Classification of Lung and Colon Cancer using a Hybrid CNN Model

One of the most dangerous diseases that people suffer from is Lung and Colon cancer. The huge number of fatalities caused by this disease forced doctors to adopt advanced methods to help them diagnose it early. In this paper, a technique based on machine learning is introduced to accelerate the process of Lung and Colon detection. The proposed method consists of using two state-of-the-art architectures called: Xception and MobileNet. First, we extract their feature maps and concatenate them. Second, we train the concatenated model from scratch. Finally, softmax layer is used to perform the classification task on Lung and Colon cancer images. This hybrid technique is applied on a challenging dataset called: LC25000 which consists of five classes with 5000 images per class. In the experiments, our proposed method achieved 99.37% accuracy which outperforms state-of-the-art methods. As a result, the new model can be used in clinics to help doctors.