Lung cancer detection is one of the most important goals of medical diagnosis. To detect lung nodules usually classical X-ray and/or computed tomography (CT) images are used. The progression of the disease can be monitored if the doubling time of the volume of a pulmonary nodule is determined and followed, which means that the volume of a nodule has to be measured/estimated. To measure the nodule volume using classical X-ray images is almost impossible, while a CT-based diagnosis is rather expensive. Recently a new radiological image-based modality, digital tomosynthesis (DTS) has been developed. DTS can be considered as a 2.5D modality, where coronal slice images of a chest can be computed.
The spatial resolution of a DTS image is much higher than that of a CT image, while the thickness of a slice is larger compared to a CT image. Thus DTS can also be used to determine lung nodule volumes although -because of the 2.5D reconstruction – volume estimation is a rather hard task. This paper proposes a new way for estimating nodule volume. The method was developed using an experimental database, which contains reconstructed images of 16 simulated small, elliptical nodules, placed into an anthropomorphic chest phantom, and was evaluated by a few real DTS images and animage base made from simulated projections from a public CT database.