Detection of Lungs Cancer Through Computed Tomographic Images Using Deep Learning

Lung cancer has become a particularly lethal disease in the last decade. Lung cancer is the second most common cause of death for women and the primary cause of death for men. Therefore, early detection of lung knobs is one of the most effective ways to treat lung infections. Similarly, computer-aided diagnosis (CAD) of lung knobs has gotten a huge interest over the last decade. As a result of the broad variety of lung knobs and the complications of the entire environment, developing a robust knob detection approach is extremely difficult. A convolutional neural network (CNN) based framework is proposed to detect tumors that are identified as risky or benign in lung disease screening using CT images. Two publicly available datasets LUNA-16 and LIDC are employed to detect lung cancer. The dataset is augmented to maximize the volume of images in it. Also, preprocessing is done on CT images for better noise removal. Additionally, segmentation is performed to specify the infected area. Three pre-trained architectures, DenseNet, AlexNet, and VGG-16, are utilized to classify the cancerous and normal images. The DenseNet classifier achieved 98% classification accuracy, 98.93% sensitivity, and 99% specificity, which exhibits outstanding performance than other classifiers. The efficient results of the proposed framework show better performance than existing state-of-art studies.