Skip to main content
. 2021 Mar 14;13(6):1291. doi: 10.3390/cancers13061291

Figure 1.

Figure 1

Block diagram of the system. The image preprocessing stage starts with patching the image and continues with the augmentation of the image patches. Pre-trained Inception-ResNet-v2 was retrained with these images. Classification and heat map generation were performed. Four pre-trained networks were independently used to analyze the results. Each network has a different number of layers (ResNet101→347-layer, VGG16→41, Inception v3→315-layer, and Inception ResNet v2→824-layer) with variation in frozen layers (ResNet101→79-layer, VGG16→no frozen, Inception v3→307-layer, and Inception ResNet v2→818-layer) and fine-tuning of middle layer numbers; we represent them as X and Y in the figure). CNN: convolutional neural network.