Abstract
Although paraquat (PQ) concentrations are the most reliable prognosis predictors of PQ poisoning, these laboratory tests are not readily available in all hospitals. In this study, we proposed an imaging related parameter, that is, the ratio of injured lung volume fraction, for the prognosis evaluation of acute PQ poisoning based on the correlation between disease progress and lung imaging features. An artificial neural network was trained and then used to classify the injured and normal lung regions. The ratio of injured lung volume fraction was calculated from the injured lung volume fractions in the first and second CT scans after three-dimensional reconstruction. Parameters of blood tests were collected. A significant difference was observed with respect to the ratio of injured lung volume fraction between survivors and nonsurvivors (0.73 ± 0.17 versus 0.40 ± 0.14, P < 0.001). No patients survived when the ratio of injured lung volume fraction was less than 0.3, while all patients survived as the ratio of injured lung volume fraction was greater than 0.8. Thus, the ratio of injured lung volume fraction may provide an alternative and informative measure for the prognosis of acute PQ poisoning.