Voxelwise lung pattern transition scores on HRCT images and their association with symptoms in patients with idiopathic pulmonary fibrosis

Objectives: To investigate regional transitions of interstitial lung disease patterns on HRCT and their association with symptom scores in patients with Idiopathic Pulmonary Fibrosis (IPF). This work uses mapping of voxels between baseline and follow-up scans to assess highly localized transitions in lung patterns. Methods: Anonymized HRCT images were obtained from 25 IPF subjects who had a baseline and follow-up surveillance HRCT scan at 6 months. A transitional interstitial lung disease (TILD) matrix of pattern transitions was computed for each paired of subject scans using the following steps: (1) lobar segmentation and classification of each voxel as one of 4 lung patterns (fibrotic, honeycomb, ground glass, or normal); (2) registration (mapping) of corresponding voxels between the paired scans; (3) for each mapped pair of voxels recording a count in the appropriate cell of the 4x4 TILD matrix; (4) summarizing the pattern transitions as percentages of total voxels (TILD scores). Associations between TILD scores and symptom scores, as well as comparisons between worse and non-worse groups, were performed. Results: TILD scores indicating worsening from ground glass to fibrotic patterns were associated with increasing the shortness of breath (rho=0.48; p=0.016). Significant differences in TILD scores were found between worse and non-worse groups in the shortness of breath (p=0.014). When symptoms were not worse the TILD scores were minimal in the right and left upper lobes. Conclusions: Using voxel-by-voxel TILD scores on paired HRCT scans 6 months apart, there were obvious transitions to worsening lung patterns when symptoms got worse.