Table 3 Comparison of segmentation of human teeth on CBCT using CNN.
Author | Year | Patients/images | CNN architecture | Training strategy | Evaluation strategy | DSC | Ac | Sn | SP | PPV | NPV |
|---|---|---|---|---|---|---|---|---|---|---|---|
Xu37 | 2019 | 1200/NA | DNN | 3D volume | VB | NA | 0.991 | NA | NA | NA | NA |
Tian36 | 2019 | 600/NA | U-Net + HN | 3D volume | VB | NA | 0.898 | NA | NA | NA | NA |
Cui33 | 2019 | 20/NA | ToothNet | 3D volume | VB | 0.921 | NA | NA | NA | NA | NA |
Li21 | 2020 | 24/1160 | AttU-Net + BDC-lstm | 2D slices | VB | 0.9526 | NA | NA | NA | NA | NA |
Lee34 | 2020 | 102/NA | UDS-Net | 2D slices | NA | 0.918 | NA | 0.932 | NA | 0.904 | NA |
Chen29 | 2020 | 25/NA | FCN + MWT | 3D volume | NA | 0.936 | NA | NA | NA | NA | NA |
Rao35 | 2020 | NA/86 | SFCRN + DCRF | 2D slices | NA | 0.917 | NA | NA | NA | NA | NA |
Wu32 | 2020 | 20/NA | GH + BADice + DASPP U-Net | 3D volume | VB | 0.962 | NA | NA | NA | NA | NA |
Wang27 | 2021 | 28/9507 | MS-D | NA | VB | 0.945 | NA | NA | NA | NA | NA |
Duan20 | 2021 | 30/NA | U-Net | 2D slices | VB | 0.957 | NA | NA | NA | NA | NA |
Shaheen23 | 2021 | 186/NA | 3D U-Net | 3D volume | VB | 0.90† | NA | 0.83 | NA | 0.98 | NA |
Lahoud31 | 2021 | 314/2924 | FPN | 2D slices | VB | 0.934 | NA | NA | NA | NA | NA |
Fontenele30 | 2022 | 175/ | 3D U-Net | 3D volume | VB | 0.95–0.97 | 0.994–0.997 | 0.91–0.94 | NA | 1 | NA |
Our study | 2022 | 24/12,552 | 2Da U-Net | 2D slices | SB | 0.839a | 0.999 | 0.925 | 0.999 | 0.852a | 0.999 |
3D U-Net | 3D volume | SB | 0.779a | 0.997 | 0.864 | 0.999 | 0.810a | 0.998 | |||
3.5Dv5 U-Net | 2D slices, 3D volume | SB | 0.911 | 0.999 | 0.888 | 1 | 0.970 | 0.999 |
