ColoRectalCADx: Expeditious Recognition of Colorectal Cancer with Integrated Convolutional Neural Networks and Visual Explanations Using Mixed Dataset Evidence

Table 9

Comparison of accuracies of CNN+LSTM architectures.

Individual CNN	CNN+LSTM training accuracy in %	CNN+LSTM testing accuracy in %
AlexNet	86.81	77.56
DarkNet-19	83.46	68.51
ResNet-50v2	79.57	83.21
DenseNet-201	87.1	84.7
EfficientNet-B7	67.53	77.56
VGG-16	72.7	71.76
VGG-19	69.46	70.16
NasNetLarge	83.57	78.1
InceptionResNetV2	84.52	80.36
Integrated CNN	CNN+LSTM training accuracy	CNN+LSTM testing accuracy
ADaRDEV^2I-22	81.43	81.9
ADaRDEV^2-22	78.56	78.56
RDEV^2-22	79.54	77.01
ADaDR-22	84.61	82.17
ADaR-22	73.23	70.57
DaRD-22	76.84	79.56
DEV-22	64.82	67.98
ADa-22	62.32	60.19
RD-22	73.1	64.6
RV-22	36.76	41.43
AD-22	78.15	67.83
DaR-22	74.26	72.94