. Author manuscript; available in PMC: 2019 Apr 11.

Published in final edited form as: Med Image Comput Comput Assist Interv. 2018 Sep 26;11070:871–879. doi: 10.1007/978-3-030-00928-1_98

Table 1.

Comparative evaluation of classification performance. In column “Inputs”, “I”and “A” refer to US images and attention maps, respectively. “SS-cls Net” refers to single-stream network trained only on attention maps to classify US video frames.

Models	Inputs	Precision	Recall	F1-score
M-SEN BCE + GAN	I	96.8	96.2	96.5
M-SEN BCE	I	96.7	90.5	93.5
M-SEN MSE + GAN	I	94.8	91.9	93.3
M-SEN MSE	I	92.4	75.6	83.2
SonoNet-32 [4]	I	79.3	82.1	80.7
SonoNet-16 [4]	I	73.6	74.1	73.8
SS-cls Net	A	71.5	76.4	73.9
SonoEyeNet-Late FT [5]	I and A	96.5	99.0	97.8