Table 2.
Comparing the segmentation accuracy of 3D, 2.5D, and 2D approaches across three auto-segmentation models to segment brain structures. The three auto-segmentation models included CapsNet, UNet, and nnUNet. These models were used to segment three representative brain structures: third ventricle, thalamus, and hippocampus, which respectively represent easy, medium, and difficult structures to segment. The segmentation accuracy was quantified by Dice scores over the test (114 brain MRIs).
| CapsNet | |||
|---|---|---|---|
| Brain Structure | 3D Dice (95% CI) | 2.5D Dice (95% CI) | 2D Dice (95% CI) |
| 3rd ventricle | 95% (94 to 96) | 90% (89 to 91) | 90% (88 to 92) |
| Thalamus | 94% (93 to 95) | 76% (74 to 78) | 75% (72 to 78) |
| Hippocampus | 92% (91 to 93) | 73% (71 to 75) | 71% (68 to 74) |
| UNet | |||
| Brain Structure | 3D Dice (95% CI) | 2.5D Dice (95% CI) | 2D Dice (95% CI) |
| 3rd ventricle | 96% (95 to 97) | 92% (91 to 93) | 91% (89 to 91) |
| Thalamus | 95% (94 to 96) | 92% (91 to 93) | 90% (88 to 92) |
| Hippocampus | 93% (92 to 94) | 86% (84 to 88) | 88% (86 to 90) |
| nnUNet | nnUNet | nnUNet | nnUNet |
| Brain Structure | Brain Structure | Brain Structure | Brain Structure |
| 3rd ventricle | 3rd ventricle | 3rd ventricle | 3rd ventricle |
| Thalamus | Thalamus | Thalamus | Thalamus |
| Hippocampus | Hippocampus | Hippocampus | Hippocampus |