Deep learning model for pleural effusion detection via active learning and pseudo-labeling: a multisite study

BMC Medical Imaging

Table 4 Performance of the AI algorithm by Each Subpopulation

	Sensitivity (95% Wilson CI)	Specificity (95% Wilson CI)	AUROC (95% DeLong’s CI)
Radiologic Findings^a	1.00 (0.80, 1.00)	0.90 (0.60, 0.99)	0.97 (0.90, 1.00)
Image Quality Issues^b	0.75 (0.30, 0.99)	1.00 (0.44, 1.00)	0.92 (0.69, 1.00)

^a Cases with radiologic findings include possible confounders as Air-fluid Level, Airspace Disease, Atelectasis, Blebs, Cardiomegaly, Fracture, Infiltrate, Mass, Nodule, Obstructive Airways Disease, Pleural Effusion, Pneumonia, and Scoliosis
^b Cases with image quality issues include possible confounders as Anatomy not complete, Artifact present, Field of view issues, and Others

ISSN: 1471-2342