Colorectal Polyp Detection Using Feedforward Neural Network with Image Feature Selection

Colorectal cancer is one of the most common cancers in the world and accounts for approximately 700,000 deaths annually. Early detection is one of the keys to detecting polyps before transforming into malignant colorectal cancer. Along with the development of computing technology, there are many techniques utilizing image processing and computing algorithms method to identify the polyps due to increased data processing speed and also faster computational algorithms. Polyp detection is a very challenging problem because polyp detection involves many factors. This research tries to perform a new improvisation in polyp detection process in terms of pre-processing stage and classification stage by utilizing statistical feature selection technique from collection bag of features combined with feedforward neural network methods. Based on experiments that have been done it is found that the proposed method is able to provide the accuracy on polyp detection at 97.85% with 6-(12)-2 neural network architecture.