Lung Cancer Identification from CT Scans using a Soft-attention enabled Deep Transfer Learning Model

Being a major contributor to the global mortality, lung cancer, owing to its fatality, calls for early and efficient diagnosis. This requires the employment of efficient computer aided diagnostics for determining the computed tomography (CT) scans as indicating lung cancer or not. Most such systems leverage the high efficiency of deep learning models to identify the lung cancer type from an input CT scan. In this work, a deep transfer learning model equipped with a soft attention block is framed to classify CT scans as either lung cancer affected or normal. Public datasets such as IQ-OTH/NCCD along with LIDC-IDRI are adopted for evaluation of the proposed technique which scored accuracies of 99.55% and 95.41% respectively, that are comparable with those of some recent standard techniques.