Watershed Segmentation with CAFIS and RCNN Classification for Pulmonary Nodule Detection

Abstract

The premier cause for high cancer deaths in the world is the lung cancer. It is important for the radiologists to predict lung cancer at an early stage. Various research works on such nodule segmentation clearly manifest that they are ineffective. These investigations brought to the development of watershed segmentation-based topological interpretation (WSBTI) and an advanced optimized segmentation method of nodule detection. The supreme goal of this research work is to effectively recognize small anomalous nodule in the lung region. The noise discrimination can be eradicated by the primary step adaptive median filter with discrete-time complex wavelet transform enhancement technique. In the consequent step, WSBTI algorithm is effectively implemented for the prediction on abnormal node of lung. Ultimately, the procurement of nodule can be done by using coactive adaptive neuro fuzzy interference system classifier (CAFIS) and recurrent convolutional neural network classifier (RCNN). The average time of segmentation is 1.05 s. The high accuracy classification is 97% by using CAFIS method and 97.6% by RCNN method.