Persistent homology analysis of longitudinal computed tomography fibrotic features in COPD

Fibrosis may coexist with emphysema in COPD, but computed tomography (CT) quantification is challenging. Persistent homology (PH), a topological data analysis technique, provides interpretable structural features in greyscale images. By using PH, this CT study aimed to quantify fibrotic lesions in nonemphysematous and emphysematous lungs and to investigate clinical implications of PH-based fibrosis quantification in patients with COPD.

The study included subjects from the Lung Cancer Screening (LCS) cohort (n=346) and two prospective COPD cohorts (Kyoto University n=234; Kyoto-Himeji n=166). Based on CT value patterns and spatial topology, PH assigned each voxel as fibrotic or nonfibrotic and calculated the percentage of fibrotic lung volume (PH-fibrosis%) in association with visually identified interstitial lung abnormality and COPD outcomes.

Higher PH-fibrosis% was associated with interstitial lung abnormalities in the LCS and Kyoto University cohorts. The two COPD cohorts consistently showed significant associations between higher baseline PH-fibrosis% and future exacerbation risk independent of emphysema and airway wall thickness (hazard ratio (HR) 1.39 and 3.35 for Kyoto University and Kyoto-Himeji, respectively). In the Kyoto University cohort, higher PH-fibrosis% was significantly associated with increased mortality (HR 1.94), with a similar trend in the Kyoto-Himeji cohort (HR 1.86). Longitudinal increases in PH-fibrosis% over median 4.98 years were associated with higher exacerbation frequency and patients experiencing greater increase in PH-fibrosis% subsequently exhibited higher mortality in the Kyoto University cohort.

PH can be used to quantify fibrotic lesions on CT and PH-fibrosis% could be a prognostic imaging marker in patients with COPD.