Table 3 The multiclass classification performance of sonologists in detecting ten fetal neurosonographic image patterns with three reading modes.

From: Deep learning system improved detection efficacy of fetal intracranial malformations in a randomized controlled trial

  

Unassisted mode

Concurrent mode

Second mode

Second mode (1)

P-values for multiple comparisons

Concurrent vs. Unassisted

Second vs. Unassisted

Concurrent vs. Second

Second (1) vs. Unassisted

AUC

Macro

0.85 (0.70, 1.00)

0.89 (0.78, 1.00)

0.90 (0.79, 1.00)

0.86 (0.71, 1.00)

< 0.001

< 0.001

0.006

0.800

Micro

0.85 (0.85, 0.86)

0.89 (0.89, 0.90)

0.90 (0.90, 0.91)

0.85 (0.85, 0.86)

< 0.001

< 0.001

0.001

0.520

SEN

Macro

0.74 (0.44, 1.00)

0.81 (0.57, 1.00)

0.83 (0.60, 1.00)

0.74 (0.44, 1.00)

0.001

< 0.001

0.022

0.880

Micro

0.73 (0.72, 0.74)

0.81 (0.80, 0.81)

0.82 (0.81, 0.83)

0.73 (0.72, 0 .74)

< 0.001

< 0.001

0.004

0.570

SPE

Macro

0.97 (0.92, 1.00)

0.98 (0.93, 1.00)

0.98 (0.94, 1.00)

0.97 (0.92, 1.00)

0.004

0.001

0.320

0.880

Micro

0.97 (0.96, 0.97)

0.98 (0.97, 0.98)

0.98 (0.97, 0.98)

0.97 (0.96, 0.97)

0.140

0.005

< 0.001

0.001

ACC

Macro

0.95 (0.89, 1.00)

0.96 (0.91, 1.00)

0.96 (0.92, 1.00)

0.95 (0.89, 1.00)

0.001

< 0.001

0.040

0.620

Micro

0.95 (0.94, 0.95)

0.96 (0.96, 0.97)

0.96 (0.96, 0.97)

0.95 (0.94, 0.95)

0.460

< 0.001

< 0.001

0.040

FK

Macro

0.70

0.78

0.80

0.70

N/C

N/C

N/C

N/C

Micro

0.70

0.79

0.80

0.71

N/C

N/C

N/C

N/C

  1. Values in parentheses are 95% Confidence Interval. Second mode (1), the results before AI assistance in second reading mode. AUC Area under the receiver operating characteristic curve, SEN Sensitivity, SPE Specificity, ACC Accuracy. N/C did not compare. Macro and Micro AUCs between each two groups were compared by paired t-test and Delong’s test, respectively. Macro and Micro SENs, SPEs, ACCs between each two groups were compared by paired t-test and chi-square test for proportions, respectively. Multiple comparisons were corrected by Bonferroni method.