A python implementation for lung tumor deep feature extraction and classification with ResNet-18. With the original ResNet-18 structure, the number of the output of its last fully-connected layer is changed from 1000 to 2 to classify whether a region has a tumor or not. The data sets used in our experiments come from the 2018 IEEE SPS Video and Image Processing (VIP) Cup competition. Please contact me if you need the training data sets.