Table 6 AUC result overview for our experiments on the official split. In this table, we present results for our best performing architecture with different depth (i.e. ResNet38-large-meta, ResNet50-large-meta, ResNet101-large-meta) and compare them to other groups.

Pathology	Wang et al.⁸	Yao et al.¹³	Guendel et al.¹⁴	“-large-meta”
Pathology	Wang et al.⁸	Yao et al.¹³	Guendel et al.¹⁴	ResNet-38	ResNet-50	ResNet-101
Cardiomegaly	0.810	0.856	0.883	0.875	0.877	0.865
Emphysema	0.833	0.842	0.895	0.895	0.875	0.868
Edema	0.805	0.806	0.835	0.846	0.842	0.828
Hernia	0.872	0.775	0.896	0.937	0.916	0.855
Pneumothorax	0.799	0.805	0.846	0.840	0.819	0.839
Effusion	0.759	0.806	0.828	0.822	0.818	0.818
Mass	0.693	0.777	0.821	0.820	0.810	0.796
Fibrosis	0.786	0.743	0.818	0.816	0.800	0.778
Atelectasis	0.700	0.733	0.767	0.763	0.755	0.747
Consolidation	0.703	0.711	0.745	0.749	0.742	0.734
Pleural Thicken.	0.684	0.724	0.761	0.763	0.742	0.739
Nodule	0.669	0.724	0.758	0.747	0.736	0.738
Pneumonia	0.658	0.684	0.731	0.714	0.703	0.694
Infiltration	0.661	0.673	0.709	0.694	0.694	0.686
Average	0.745	0.761	0.807	0.806	0.795	0.785
No Findings	—	—	—	0.727	0.725	0.720

Additionally we provide an average AUC over all pathologies in the last row. Bold text emphasizes the overall highest AUC value.

Quick links

Search