Table 4.
Summary of common deep learning architectures and methods.
| References | Architecture | Methods | Detailed method | Applications | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ML | CNN | 3D CNN | 2D CNN | DenseNet | ResNet | UNet | AlexNet | FCN | Classification | Detection | Segmentation | |||
| [31] | √ | √ | √ | 2DCNN + 3DCNN + Inception CNN | Head and neck cancer | |||||||||
| [101] | √ | √ | CNN extracts topological embeddings, and in using binary classification | |||||||||||
| [32] | √ | √ | DenseNet classification after dimensionality reduction using convolutional gated cyclic units | In vivo Tumors | ||||||||||
| [33] | √ | √ | √ | 3D CNN and 2D inception CNN | Head and neck cancer | |||||||||
| [63] | √ | √ | CNN classifier | Head and neck cancer | ||||||||||
| [34] | √ | √ | KidneyResNet consisting of Resnet-18 | Ambient infusion | ||||||||||
| [80] | √ | √ | Combining modulated Gabor and CNN in the MGCNN framework |
Red blood cells | ||||||||||
| [35] | √ | √ | √ | Spectral-Spatial-CNN with 3D convolution | Stomach Cancer | |||||||||
| [17] | √ | √ | CNN training with different patch sizes after PCA dimensionality reduction | Red blood cells | ||||||||||
| [36] | √ | √ | Gabor filter and CNN | Red blood cells | ||||||||||
| [37] | √ | √ | CNN | Tissue classification | ||||||||||
| [81] | √ | √ | √ | Compare the classification performance using (RBF-SVM), MLP, and 3DCNN | Stomach and Colon Cancer | |||||||||
| [82] | √ | √ | Combining PCA, SVM, KNN classification with K-means for final weighted voting classification | Brain tumor | ||||||||||
| [83] | √ | √ | SVM combined with ANN for classification | Identification of cancer cells | ||||||||||
| [39] | √ | √ | √ | HybridSpectraNet (HybridSN) composed of 3D CNN and 2D CNN in spectral space | Colon Cancer | |||||||||
| [84] | √ | √ | 3D CNN combined with 3D attention module for deep hypernetworks | White blood cells | ||||||||||
| [40] | √ | √ | SICSURFIS HSI-CNN system composed of SICSURFIS imager and CNN | Skin disease | ||||||||||
| [85] | √ | √ | Stacked auto encoder (SAE) | Tongue coating | ||||||||||
| [93] | √ | √ | White blood cells | |||||||||||
| [41] | √ | √ | K-means and SAM | Skin disease | ||||||||||
| [86] | √ | √ | Two-channel deep fusion network EtoE-Fusion CNN for feature extraction | White and red blood cells | ||||||||||
| [42] | √ | √ | Mapping RGB to high broad-spectrum domain with 2D CNN classification | Breast cancer | ||||||||||
| [95] | √ | √ | The external U-Net handles spectral information, and the internal u handles spatial information, making up the UwU-Net classification | Drug position | ||||||||||
| [18] | √ | √ | Regression-based partitioned deep convolutional networks | Head and neck cancer | ||||||||||
| [94] | √ | √ | √ | √ | 1D, 2D, 3D CNN, RNN, MLP, SVM for comparison | Blood Classification | ||||||||
| [87] | √ | √ | √ | √ | U-Net, 2D CNN, 1D DNN combined with classification | Brain cancer | ||||||||
| [43] | √ | √ | Extracting image elements into patches into CNN | Head and neck cancer | ||||||||||
| [44] | √ | √ | RF, SVM, MLP and K-Nearest Neighbor Comparison | Esophageal Cancer | ||||||||||
| [45] | √ | √ | Pixel-level classification | Head and neck cancer | ||||||||||
| [89] | √ | √ | √ | AlexNet combined with SVM | Corneal epithelial tissue | |||||||||
| [90] | √ | √ | √ | Hybrid 3D-2D network for extracting spatial and spectral features | Brain cancer | |||||||||
| [91] | √ | √ | CNN with support vector machine (SVM), random forest (RF) synthetic classification | Tissue classification | ||||||||||
| [102] | √ | √ | LDA | Septicemia | ||||||||||
| [48] | √ | √ | CNN architecture for inception-v4 | Head and neck cancer | ||||||||||
| [103] | √ | √ | CNN architecture for inception-v4 | Head and neck cancer | ||||||||||
| [51] | √ | √ | 2D CNN classification | Brain cancer | ||||||||||
| [52] | √ | √ | RF, logistic regression, SVM comparative classification | Head and neck cancer | ||||||||||
| [58] | √ | √ | ResNet34 | Stomach Cancer | ||||||||||
| [92] | √ | √ | RF, SVM, MLP | Colon Cancer | ||||||||||
| [59] | √ | √ | PCA downscaling, Spectral Angle Mapper (SAM) | Stomach Cancer | ||||||||||
| [60] | √ | √ | Discrete Wavelet Transform (DWT) based feature extraction, SVM | Head and neck cancer | ||||||||||
| [96] | √ | √ | Dual-stream convolution model | Tongue Tumor | ||||||||||
| [97] | √ | √ | DenseNet-Blocks combined with 3D CNN to extract spatial spectral information | Head and neck cancer | ||||||||||
| [46] | √ | √ | CNN with Deep Local Features (DELF) | Skin Features | ||||||||||
| [49] | √ | √ | CNN and SVM + PCA + KNN are used, respectively | Head and neck cancer | ||||||||||
| [99] | √ | √ | Select the channel and use U-Net | Head and neck cancer | ||||||||||
| [55] | √ | √ | 3D full convolutional network with extended convolutional fast and fine-grained feature dual path | Melanoma | ||||||||||
| [100] | √ | √ | The encoding part of U-Net uses transformer to extract the spectral information and convolution to extract the spatial information jointly | Carcinoma of bile duct | ||||||||||
| [56] | √ | √ | Pixel-based, superpixel-based, patch-based, and full image-based data are fed into the CNN and U-Net, respectively | |||||||||||
| [57] | √ | √ | Seven machine learning models and U-Net were used for the study, respectively | Image-guided surgery | ||||||||||
| [98] | √ | √ | SegNet and dense full convolutional neural networks are used | Eye diseases | ||||||||||