Table 1 Final segmentation results of LV and RV test sequences, comparison between the proposed network EchoRCNN and the MaskRNN and RGMP architectures
From: Automatic cardiac evaluations using a deep video object segmentation network
| Network | Region | DSCa | JSCb | Precision | Recall |
|---|---|---|---|---|---|
| EchoRCNN | Left ventricle | 94.03 ± 0.01 | 88.83 ± 0.03 | 94.85 ± 0.02 | 93.37 ± 0.02 |
| Right ventricle | 94.97 ± 0.02 | 90.61 ± 0.04 | 97.03 ± 0.01 | 93.14 ± 0.04 | |
| MaskRNN | Left ventricle | 85.04 ± 0.01 | 70.56 ± 0.02 | 83.31 ± 0.01 | 90.99 ± 0.006 |
| Right ventricle | 69.25 ± 0.27 | 59.68 ± 0.3 | 80.78 ± 0.21 | 74.23 ± 0.32 | |
| RGMP | Left ventricle | 81.96 ± 0.1 | 70.97 ± 0.12 | 81.02 ± 0.16 | 74.55 ± 0.14 |
| Right ventricle | 65.41 ± 0.26 | 55.26 ± 0.25 | 75.76 ± 0.15 | 63.03 ± 0.3 |