A Computational Imaging Method for Small Airways Disease Staging in COPD

This is the authors' abstract. We don't add key points for this paper.

RATIONALE: COPD is characterized by airflow limitation due to narrowing and ultimately loss of terminal airways (small airways disease, SAD) and/or irreversible alveolar destruction (emphysema). To detect SAD in the presence of emphysema, we developed a quantitative CT-based technique called Parametric Response Mapping (PRM). SAD detected by PRM (‘PRM SAD’) correlates with airway narrowing, obstruction and loss, as measured using microCT of lung explants from end-stage COPD patients. However, in early disease, some PRM SAD can revert to radiologically normal tissue, suggesting some SAD in early disease may represent reversible bronchoconstriction, not permanent narrowing and/or loss. We have developed a model based on elastic-principal graphs that better captures SAD evolution. Our approach of “SAD Staging” provides a more precise readout of SAD that will enhance detection of local disease and improve prognosis for early COPD. METHODS: Demographic, HRCT, PFT and mortality data were obtained from 8856 participants from Phase 1 of the COPDGene longitudinal study. Of these, 3636 had Phase 2 data (∼5 years later). PRM was generated from all paired HRCT. SAD Staging was performed at a sub-volume of 163 voxels. For each sub-volume in the lungs, relative volumes of PRM classes were determined. Using these PRM values, the sub-volume was projected onto our model to determine its stage assignment and relative position. This approach was applied to all sub-volumes to generate a 3D SAD Staging Map (see Staging Workflow, Figure 1A). SAD staging evolution was determined using PRM data at year 0 aligned to year 5. We evaluated SAD stages as surrogates of mortality using ROC analysis and Cox regression. RESULTS: The relative lung volume of stage 1 increased from a mean of 0.1 (0.11 SD) to 0.23 (0.14 SD) in “at-risk” and GOLD 1 subjects, with no change at higher GOLD scores. In contrast, stages 2 and 3 increased with increasing GOLD. Figure 1B illustrates stage evolution of a representative GOLD 2 subject with sub-volumes progressing to increased stages over 5 years. Subjects with a baseline mean relative position of Stage 1 greater than 0.47 (Stage1p0.47) had poorer survival than those < 0.47 (Figure 1C). CONCLUSIONS: We developed a method to track CT-based lung SAD evolution at the local level. SAD Staging may equip healthcare professionals with the capability of identifying local SAD earlier in asymptomatic individuals at risk for COPD and may suggest a role for expiratory CT images in lung cancer screenings.