Table 1 The related research
Parts | Authors | Methods | Deficiencies |
|---|---|---|---|
Thyroid | Chen (2023) | Unet + sobel | Need to do a lot of preorder calculations to get training inputs |
Thyroid | Balachandran (2023) | ACU2E-Net | Huge computation |
Nodules | Bi (2023) | BPAT-UNet | Huge computation |
Nodules | Shao (2023) | FCG-Net | Not much improved compared with the basic model |
Nodules | Dai (2023) | SK-Unet++ | Higher image quality requirements, large error distance for mis-segmentation |
Multi-target | Kumar (2020) | MPCNN | Not good for small nodules with internal vesicles |
Multi-target | Webb (2020) | DeepLabv3 + LSTM | Poor segmentation results for nodules |
Multi-target | Luo (2021) | Cascade R-CNN | No segmentation of nodules, internal vessels, poor segmentation results for small targets |
Multi-target | Ma (2022) | SPRMaskR-CNN | Detection identification and segmentation is done in two steps. Cannot identify some quite small organs |
Multi-target | Zheng (2023) | DSRU-Net | Poor segmentation of small nodules |