Table 1. Summary of the related work.
| Author | Modality | Methodology | Discussion |
|---|---|---|---|
| Wang et al. (2021) | COVID-19 Images | GCN + CNN | The fusion of both individual image-level features and relation-aware features to produce Graph Convolutional Networks (GCN), and CNN respectively |
| Wang et al. (2020) | kurtosis map | Hybrid decomposition of NSCT and morphological sequential toggle operator (MSTO) | Their methodology extracted major feature information of the source images and preserved the unambiguous edges with a little produced noise in both visible and infrared image fusion |
| Chandra et al. (2020) | CXR images | gray level co-occurrence matrix (GLCM) | extract shape feature from CXR images based on gray level co-occurrence matrix (GLCM) with an improved abnormality detection |
| Bhandary et al. (2020) | CXR images | CNN and PCA | The combination of one-dimensional feature vectors and the dimensionality reduction is performed using PCA are then applied to the source CXR images and tested for the normal bacterial pneumonia |
| Ozkaya, Ozturk & Barstugan (2020) | CT images | DCNN | pre-trained CNN to fuze different subsets and transfer learning classifiers to classify COVID-19 cases |
| Haskins, Kruger & Yan (2020) | MR, CXR, and CT modality. | DCNN | To visualize different registration of essential data of the source images using fixed and moving data labels as well as fixed, and moving images |
| Lin et al. (2020) | CT, MR, and PET scan | stacked convolution neural network (DSCNN) for multi band images | DSCNN and NSCT fuze multiband images reconstructed by long short-term memory (LSTM) and DSCNN to overcome the data-driven approach's controllability problem. |
| Huang et al. (2020b) | MRI-CT | NSCT and DCNN | (NSCT) by which the fusion technique able to avoid the spectral aliasing and provide more characteristic of the invariance translation |
| Maharjan et al. (2020) | Brain Tumor CT | ELM and NSCT | Detect brain tumor using NSCT and extreme learning machinery (ELM]. |
| Pereira et al. (2020) | CXR images COVID-19 |
multi-class hierarchical CNN |
The fusion strategy based on weighted sum, weighted product, and the voting strength of the enrolled features. They achieved average F1-Score of 0.65, and 0.89 multi-class, and hierarchical classification respectively. |
| Panwar et al. (2020) | Chest X-ray and CT-Scan images of COVID-19 cases | deep learning and grad-CAM based color visualization | Color visualization approach to make the deep learning model more interpretable and explainable |
| Chen et al. (2020) | Multi-label CXR image | self-adaptive weighted fusion scheme |
Contextual information in both global and lung field with the mean area under the curve AUC = 0.82 |
| Chowdhury, Rahman & Kabir (2020) | 2905 chest X-ray images COVID-19 |
PDCNN+CNN | Accuracy = 96.58 for COVID-19, Normal, and Pneumonia cases |
| Bashir et al. (2019) | CXR, CT, and MRI | multimodal imagery based on using SWT and principal component analysis (PCA) | A dimensionality reduction is performed using PCA and then SWT to extract features |
| Lee, Lee & Kang (2018) | CXR images | Poisson-Gaussian noise analysis | NSCT in fusion strategy, the noise distribution produced by CXR images |
| Rajalingam, Priya & Bhavani (2018) | CXR and CT | NSCT and DCNN | Hybrid multimodality medical image fusion are applied in both CXR and CT images |
| Li et al. (2017) | MRI and PET images | Multi-scale transformation coefficients to produce a fuzed image with inter-scale correlation | They apply MRI and PET images to observe the objectives of fusion performance and determine the source images' miss-registration |