Radiomics for severe asthma phenotyping and endotyping

Background: Non-invasive measures of airway remodeling are essential for further understanding severe asthma pathophysiology. Aim: We investigated whether chest high resolution computed tomography (CT) could be used for asthma phenotyping and endotyping in 105 adults from the severe asthma European U-BIOPRED study. Methods: Seventy-three severe asthma ex-smokers [45/28 f/m, age, 51±12 years, FEV1 67.6±25.5% pred, mean±SD), 20 severe asthma current smokers (54±9 years, 10/10 f/m, FEV1 69.3±21.0% pred) and 12 mild-to-moderate asthma non-smokers (4/8 m/f, 41±17 years, FEV1 92.3±16.3% pred) were studied. Lung CT imaging was acquired at total lung capacity and residual volume, 25 parameters were quantified (APOLLO platform, VIDA). CT and serum proteomics datasets were integrated using similarity network fusion; spectral and consensus clustering were applied. Results: Three clusters were identified (Figure 1). Compared with cluster 2 (n=34) and 3 (n=41), cluster 1 (n=30) showed a female prevalence (p<0.005) and higher BMI (p<0.01); FEV1 (p<0.0001) and FVC% predicted were lower in cluster 3 (p<0.002). Serum leptin and complement factor I were elevated in cluster 1 (FDR<0.05). Forty-three serum proteins were differentially expressed between clusters 2 (higher NOTCH signaling) and 3 (higher Insulin-like Growth Factor 1 Receptor signaling). Conclusions: Preliminary data show that integration of radiomics and serum proteomics datasets enables severe asthma endotyping.