Contrastive learning and subtyping of post-COVID-19 lung computed tomography images

Abstract

Patients who recovered from the novel coronavirus disease 2019 (COVID-19) may experience a range of long-term symptoms. Since the lung is the most common site of the infection, pulmonary sequelae may present persistently in COVID-19 survivors. To better understand the symptoms associated with impaired lung function in patients with post-COVID-19, we aimed to build a deep learning model which conducts two tasks: to differentiate post-COVID-19 from healthy subjects and to identify post-COVID-19 subtypes, based on the latent representations of lung computed tomography (CT) scans. CT scans of 140 post-COVID-19 subjects and 105 healthy controls were analyzed. A novel contrastive learning model was developed by introducing a lung volume transform to learn latent features of disease phenotypes from CT scans at inspiration and expiration of the same subjects. The model achieved 90% accuracy for the differentiation of the post-COVID-19 subjects from the healthy controls. Two clusters (C1 and C2) with distinct characteristics were identified among the post-COVID-19 subjects. C1 exhibited increased air-trapping caused by small airways disease (4.10%, p = 0.008) and diffusing capacity for carbon monoxide %predicted (DLCO %predicted, 101.95%, p < 0.001), while C2 had decreased lung volume (4.40L, p < 0.001) and increased ground glass opacity (GGO%, 15.85%, p < 0.001). The contrastive learning model is able to capture the latent features of two post-COVID-19 subtypes characterized by air-trapping due to small airways disease and airway-associated interstitial fibrotic-like patterns, respectively. The discovery of post-COVID-19 subtypes suggests the need for different managements and treatments of long-term sequelae of patients with post-COVID-19.

Lung pathophysiology in patients with long COVID-19: one size definitely does not fit all

“…However, this is in contrast with the findings of R inaldo et al [ 13 ] and B eaudry et al [ 5 ], who did not find any association between dyspnoea and cardiopulmonary impairment or lung function parameters in COVID-19 patients 3–6 months post-infection. Indeed, L i et al [ 19 ] demonstrated the presence of two different latent phenotypes using a volume-independent contrastive learning model with inspiratory and expiratory chest computed tomography (CT) images in 140 post-COVID-19 patients, one with normal volumes and D LCO but CT signs of air trapping and one with reduced volumes and D LCO but only an interstitial fibrotic like pattern.…”

Is Pulmonary Involvement a Distinct Phenotype of Post-COVID-19?

“…Regardless of calculated percentages, modern methods of radiological analysis such as a contrastive learning model used by Li et al [51] manage to distinguish clusters of post-COVID-19 patients-in the mentioned study, it was an air-trapping pattern (due to some small airways disease involvement) and an interstitial fibrotic-like pattern, while the latter were hospitalized more often (90 vs. 67%) and had lower FVC%, TL,CO%, TLC, and higher RV/TLC as well as more ground glass opacities during assessment after the median time of 113 days from COVID 19. These data look promising regarding its comparability to our clinical phenotype distinction; the pulmonary phenotype group presented with alterations in T L,CO %, RV%, TLC% and absolute values of FVC and FEV1.…”

Investigating distributions of inhaled aerosols in the lungs of post-COVID-19 clusters through a unified imaging and modeling approach

“…The cluster-specific characteristics, as detailed in ( Li et al, 2022 ), are outlined below (see also Table S1 ) to aid in establishing correlations with CFPD-based variables. C1 had a female-dominant composition (68.37 % females), while C0 and C2 had a more balanced distribution of males and females.…”

“…These biomarkers serve to monitor disease progression, define disease subtypes, predict health trajectories, and inform product and clinical trial designs ( Couper et al, 2014 ). For instance, a recent study, using deep learning in post-COVID-19 CT lung images, identified two distinct clusters ( Li et al, 2022 ). Cluster 1 (C1) was characterized by a female-dominant group with small airway disease, while Cluster 2 (C2) consisted of older individuals with fibrotic-like lung characteristics.…”