Assessment of a Radiomics-Based Computer-Aided Diagnosis Tool for Cancer Risk Stratification of Pulmonary Nodules

PROJECT SUMMARY/ABSTRACT This proposal describes a comprehensive 5-year mentored career development plan with the goal of training the candidate to become a leading independent physician-scientist focused on improving the diagnostic evaluation and management of pulmonary nodules (PNs) to optimize early detection of thoracic cancer and

minimize unnecessary harms to patients. The candidate is currently a Post-Doctoral Research Fellow and Attending Pulmonologist at the University of Pennsylvania (Penn). The proposal builds upon Dr. Kim’s previous research training in epidemiology and biostatistics and clinical experience in thoracic oncology. PNs are

commonly detected by computed tomography (CT). Lung biopsy, a highly invasive procedure, is required for a definitive diagnosis but carries significant risks and costs. Thus, clinicians face the diagnostic challenge of PN malignancy risk estimation when deciding which patients should undergo a biopsy, and which should be

surveilled with repeat imaging. The overall goal of this project is to address the current inadequacy of estimating malignancy risk within the diagnostic process of PN evaluation by assessing the clinical utility and effectiveness of a radiomics-based computer-aided diagnosis (CAD) tool. This novel technology synthesizes

quantitative features from raw CT imaging data invisible to the human eye and has been previously demonstrated by the candidate’s team to improve clinicians’ PN diagnostic accuracy. This project’s goal will be accomplished via three complementary specific aims. In Aim 1, a retrospective cohort study will be performed

to determine the clinical utility of a CAD-based risk stratification strategy using net reclassification indices, decision curve analysis, and relative utility curves. In Aim 2, a pilot, single-center pragmatic randomized clinical trial will be conducted to compare the clinical effectiveness of a CAD-based risk stratification strategy to usual

care for appropriate management of PNs, defined as biopsy or empiric treatment for malignant PNs and surveillance for benign PNs. Finally, in Aim 3, the cost-effectiveness of a CAD-based risk stratification strategy will be evaluated using decision analytic models for a simulated cohort of individuals with newly detected PNs.

Dr. Kim has outlined a rigorous training plan of coursework, skills acquisition (with a focus on clinical utility analysis, clinical trial design, decision analytic modeling, and cost-effectiveness analysis), and professional career development. To realize this vision, he has assembled a distinguished, multidisciplinary mentorship and

advisory team, led by his primary mentor, Dr. Anil Vachani, the Director of Clinical Research in the Section of Interventional Pulmonology and Thoracic Oncology at Penn, and co-mentor, Dr. Katharine Rendle, Deputy Director for Research at the Penn Center for Cancer Care Innovation. Penn provides an outstanding

intellectual, collaborative, and supportive environment for this proposal, positioning Dr. Kim to successfully complete the aims and training plan so that he will be a competitive applicant for an R01 award to perform a multicenter comparative effectiveness trial assessing PN malignancy risk stratification strategies.