Comparing different deep learning backbones for segmentation of lung nodules

Deep Learning has made great impact on computer vision applications such as object detection, image classification, segmentation and localization. Computer based system for lung cancer detection and diagnosis helps in detecting the lung nodules and analysing the nodule features for further segmentation and classification. This paper explains the effect of UNet model on various back bone architectures pretrained on ImageNet dataset. This work was carried out on LIDC-IDRI dataset and the efficiency interms of IoU score and accuracy both in training and validation datasets are listed. Results show that EfficientNet backbone outperforms other backbone architectures.