We present a Deep Convolutional Neural Network based on VGG16 for the classification of patch images from the benchmark of PatchCamelyon 2016. This dataset consists of 327680 color images from histopathologic scans of lymph node sections. The images were originally acquired at 2 different hospital centers with a 40x objective.The main way to optimize the construction of our model was using a train, validation and test set, which is the standard approach to do it. We made changes to the architecture and its parameters one by one, observing its performance in the validation set in terms of accuracy and average loss.Finally, we could obtain a optimized CNN architecture that performed better than the VGG16 model by itself. Also, the AUĆs obtained in the valid and test set were 0.91 and 0.90 respectively, which shows the neural network was trained in a optimal way so that it did not perform overfitting. The final accuracy for valid was 89% and 86% for test set. This shows that our model is between those in the state of the art in classification and can be used for further harder problems such as semantic segmentation in the actual Camelyon dataset.