Project Summary/Abstract Lung cancer is the leading cause of cancer death and one of the most common cancers among both men and women in the United States. Recent advances in high-resolution imaging set the stage for radiomics to become an active emerging field in cancer research. However, the promise of radiomics is limited by a lack of image standardization tools, because computed tomography (CT) images are often acquired using scanners from different vendors with customized acquisition parameters, posing a fundamental challenge to radiomic studies across sites. To overcome this challenge, especially for large-scale, multi-site radiomic studies, advanced algorithms are required to integrate, standardize, and normalize CT images from multiple sources. We propose to develop STAN-CT, a deep learning software package that can automatically standardize and normalize a large volume of diagnostic images to facilitate cross-site large-scale image feature extraction for lung cancer characterization and stratification. By precisely mitigating the differences in advanced radiomic features of CT images, STAN-CT will overcome research silos and promote medical image resource sharing, ultimately improving the diagnosis and treatment of lung cancer. Our goal will be achieved through two Aims. In Aim 1, we will develop a working prototype to standardize CT images. First, we will collect raw image data from lung cancer patients and reconstruct CT images using multiple image reconstruction parameters, and we will scan a multipurpose chest phantom along with five different nodule inserts. Then, we will develop and train STAN-CT for CT image standardization. An alternative training architecture will be developed to achieve the improved model training stability. In Aim 2. We will deploy and test STAN-CT for image standardization locally and across three medical centers. First, we will make the STAN-CT software package available to the public by providing a menu-driven web-interface so that that users can conveniently convert medical images that were taken using non-standard protocols to one or multiple standards that they specify. Second, we will deploy STAN-CT at the University of Kentucky for local performance validation. We will test the functionality, reliability, and performance of STAN-CT using both patient chest CT image data collected at large-scale and the phantom image data, both independent to training. Third, we will deploy and test STAN-CT at the University of Kentucky as well as the University of Texas Southwestern Medical Center and Emory University for cross- center performance validation. We will use the same multipurpose chest phantom and both standard and non- standard protocols to validate STAN-CT at the three centers. We will test the generalizability of STAN-CT using clinical CT images of human patients and will determine whether a model trained using the data from one medical center are applicable for images collected at another place. Finall...