Lung_RUNET: a Segmentation Framework for Lung Nodules
P. Malin Bruntha, S. Dhanasekar, L. Jubair Ahmed, V. Govindaraj, S. Immanuel Alex Pandian, Siril Sam Abraham
Abstract
In the modern-day world, deaths due to cancer are on the increase alarmingly. Among various types of cancer, lung cancer is very deadly. If detected early, however, the mortality rates due to lung cancer can be brought down drastically. Lung nodule segmentation is now possible using image analysis of Computed Tomography images of lung thoracic regions. The present study suggests a new model namely Lung Residual UNet (Lung_RUNET) based on UNet architecture for the effective segmentation of lung nodules. The proposed model was checked on the publicly available LIDC-IDRI dataset. The results were compared with the UNet model and have shown significant improvement in segmentation with 87.2% in DSC, 81.1% in sensitivity and 89.1% in precision.