A Dynamic Model of Average Lung Deformation Using Capacity-Based Reparameterization and Shape Averaging of Lung MR Images

We present methods for extracting an average representation of respiratory dynamics from free-breathing lung MR images. Due to individual variations in respiration and the difficulties of real-time pulmonary imaging, time of image acquisition bears little physiologic meaning. Thus, we reparameterize each individual’s expiratory image sequence with respect to normalized lung capacity (area, as a substitute for volume), where 1 represents maximal capacity and 0 minimal capacity, as measured by semi-automated image segmentation. This process, combined with intra-subject pairwise non-rigid registration, is used to interpolate intermediate images in the normalized capacity interval [0,1]. Images from separate individuals with the same normalized capacity are taken to be at corresponding points during expiration. We then construct an average representation of pulmonary dynamics from capacity-matched image sequences. This methodology is illustrated using two coronal 2-D datasets from healthy individuals.