An Effective and Robust Cancer Detection in the Lungs with BPNN and Watershed Segmentation

Lung cancer, a massively aggressive, quickly metastasizing and widespread disease, is the primary killer among both men and women worldwide. Regrettably, while the incidence of lung cancer decreased steadily in men over the past several years, it has increased alarmingly in women. In Computed Tomography (CT) lung cancer shows up as an isolated nodule. An Automatic Lung Cancer Detection System using improved Haar Wavelet Transform, Scale-Invariant Feature Transform (SIFT), Back Propagation Neural Network (BPNN), and Watershed Segmentation was proposed in this paper. Further, this work involves the usage of Bag of Visual Words (BOVW) based on K means Clustering to the extracted features from SIFT in the previous step. Later, classification is performed using BPNN which is a supervised learning algorithm from the field of Artificial Neural Networks (ANN). Finally, we detect the nodule in the cancerous lung image using watershed segmentation technique. The validation results have been proposed to be 91% accurate when compared to applying different algorithms.