Efficient Computerized Lung Cancer Detection Using Bag of Words

Abstract

The automatic detection of diseases in the medical field is growing very fast nowadays. It is widely being accepted as this system can reduce the burden on doctors. Among the available examination of diseases, attention on lung cancers is required more as these play a major role in increasing the mortality rate in the present day. Though many computerized cancer detection techniques were proposed earlier, those techniques gets failed in managing the better accuracy rate due to their combinations of filtering techniques, segmentation techniques, and classifiers. An MLP-BPNN(Multi-Layered Perceptron Back Propagation Neural Network based on SIFT(Scale Invariant Feature Transform) feature extraction along with Bag of Words(BOW) is proposed which gives the better accuracy rate of 89% when compared to any other Cancer detection technique proposed earlier. Lung images of 300 are collected from the Rajiv Gandhi Cancer Institute and Research Centre, Delhi as a dataset out of which 100 images are used for testing and 200 images are used for training.