Correction of lung boundary using the gradient and intensity distribution

Abstract

We propose a new method for correcting the segmented lung boundary in expiratory and inspiratory CT. First, the initial lung boundary is extracted by using density-based segmentation. Second, the scope for the boundary propagation is computed by generating and analyzing the gradient profiles with an adaptive length. The definition of the scope helps to prevent the leakage outside the scope and improves the efficiency of the propagation. Finally, the boundary is propagated within the defined scope using a speed function. The speed function is based on the gradient and intensity distribution and prevents the boundary from converging to the local gradient maxima. The results of the lung boundary correction are evaluated by visual inspections, accuracy evaluations and processing time. Experimental results show that the proposed method corrects the lung boundary reliably and reproducibly.