Table 4.

The auxiliary performance of FGGDL interface in the real world.

	Diagnosis accuracy
	No AI	With AI	P-value
Clinician 1	0.838 (0.801–0.874)	0.885 (0.854–0.916)	0.024
Clinician 2	0.833 (0.796–0.869)	0.875 (0.842–0.908)	0.024
Senior student 1	0.758 (0.715–0.800)	0.798 (0.758–0.837)	0.059
Senior student 2	0.753 (0.710–0.795)	0.863 (0.829–0.896)	0.000
Junior student 1	0.588 (0.539–0.636)	0.820 (0.782–0.858)	0.000
Junior student 2	0.630 (0.582–0.678)	0.758 (0.715–0.800)	0.000
Junior student 3	0.460 (0.411–0.509)	0.813 (0.774–0.851)	0.000
Junior student 4	0.695 (0.650–0.740)	0.875 (0.842–0.908)	0.000
Machine	0.893 (0.862–0.923)

Diagnosis accuracy and progression prediction accuracy are presented with 95% CIs.