Table 2.

Diagnostic performance of expert subjective assessment (SA) ¹⁶ , IOTA simple rules (SR) ¹⁷ , ¹⁸ , IOTA simple‐rules risk (SRR) ¹⁹ and deep neural network models Ovry‐Dx1 and Ovry‐Dx2, for discriminating between benign and malignant ovarian lesions in test set (n = 150)

		Sensitivity			Specificity			Accuracy
Diagnostic model	Percent that could be classified	n/N	% (95% CI)	P	n/N	% (95% CI)	P	n/N	% (95% CI)	P
SA in all cases	100	72/75	96.0 (89.7–98.9)	N/A	66/75	88.0 (79.2–93.9)	N/A	138/150	92.0 (86.8–95.6)	N/A
Ovry‐Dx1 in all cases	100	72/75	96.0 (89.7–98.9)	1.0‡	65/75	86.7 (77.6–92.9)	1.0‡	137/150	91.3 (86.0–95.1)	1.0‡
Ovry‐Dx2 excluding inconclusive cases	87.3	66/68	97.1 (90.9–99.4)	N/A	59/63	93.7 (85.6–97.8)	N/A	125/131	95.4 (90.8–98.1)	N/A
Ovry‐Dx2 + SA in inconclusive cases	100	72/75	96.0 (89.7–98.9)	1.0‡	67/75	89.3 (80.9–94.8)	1.0‡	139/150	92.7 (87.7–96.0)	0.75‡
SR only	72.6	52/55	94.5 (86.2–98.4)	N/A	50/54	92.6 (83.3–97.4)	N/A	102/109	93.6 (87.8–97.1)	N/A
SR with inconclusive cases as malignant*	100	72/75	96.0 (89.7–98.9)	1.0§	50/75	66.7 (55.5–76.5)	0.003§	122/150	81.3 (74.5–86.9)	0.006§
SR + SA in inconclusive cases	100	71/75	94.7 (87.8–98.2)	1.0§	65/75	86.7 (77.6–92.9)	1.0§	136/150	90.7 (85.2–94.5)	1.0§
SRR in all cases†	100	72/75	96.0 (89.7–98.9)	1.0§	50/75	66.7 (55.5–76.5)	0.003§	122/150	81.3 (74.5–86.9)	0.006§

^*Cases in which SR yielded inconclusive result were classified as malignant.

^†Using cut‐off of 0.2 to define women with suspected malignancy. McNemar's test used for comparison of:

^‡Ovry‐Dx1 and Ovry‐Dx2 to SA; §Ovry‐Dx1 to SR or SRR. N/A, not applicable.