Registration of Paired Inspiratory-Expiratory Lung CT Images Using Fissural Information

Longitudinal thoracic CT imaging receives high prevalence among pulmonary disease analysis and pulmonary image processing due to its allowance of tracking lungs within a respiratory cycle. A variety of disease biomarkers and clinical applications could be derived from the paired peak breath-holding image mapping information. However, it still remains challenging to align intra-patient multi-inflated lungs, especially the fissural areas when accurate lobe segmentation is unachievable. In this paper, a fully automated paired inspiratory-expiratory lung CT image registration method was presented. The method pipeline included: data preprocessing, a hybrid similarity measurement combined calculated fissureness, gradient distance transformation field and image intensity features, and a multi-resolution driven B-spline deformation. The proposed method was evaluated on a public standard platform to demonstrate its performance. Qualitative and quantitative experiments have shown that the proposed method is effective in reducing landmark registration errors and in aligning fissural areas without any specific involvement of pulmonary lobe labeling.