Fast Automated Segmentation and Reproducible Volumetry of Pulmonary Metastases in CT-Scans for Therapy Monitoring

The assessment of metastatic growth under chemotherapy belongs to the daily radiological routine and is currently performed by manual measurements of largest nodule diameters. As in lung cancer screening where 3d volumetry methods have already been developed by other groups, computer assistance would be beneficial to improve speed and reliability of growth assessment. We propose a new morphology and model based approach for the fast and reproducible volumetry of pulmonary nodules that was explicitly developed to be applicable to lung metastases which are frequently large, not necessarily spherical, and often complexly attached to vasculature and chest wall. A database of over 700 nodules from more than 50 patient CT scans from various scanners was used to test the algorithm during development. An in vivo reproducibility study was conducted concerning the volumetric analysis of 105 metastases from 8 patients that were subjected to a low dose CT scan twice within several minutes. Low median volume deviations in inter-observer (0.1%) and inter-scan (4.7%) tests and a negligible average computation time of 0.3 seconds were measured. The experiments revealed that clinically significant volume change can be detected reliably by the method.