An end-to-end framework for pulmonary nodule detection and false positive reduction from CT Images

Abstract

The automatic pulmonary nodule detection in thoracic computed tomography (CT) scans plays a crucial role in the early diagnosis of lung cancer. The automated lung nodule detection is challenging due to the high variance in appearance and shape of the targeting nodules. To overcome the challenge, we present an end-to-end effective 3D nodule detection framework. It involves two networks: an U-Net equipped with ResNet block as well as region proposal generation module for suspicious nodule detection, and a CNN with focal loss for false positive reduction. Extensive experiments on LUNA16 datasets to validate the proposed framework. Our experiments demonstrate that our model achieves an average sensitivity (84.2%) and FPs/scan (1.14), respectively. The experimental results have indicated the effectiveness of the proposed improvements and the potential of our proposed method for real clinical practice.