Fig 1.
Overall scheme of the pre-processing and deep learning and its evaluation procedure.
Table 1.
Description of ultrasound diagnosis equipment used in this study.
Fig 2.
Image classification criteria defined in this study.
In a schematic diagram of a blood vessel, black indicates the visualized condition, and white indicates the invisible condition.
Table 2.
Description of the dataset.
Fig 3.
Receiver operating characteristic curves for the learned model developed in this study.
(a: Stationary type, b: Portable type).
Fig 4.
Examples of correctly classified unsatisfactory images.
(a) Short axis acquired by stationary ultrasonography equipment, (b) Long axis acquired by stationary ultrasonography equipment, (c) Short axis acquired by portable ultrasonography equipment, and (d) Long axis acquired by portable ultrasonography equipment.
Fig 5.
Examples of heatmaps for correctly classified unsatisfactory images.
(a) Short axis acquired by stationary ultrasonography equipment, (b) Long axis acquired by stationary ultrasonography equipment, (c) Short axis acquired by portable ultrasonography equipment, and (d) Long axis acquired by portable ultrasonography equipment.
Table 3.
Normalized confusion matrix for stationary ultrasonography.
Table 4.
Normalized confusion matrix for portable ultrasonography.
Table 5.
Classification accuracy using the learned model.
Fig 6.
Examples of correctly classified satisfactory images.
(a) Short axis acquired by stationary ultrasonography equipment, (b) Long axis acquired by stationary ultrasonography equipment, (c) Short axis acquired by portable ultrasonography equipment, and (d) Long axis acquired by portable ultrasonography equipment.
Fig 7.
Examples of heatmaps for correctly classified satisfactory images.
(a) Short axis acquired by stationary ultrasonography equipment, (b) Long axis acquired by stationary ultrasonography equipment, (c) Short axis acquired by portable ultrasonography equipment, and (d) Long axis acquired by portable ultrasonography equipment.
Fig 8.
Examples of correctly classified Moderately Satisfactory images.
(a) Short axis acquired by stationary ultrasonography equipment, (b) Long axis acquired by stationary ultrasonography equipment, (c) Short axis acquired by portable ultrasonography equipment, and (d) Long axis acquired by portable ultrasonography equipment.
Fig 9.
Examples of heatmaps for correctly classified Moderately Satisfactory images.
(a) Short axis acquired by stationary ultrasonography equipment, (b) Long axis acquired by stationary ultrasonography equipment, (c) Short axis acquired by portable ultrasonography equipment, and (d) Long axis acquired by portable ultrasonography equipment.
Fig 10.
Examples of Moderately Satisfactory images misclassified as satisfactory images and the corresponding heatmap images.
(a) Long axis acquired by stationary ultrasonography equipment, (b) Long axis acquired by Portable ultrasonography equipment, (c) Long axis acquired by stationary ultrasonography equipment, and (d) Long axis acquired by portable ultrasonography equipment.