Table 1.
Summary of related work for COVID-19 detection.
| Cited | Methodology | Dataset | Results | Parameters | Limitations |
|---|---|---|---|---|---|
| Yujin et al. [12] | A Patch-based CNN method was developed based on ResNet-18 architecture. | JSRT + NLM | ACC = 88%, F1-score = 84%, SEN = 96%, SPE = 81% |
NIL | The method is considered computationally complex because of complex image pre-processing and classification steps. |
| Tulin et al. [21] | The DCN model was proposed. | OWN Datasets | ACC = 98%, F1-score = 96%, SEN = 95%, SPE = 95% |
678 Trainable | The limited number of COVID-19 samples used. |
| Fan et al. [22] | The inf-Net deep learning model was developed. | COVID-19 CT Collection | SEN = 87%, SPE = 97%. |
NIL | Only focused on lung infection segmentation rather than classifying COVID-19 patients. Complex approach to obtain multi-class label infection label. |
| Linda et al. [18] | The covid-Net deep learning model for detection. | COVIDx | ACC = 93% | 11.78 million | The method has degraded performance. |
| Ezz et al. [23] | The covidXNet model was proposed | OWN Datasets | ACC = 90% | NIL | Training complexity; several deep learning models were used to train. |
| Sakshi et al. [21,24] | Several pre-trained models were used in the study. | CT scan Dataset | ACC = 99%. SEN = 100%, SPE = 98% |
NIL | No effective evaluation of the deep learning model because only one train–test split strategy is used. |
| Ali et al. [25] | Several deep learning models are use in this work. | Dr Joseph Dataset, ChestX-ray8 Dataset, Kagge ChestXray Datasets |
ACC = 96%, ACC = 99%, ACC = 99% |
210.4 Millions | Several deep learning models used in the proposed work, making them complex. |
| Abbasian et al. [26] | A comparative study of 10 deep learning models was presented. | Own Datasets | ACC = 100% | NIL | Limited in terms of model complexity. |
| Lamia et al. [28] | A multi-level threshold and SVM method were proposed. | Montgomery County X-ray Set | ACC-97%, SPE = 99%, SEN = 95% |
NIL | Low generalizability problem. |
| Constantinou et al. [27] | Several deep learning models are used in this work. | Chest X-ray Image | PRE = 96%. RECALL = 96% ACC = 96% |
NIL | Low generalization problem. |
| Oguz et al. [30] | ResNet 50 and SVM Model. | CT Images | ACC = 96%, F1_score = 95% |
23 Million | Limited in terms of data used in work |
| Ieracitano et al. [31] | A fuzzy logic-based deep learning (DL) approach called CovNNet | CXR Images | ACC = 81% | NIL | Degraded performance |
| Chakraborty et al. [14] | DLM method is used to detect COVID. | CXR Images | ACC = 96%. SEN = 93% |
11 Million | Lack of model validation because of limited data. |
| Nasiri et al. [32] | DenseNet 169 and extreme gradient boosting | CXR Images | ACC = 98%. | NIL | Low sensitivity achieved in two-class problem |