|
| Reference | Year | Method | Advantages | Disadvantages |
|
| Darwiesh et al. [57] | 2016 | The method of Brownian motion of water molecules to produce contrast | Detecting edge areas to separate sections with tumors and nontumor sections | (i) Lack of detection of tumors in other tumors or other areas
(ii) High computational complexity and slow method
(iii) Lack of separation of areas with benign and malignant tumors |
| Aslam et al. [58] | 2015 | Edge detection | Detecting edge areas to separate sections with tumors and nontumor sections | (i) Lack of detection of tumors in other tumors or other areas
(ii) High computational complexity and slow method
(iii) Low accuracy
(iv) Lack of separation of areas with benign and malignant tumors |
| Qiao et al. [59] | 2021 | Watershed and hierarchical clustering algorithm | Detecting edge areas to separate sections with tumors and nontumor sections | (i) Lack of diagnosis of tumors in other tumors or other areas
(ii) High computational complexity and slow method
(iii) Low accuracy
(iv) Lack of separation of areas with benign and malignant tumors |
| Ain et al. [60] | 2014 | Concrete anisotropic emission based on group classification, support vector machine (SVM), and FCM | High accuracy in diagnosing and classifying areas with tumors | Lack of comparison with previous methods and lack of consideration for comparison with DL methods or other neural networks |
| Mobahi et al. [48] | 2011 | Genetic algorithm and discrete wavelet transform threshold method | Detecting edge areas to separate sections with tumors and nontumor sections | |
| Karnan, and Selvanayaki [65] | 2010 | The combined approach of ant colony optimization algorithms and genetic algorithm | Detecting edge areas to separate sections with tumors and nontumors sections | (i) Lack of diagnosis of tumors in other tumors or other areas
(ii) Very high computational complexity and slowness of the method
(iii) Low accuracy
(iv) Lack of separation of areas with benign and malignant tumors |
| Ghosh et al. [66] | 2018 | FCM-based chaotic firefly algorithm | (i) Detecting edge areas for separating tumor and nontumor sections
(ii) High execution speed with the complexity of the method
(iii) Accurate detection of features | (i) High computational complexity
(ii) Lack of separation of areas with benign and malignant tumors |
| Zhu et al. [67] | 2018 | Particle swarm optimization (PSO) | (i) Detecting edge areas for separating tumor and nontumor sections
(ii) High execution speed with the complexity of the method
(iii) Accurate detection of features | (i) Lack of diagnosis of tumors in other tumors or other areas
(ii) High computational complexity and slow method
(iii) Lack of separation of areas with benign and malignant tumors |
| Alagarsamy et al. [68] | 2019 | Bat algorithm | (i) Detecting edge areas for separating tumor and nontumor sections
(ii) High execution speed with the complexity of the method
(iii) Accurate detection of features | (i) Lack of detection of tumors in other tumors or other areas
(ii) High computational complexity and slow method
(iii) Lack of separation of areas with benign and malignant tumors |
| Memiş et al. [158] | 2020 | Deep CNN | Finding the head bone femoral and femur properties for low-quality MRI images | (i) Small volume of the dataset for validation and verification
(ii) Unable to support any types of disease |
| Duran et al. [159] | 2020 | Self-attention model | End-to-end attention model with multiple classes | (i) Only unable to detect prostate cancer
(ii) An additional mechanism for CAD models |
| Hu et al. [137] | 2019 | 3D-DenseUNet-569 | (i) Adaptable to depthwise separable convolution
(ii) Drop the GPU processing time | (i) Low-level feature extraction
(ii) Improper for big data
(iii) Unable to adapt to 2D images |
| Karayegen & Feyzi [161] | 2021 | Deep learning models | (i) High prediction method
(ii) Differing modality of MRI images
(iii) 3D image analysis | (i) Limited dataset for verification
(ii) Do not use all image area
(iii) Needs ground truth |
| Ahmadi et al. [162] | 2021 | Deep spiking neural network | (i) Low computational complexity
(ii) Used quantum filter
(iii) High accuracy | (i) Multistep method
(ii) Overfitting in some analysis |
| Ahmadi et al. [163] | 2021 | Robust PCA and CNN | (i) Clustering and segmentation method
(ii) Automated clustering
(iii) Used remove outliers
(iv) High accuracy and sensitivity | (i) High complexity
(ii) Do not support 3D images |
|
