Chronic lung allograft dysfunction phenotype and prognosis by machine learning CT analysis

Abstract

BackgroundChronic lung allograft dysfunction (CLAD) is the principal cause of graft failure in lung transplant recipients and prognosis depends on CLAD phenotype. We used machine learning computed tomography (CT) lung texture analysis tool at CLAD diagnosis for phenotyping and prognostication compared to radiologists’ scoring.MethodsThis retrospective study included all adult first double-lung transplant patients (01/2010–12/2015) with CLAD (censored 12/2019) and inspiratory CT near CLAD diagnosis. The machine learning tool quantified ground-glass opacity, reticulation, hyperlucent lung, and pulmonary vessel volume (PVV). Two radiologists scored for ground-glass opacity, reticulation, consolidation, pleural effusion, air trapping and bronchiectasis. Receiver operating characteristic curve analysis was used to evaluate the diagnostic performance of machine learning and radiologist for CLAD phenotype. Multivariable Cox proportional-hazards regression analysis for allograft survival controlled for age, sex, native lung disease, cytomegalovirus serostatus, and CLAD phenotype (bronchiolitis obliterans syndrome [BOS] and restrictive allograft syndrome [RAS]/mixed).Results88 patients were included (57 BOS, 20 RAS/mixed, and 11 unclassified/undefined) with CT a median 9.5 days from CLAD onset. Radiologist and machine learning parameters phenotyped RAS/mixed with PVV as the strongest indicator (AUC 0.85). Machine learning hyperlucent lung phenotyped BOS using only inspiratory CT (AUC=0.76). Radiologist and machine learning parameters predicted graft failure in the multivariable analysis, best with PVV (HR=1.23, 95%CI 1.05–1.44, p=0.01).ConclusionsMachine learning discriminated between CLAD phenotypes on CT. Both radiologist and machine learning scoring were associated with graft failure, independent of CLAD phenotype. PVV, unique to machine learning, was the strongest in phenotyping and prognostication.

Percentage of low attenuation area on computed tomography detects chronic lung allograft dysfunction, especially bronchiolitis obliterans syndrome, after bilateral lung transplantation

“…Compared with the previously reported CT‐based diagnostic approaches for CLAD, the %LAA was not included in the CT‐scan score, which was evaluated by blinded radiologists for the evidence of consolidation, bronchiectasis, reticular change, pleural effusion, and ground‐glass opacities 9 . Although a machine‐learning CT analysis included hyperlucent lung, which corresponds to the %LAA in the present study, as well as ground‐glass opacity, reticulation, and pulmonary vessel volume, 24 the %LAA can be obtained objectively and easily using diagnostic imaging software without requiring machine learning. Unlike a quantitative lung density analysis, 11,25 the %LAA was calculated based on lung field areas with attenuation values less than the designated threshold only, and not a lung histogram analysis.…”

“…CLAD can present either as a predominantly obstructive ventilatory pattern, a restrictive pattern, or a mixed obstructive and restrictive pattern that is not explained by other conditions or as a combination of these 7 . Recently, advanced diagnostic imaging methods including lung perfusion scintigraphy, inspiratory and expiratory computed tomography (CT) volumetry, the CT‐scan score, quantitative CT analysis, and machine learning CT analysis have opened more doors to diagnosing CLAD 8–12 …”

“…7 Recently, advanced diagnostic imaging methods including lung perfusion scintigraphy, inspiratory and expiratory computed tomography (CT) volumetry, the CT-scan score, quantitative CT analysis, and machine learning CT analysis have opened more doors to diagnosing CLAD. [8][9][10][11][12] In the CT analysis of lung disease, the percentage of low attenuation area (%LAA) is a useful parameter for evaluating the severity of lung emphysematous changes on CT in patients with chronic obstructive pulmonary disease (COPD). 13 The %LAA is automatically calculated using diagnostic imaging software, enabling an objective evaluation of the severity of COPD.…”

Biomarkers for Chronic Lung Allograft Dysfunction: Ready for Prime Time?

“…13 A comprehensive study compared scoring by radiologists with machine learning prediction‚ and although both radiologist and machine learning scoring were associated with graft failure, the pulmonary vasculature volume, unique to machine learning, was the strongest in phenotyping and prognostication of patients. 14 Lastly, the parenchyma can be comprehensively investigated where a subdivision can be made in those tools that differentiate signs of hyperinflation and air trapping versus those focusing on parenchymal changes related to fibrosis. Simple measures of lung volume and lung density have been used to differentiate the BOS and RAS phenotype with the latter being characterized with a lower lung volume and an increased lung density.…”

Prognostic Value of Chest CT Findings at BOS Diagnosis in Lung Transplant Recipients

“…2 In recent years, reader-independent CT quantification methods have been studied, mainly aiming to better discriminate between different CLAD phenotypes. [4][5][6] However, particular CT features may also provide additional prognostic information in established CLAD. In RAS, which is radiologically characterized by persistent subpleural and/or diffuse interstitial fibrosis, worse outcomes have been attributed to specific CT findings, such as the presence of diffuse or basal versus apical parenchymal infiltrates, increased lung density, and decreased graft volume.…”