An Enhanced Technique of COVID-19 Detection and Classification Using Deep Convolutional Neural Network from Chest X-Ray and CT Images

Table 7

Performance comparisons between the IDConv-Net and state-of-the-art models on CT images.

Model	Dataset	Precision	Recall	-score	Training accuracy	Testing accuracy
Noisy-OR Bayesian [64]	Transverse-section CT	86.87	86.67	86.70	87.05	86.70
Decision fusion [65]	COVID-CT	88.14	88.79	86.70	89.78	88.34
Ensemble [66]	CT (HRCT)	90.63	93.55	92.06	92.83	91.94
DenseNet [67]	HRCT images	96.00	97.00	93.00	91.36	92.00
DenseNet161 [68]	COVID19Net	85.39	77.55	81.28	84.36	82.76
Ensemble (Hard voting) [68]	COVID19Net	82.80	78.57	80.63	82.88	81.77
Ensemble (Soft voting) [68]	COVID19Net	83.70	78.57	81.05	84.21	82.27
VGG16-based DL [69]	Hospital of Tabriz Data	91.50	89.78	90.63	91.68	90.14
3D-ResNets with attention [63]	Several cooperative hospitals	86.27	92.33	85.20	—	93.30
U-NET [70]	CT segmentation dataset	96.38	96.04	86.10	95.93	95.60
Proposed IDConv-Net model	Merged 2D-CT	98.64	98.31	98.48	99.53	98.41