Diagnostic accuracy, sensitivity and specificity for detection of ulcerations in the small bowel and colon in patients with suspected or known Crohnʼs disease.

	TP	TN	FP	FN	Accuracy (%)	95 %CI	Sensitivity (%)	95 %CI	Specificity (%)	95 %CI
Patient split
Small bowel	317	602	2	12	98.50	97.50–99.18	96.35	93.72–98.10	99.67	98.81–99.96
Colon	130	283	0	8	98.10	96.29–99.18	94.20	88.90–97.46	100	98.70–100
Overall	447	885	2	20	98.38	97.55–98.98	95.72	93.46–97.36	99.77	99.19–99.97
Random split
Small bowel	359	620	0	15	98.49	97.52–99.15	95.99	93.47–97.74	100	99.41–100
Colon	174	302	0	6	98.76	97.31–99.54	96.67	92.89–98.77	100	98.79–100
Overall	533	922	0	21	98.58	97.83–99.12	96.21	94.26–97.64	100	99.60–100

Data are shown for two different split of images used training, validation and testing of the deep learning framework.

TP, true positive; TN, true negative; FP, false positive; FN, false negative.