Skip to Main Content
About this Publication
Title
Joint Learning for Deformable Registration and Malignancy Classification of Lung Nodules
Digital Object Identifier
ISBN-13
978-1-6654-1246-9
ISBN-10
1-6654-1246-1
Publication
IEEE. 2021 May 25; Pages 1807-1811
Authors

Aryan Ghazipour
Medical Imaging Lab, University of Louisville, Louisville, KY, USA
Benjamin Veasey
Medical Imaging Lab, University of Louisville, Louisville, KY, USA
Albert Seow
University of Louisville, Louisville, KY, USA
Amir A. Amini
Medical Imaging Lab, University of Louisville, Louisville, KY, USA

Abstract

The use of low-dose Computer Tomography (CT) has been effective in reducing the mortality rate of lung cancer. With the rapid increase in the number of studies, computer aided diagnosis (CAD) systems need to be developed to further assist radiologists in detecting and classifying lung nodules in CT scans. In this paper we propose a new system based on Deep Convolutional Neural Networks, U-net, and ResNet architectures. This network performs the dual task of registration of lung nodule scans in two time points and additionally performs benign/malignant classification. The training and testing data were put together based on a subset of data from the National Lung Screening Trial (NLST), referred to as NLSTx which has biopsy confirmed diagnoses for the nodules. The combination of deformable registration and binary classification performed by the network will increase the usefulness of this CADx system allowing both measurement of the growth and classification of the nodule.

Related CDAS Studies
Related CDAS Projects