Semantic context-aware attention UNET for lung cancer segmentation and classification

摘要

Lung cancer is a serious type of cancer, leading to increased mortality to deathin both men and women as the symptoms are noticed only at later stages. Lifespan of individuals' may be extended if lung cancer is detected in its earlystages. One of the imaging modalities used to diagnose lung cancer is computedtomography (CT). A nodule or mass is a small abnormal growthobserved in a lung CT scan that, in most cases, may turn out to be benign. Acomputer-aided system is essential to help physicians in precisely diagnosingthe disease. The main objective of this work is to detect and classify the nodulesin the lung CT scan images as benign or malignant. A context-awareattention UNET architecture is proposed to segment the nodule from the lungCT scan image. Further, the segmented nodule is classified as benign or malignantusing a Convolutional Neural Network architecture. The experiments areperformed using the LUNA 16 and LIDC-IDRI lung CT scan image datasets.From the results obtained, it is observed that the context-aware attentionUNET shows a noteworthy improvement in the following metrics: Dice Score,Sensitivity, Specificity, and F-Measure. A significant improvement is obtainedcompared to the existing systems in detecting the lesion as a benign nodule ormalignant nodule. Further, an ablation study is performed to validate the significanceof each component in the architecture. The experimental resultshave reported 98.81 and 99.15 for specificity and sensitivity, respectively,and therefore the proposed system has potential clinical value in the detectionof lung cancer.