Multi-label annotation of text reports from computed tomography of the chest, abdomen, and pelvis using deep learning

Table 1 Classification performance of our final RBAs compared to previously reported RBAs

Organ	Label	# Pos	Han et al. [15]		Faryna et al. [16]		Our Final RBAs
Organ	Label	# Pos	Acc	F-Score	Acc	F-Score	Acc	F-Score
Lungs/Pleura	Atelectasis	251	0.86	0.74	0.97	0.95	0.98	0.97
	Nodule	296	0.77	0.74	0.92	0.89	0.92	0.89
	Emphysema	193	0.82	0.45	0.98	0.96	0.99	0.98
	Effusion	205	0.82	0.53	0.84	0.58	0.98	0.97
	Normal	209	0.79	0.44	0.96	0.94	0.98	0.96
Liver/Gallbladder	Stone	144	0.87	0.62	0.95	0.9	0.96	0.91
	Lesion	224	0.92	0.88	0.94	0.91	0.95	0.92
	Dilatation	87	0.86	0.1	0.9	0.7	0.98	0.92
	Fatty	166	0.97	0.94	0.98	0.96	0.98	0.96
	Normal	166	0.94	0.9	0.95	0.9	0.96	0.93
Kidneys/Ureters	Stone	174	0.91	0.82	0.93	0.85	0.93	0.85
	Atrophy	94	0.96	0.85	0.99	0.97	0.99	0.97
	Lesion	238	0.91	0.87	0.91	0.86	0.91	0.86
	Cyst	234	0.95	0.92	0.96	0.94	0.96	0.94
	Normal	194	0.94	0.89	0.96	0.92	0.96	0.92

Comparison of classification performance between previously reported RBAs and our final RBAs using the manually annotated test set. “# Pos” is the number of positive examples for that label in the test set, Acc = Accuracy. Bolded values represent an equivalent F1-Score or increase in performance

ISSN: 1472-6947