Chronic diffuse infiltrative lung disease: determination of the diagnostic value of clinical data, chest radiography, and CT and Bayesian analysis.

Abstract

To assess the value of clinical, chest radiographic, and computed tomographic (CT) findings in classifying chronic diffuse infiltrative lung disease (CDILD) MATERIALS AND METHODS: Two samples from the same population were consecutively studied: the training set (group A, n = 208) for the development of the decision aid and the test set (group B, n = 100) for validation. Computer-aided diagnoses were made with a Bayesian model that assigned to each patient diagnostic probabilities based on clinical, radiographic, or CT variables. In group A, a correct diagnosis based on clinical data was obtained in 29% of cases; radiography, 9%; and CT, 36%. This increased to 54% when clinical and radiographic variables were combined (P < .0001) and to 80% when data from all three were analyzed together (P < .0001). With prior and conditional probabilities determined from group A, the frequency of correct diagnosis in group B was 27% with clinical data, which increased to 53% (P < .0001) with radiographic findings and 61% after including CT data (P = .07). CT can help determine the specific diagnosis in patients with CDILD.