Malignancy risk estimation of screen-detected pulmonary nodules

S. van Riel

• Promotor: B. van Ginneken and C. Schaefer-Prokop
• Copromotor: C. Jacobs
• Graduation year: 2020
• Radboud University, Nijmegen

Abstract

Lung cancer is still the leading cause of cancer-related mortality worldwide in both men and women. Since patients only develop clinical symptoms in an advanced stage of the disease, most lung cancers are diagnosed in a late stage which has a vastly lower survival rate than early stage disease. Lung cancer screening through CT scans can help to detect early stage lung cancer, which manifests itself as a pulmonary nodule, a round focal opacity in the lungs. Radiologists need to differentiate malignant from benign nodules on these CT scans to recommend the most appropriate follow-up procedure. In order to aid radiologists and to standardize the interpretation, several categorical systems and computer models have been developed for the risk assessment of pulmonary nodules. This thesis presents the evaluation of the performance of human observers and computer models to characterize pulmonary nodules mainly with respect to differentiating malignant from benign nodules.