Table 4 Comparison of our proposed GraphCovidNet model with some previous works on all the datasets (Oh et al.30, Chandra et al.37, Nour et al.3, Hemdam et al.45, Turkoglu et al.26 have combined other dataset; Oh et al.30, Chandra et al.37, Hemdam et al.45 have considered the first dataset only; Nour et al.3, Turkoglu et al.26 have considered the second dataset only).
Dataset | Authors | Methodology | Accuracy (%) | Precision (%) | Recall (%) | F1-score (%) |
|---|---|---|---|---|---|---|
SARS-COV-2 Ct-Scan Dataset8 | Soares et al.8 | xDNN | 97.38 | 99.16 | 95.53 | 97.31 |
Pedro et al.15 | EfficientNet with transfer learning | 98.99 | 99.20 | 98.80 | 99 | |
Proposed | GraphCovidNet | 100 | 100 | 100 | 100 | |
COVID-CT dataset11 | Pedro et al.15 | EfficientNet with transfer learning | 87.68 | 93.98 | 79.59 | 86.19 |
Yang et al.11 | Segmentation masks with CSSL | 89.1 | – | – | 89.6 | |
Proposed | GraphCovidNet | 100 | 100 | 100 | 100 | |
covid-chestxray-dataset23+Chest X-Ray Images (Pneumonia) dataset24 | Makris et al.53 | VGG16 and VGG19 with transfer learning | 95.88 | COVID-96 normal-95 Pneumonia-95 | COVID-96 normal-100 Pneumonia-91 | COVID-98 normal-98 Pneumonia-98 |
Elaziz et al.21 | MRFO + KNN | 96.09 | 98.75 | 98.75 | 98.75 | |
Zhong et al.54 | VGG16 based CNN model | 87.3 | 89.67 | 84.4 | 86.96 | |
Oh et al.30 | DenseNet103 for segmentation + ResNet-18 | 88.9 | 83.4 | 85.9 | 84.4 | |
Chandra et al.37 | Majority voting of SVM, KNN, DT, ANN, NB | 93.41 | – | – | – | |
Nour et al.3 | CNN for feature extraction + SVM | 98.97 | – | 89.39 | 96.72 (F-score) | |
Hemdam et al.45 | VGG19 or DenseNet201 | 90 | COVID-83 Normal-100 | COVID-100 Normal-80 | COVID-91 Normal-89 | |
Turkoglu et al.26 | AlexNet+ Relief feature selection algorithm and SVM | 99.18 | 99.48 | 99.13 | 99.30 | |
Proposed | GraphCovidNet | 99.84 | 99.84 | 99.84 | 99.84 |
