Variational approach for small-size lung nodule segmentation

This paper describes a novel variational approach for segmentation of small-size lung nodules which may be detected in low dose CT (LDCT) scans. These nodules do not possess distinct shape or appearance characteristics; hence, their segmentation is enormously difficult, especially at small size (≤ 1 cm). Variational methods hold promise in these scenarios despite the difficulties in estimation of the energy function parameters and the convergence. The proposed method is analytic and has a clear implementation strategy for LDCT scans. We show the effectiveness of the algorithm for segmenting various types of nodules regardless of their location in the lung tissue.