FIGURE 4.

Architecture of the BC classification based on the gemcitabine resistance level. The proposed CNN mainly comprised three sets of convolutional layers and fully connected layers. The network parameters were trained by minimizing the cross-entropy loss function using the Adam optimizer. In this process, we fine-tuned the learning rate using predefined schedules, including exponential or step decay. The trained network was ultimately tested with unseen data to validate its effectiveness in predicting the four levels of gemcitabine resistance from the BC cell images. CNN: convolutional neural network; BC: bladder cancer.