Table 1 Comparison of brain tumor segmentation models on Figshare dataset.
Reference | Model/approach | Dataset | Key features and limitations |
|---|---|---|---|
Sahoo et al. (2023) | Residual U-Net based transfer learning model | Figshare | \(\bullet\) No pre-processing used \(\bullet\) Reduced feature extraction for large tumors |
Mayala et al. (2022) | MST-based method | Figshare | \(\bullet\) Interactive segmentation \(\bullet\) Graph-based approach \(\bullet\) Limited to Gaussian filter |
El-Shafai et al. (2022) | Hybrid segmentation approach | Figshare | \(\bullet\) Combines traditional and deep learning methods \(\bullet\) Noise sensitivity and difficulty with flat images |
Kasar et al. (2021) | DNNs - U-Net, SEGNET | Figshare | \(\bullet\) Semantic segmentation approach \(\bullet\) No separate DSC for tumor and background |
Sobhaninia et al. (2020) | Cascaded Deep Neural Networks | Figshare | \(\bullet\) Uses multiple image scales \(\bullet\) Needs improved Dice value |
Díaz-Pernas et al. (2021) | Multiscale CNN | Figshare | \(\bullet\) Combines multiscale features \(\bullet\) False positives from skull and vertebral parts |
Razzaghi et al. (2022) | Multimodal Deep Transfer Learning | Figshare | \(\bullet\) Combines multimodal data \(\bullet\) Scalability issues with more modalities |
Verma et al. (2024) | RR-U-Net | Figshare | \(\bullet\) U-Net with residual modules \(\bullet\) Lacks advanced architectures like DenseNets |
