Detection of pulmonary nodules on CT and volumetric assessment of change over time

Abstract

We present a computer vision system that automatically detects pulmonary nodules in computed tomography (CT) scans of oncology patients, performs size analysis andassesses for change in volume over time. Thresholding, backtracking, and smoothing algorithms have been developedt o recognize the thorax and trace the lung border. The regions within the lung that potentially contain nodules are evaluated for their shape, size, and position. These candidate regions are then characterized as nodules versus other structures by comparing consecutive CT slices in the same study. A preliminary system for the registration of studies has also been developed. It estimates nodule volume in each study and evaluates the volumetric change over time. Our system has been testedon initial andfollo w-up studies of four patients. Preliminary results were detection of 284 nodules ranging between 1 and32 mm at various locations and assessment of their volumetric change over time. Our techniques have future applications for determining disease progression, remission, and stability in oncologic patients in addition to coregistration of different modalities within the thorax.