Effectiveness of bone suppression imaging in the diagnosis of tuberculosis from chest radiographs in Vietnam: An observer study

Abstract

ObjectiveTo assess the effectiveness of bone suppression imaging (BSI) in the diagnosis of tuberculosis from chest radiographs (CXRs) in Vietnam. Materials and methodsA total of 97 images (50 tuberculosis and 47 normal) comprised the dataset for this observer study with unanimous consensus of a panel of 3 expert radiologists. The participants were 9 Vietnamese radiologists (6 chest radiologists and 3 non-chest radiologists). Participants recorded their confidence levels regarding the presence of tuberculosis after reading a standard chest radiograph directly first and then after BSI processing. Receiver operating characteristic (ROC) analysis was used to evaluate participant performance. In addition, the change in participants' decision regarding the presence or absence of tuberculosis after BSI processing was recorded for each patient. Improvements in sensitivity and specificity were calculated. ResultsThe average AUC for non-chest radiologists improved from 0.882 without BSI to 0.933 with BSI (P = 0.048). In addition, BSI improved sensitivity by 10.0% whereas specificity decreased by 2.8% among non-chest radiologists. ConclusionUsing BSI improved the accuracy of tuberculosis diagnosis from CXRs, particularly by non-chest radiologists.