ABSTRACT 419

Abstract

Background and aims: Acute respiratory distress syndrome (ARDS) is the most severe form of acute respiratory failure both in adult and children characterized by increased capillary permeability, inflammation and alveolar repair. Both Consensus Conference performed in 1994 and 2011 require the presence of bilateral pulmonary infiltrates on chest X-ray (CXR). To be consistently useful, interpretation of the CXR must be reliable. Adult studies on radiographic interpretation in ARDS have shown limited inter-observer agreement and concluded that intensivists without formal consensus training can only achieve moderate level of agreements. Aims: A computer-aided diagnosis system (CAD) was developed to automatically detect ARDS. The system is based on texture analysis of many patch of parenchyma lung among the four quadrants. The aim of the present study is to compare clinical assessment to the CAD system. Methods: CXR database was done using 111 CXR selected among 843 children included in a previous study, 60/111 CXR presented an ARDS based on gold standard reading. CXR reading was first performed by two intensivists. In case of disagreement, a consensus was performed with a third intensivist. The 111 CXR were diagnosed automatically by our CAD system. Results: Inter-observer variability was fund between the first two intensivists with a moderate kappa (kappa=0.55). The CAD system was able to significantly improve the kappa (kappa=0.75). Specificity and sensibility of the CAD system was 84.4 and 90% respectively. Conclusions: Our study confirms the inter-observer variability with clinical assessment and the use of a CAD system for CXR interpretation in ARDS is able to reduce it.