Table 2:

Diagnostic accuracy of the 3 observers for the detection of cranial AVFs (n = 31), with DSA serving as the reference standard^{^a}

Detection of AVF	Observer 1, Method 1^{^b}		Observer 2, Method 1^{^b}		Observer 3, Method 1^{^b}		Observer 1, Method 2^{^c}		Observer 2, Method 2^{^c}		Observer 3, Method 2^{^c}
True-positive, true-negative	19	10	21	10	18	7	20	10	21	10	18	7
False-positive, false-negative	0	2	0	0	3	3	0	1	0	0	3	3
Sensitivity (%) (95% CI)	90 (70–99)		100 (81–100)		86 (63–96)		95 (76–100)		100 (81–100)		86 (63–96)
Specificity (%) (95% CI)	100 (69–100)		100 (66–100)		70 (35–92)		100 (69–100)		100 (66–100)		70 (35–92)
PPV (%) (95% CI)	100 (80–100)		100 (81–100)		86 (63–96)		100 (80–100)		100 (81–100)		86 (63–96)
NPV (%) (95% CI)	83 (57–94)		100 (66–100)		70 (35–92)		91 (60–98)		100 (66–100)		70 (35–92)
Fleiss κ	0.619 (Substantial agreement)						0.661 (Substantial agreement)

Note:—PPV indicates positive predictive value; NPV, negative predictive value.
↵a The Fleiss κ for interobserver agreement is shown per method.
↵b Method 1, color-mapping only.
↵c Method 2, color-mapping and dynamic subtraction rendering of 4D-CTA.