Figure - PMC

Skip to main content

View full-text article in PMC

. Author manuscript; available in PMC: 2023 Jun 10.

Published in final edited form as: Radiology. 2019 Jul 9;292(3):695–701. doi: 10.1148/radiol.2019181343

Figure 6: — Receiver operating characteristic (ROC) curves comparing deep learning and radiologists using ACR TI-RADS on (a) 49 easy test cases and (b) 50 difficult test cases. Easy subset includes 10 malignant nodules based on unanimous correct management decisions from 9 readers and expert consensus and 39 benign cases based on at least 6 out of 10 correct management decisions for FNA recommendation. Difficult subset contains the remaining 5 malignant and 45 benign test cases.

Figure 6: — Receiver operating characteristic (ROC) curves comparing deep learning and radiologists using ACR TI-RADS on (a) 49 easy test cases and (b) 50 difficult test cases. Easy subset includes 10 malignant nodules based on unanimous correct management decisions from 9 readers and expert consensus and 39 benign cases based on at least 6 out of 10 correct management decisions for FNA recommendation. Difficult subset contains the remaining 5 malignant and 45 benign test cases.