Discuss about the helping of deep learning on diagnosis cancer?

Table 4. shows the quantitative results for prostate cancer classification. Similar to colorectal cancer classification, MSBP-Net-ResNet achieved accurate classifications: 84.58 Accuracy. In the comparative experiments, the single scale network (ResNet) was substantially inferior to MSBP-Net-ResNet;

for instance

0.03 in specificity

also

outperformed the fusion networks in all the evaluation metrics.

Furthermore

, MSBP-Net-MobileNet was superior to the single-scale network (MobileNet);

for example

, 0.09 increase in Macro F1. As it is compared to the fusion networks, MSBP-Net-MobileNet provided a substantial performance gain in the entire evaluation metrics: 2 increase in three metrics related to accuracy as well as >0.05 increase in sensitivity. As we observed in colorectal cancer classification, MSBP-Net-ResNet was the best performing network, achieving the best performance in all 8 evaluation metrics.