Skip to main content
. Author manuscript; available in PMC: 2020 Jun 1.
Published in final edited form as: Inf Process Med Imaging. 2019 May 22;26:592–604. doi: 10.1007/978-3-030-20351-1_46

Table 3.

Standard plane detection results after fine-tuning (mean ± standard deviation [%]). Rand. Init. denotes the SE-ResNeXt model trained from scratch. The best score among the first three models is marked in bold. Scores of the fine-tuned SonoNet that exceed all three models are marked in bold as well. The literature SonoNet scores are given in parenthesis.

Rand. Init. Gaze-FT Saliency-FT Δ(Saliency, Rand. Init.) SonoNet-FT (Lit. value [1])
Precision 70.4 ± 2.3 67.2 ± 3.4 79.5 ± 1.7   9.1 ± 2.1 82.3 ± 1.3 (81)
Recall 64.9 ± 1.6 57.3 ± 4.5 75.1 ± 3.4 10.2 ± 1.9 87.3 ± 1.1 (86)
F1-score 67.0 ± 1.3 60.7 ± 3.9 76.6 ± 2.6   9.6 ± 2.1 84.5 ± 0.9 (83)

F1-scores per class:
RVOT 37.9 ± 3.8 30.4 ± 4.9 58.7 ± 2.7 20.8 ± 5.5 71.2 ± 2.8
LVOT 30.3 ± 4.7 25.8 ± 5.9 48.6 ± 3.3 18.4 ± 7.3 69.9 ± 5.3
4CH 43.1 ± 6.7 33.5 ± 8.5 57.3 ±10.8 14.2 ±11.9 75.7 ± 9.1
Kidneys 71.4 ± 5.5 68.5 ±12.1 84.7 ± 6.3 13.3 ± 5.7 81.0 ± 5.0
Profile 77.5 ± 7.2 61.7 ± 7.7 87.2 ± 7.5   9.7 ± 3.7 88.1 ± 4.5
Lips 76.7 ± 2.6 74.2 ± 7.3 85.6 ± 4.5   8.8 ± 6.7 92.9 ± 0.8
Brain (Cb.) 84.9 ± 7.0 83.2 ± 1.5 93.7 ± 4.6   8.8 ± 2.3 92.8 ± 1.1
3VT 50.4 ± 1.9 47.1 ± 7.1 58.3 ± 7.1   7.9 ± 5.6 77.9 ± 1.6
Brain (Tv.) 86.1 ± 7.7 88.8 ± 2.8 92.9 ± 5.0   6.8 ± 2.8 92.1 ± 4.5
Spine (cor.) 72.9 ± 3.6 57.2 ± 2.8 79.0 ± 3.7   6.1 ± 6.2 90.3 ± 4.9
Abdominal 67.9 ± 5.1 60.8 ± 6.7 72.9 ± 2.9   5.0 ± 3.7 85.0 ± 1.4
Spine (sag.) 86.5 ± 3.5 80.2 ± 2.5 89.1 ± 2.1   2.7 ± 5.2 91.6 ± 2.5
Femur 85.7 ± 2.0 77.7 ± 0.1 87.6 ± 1.3   1.9 ± 1.5 89.5 ± 1.8
Background 85.2 ± 0.9 83.3 ± 0.7 89.0 ± 0.4   3.8 ± 1.2 90.3 ± 0.4

RVOT: right ventricular outflow tract; LVOT: left ventricular outflow tract; 4CH: four chamber view; 3VT: three vessel and trachea view; Brain (Cb.): brain cerebellum suboccipitobregmatic plane; Brain (Tv.): brain transventricular plane; Cor.: Coronal plane; Sag.: Sagittal plane.