RESUMO
OBJECTIVE: The purpose of our study was to evaluate chest radiographic features of nontraumatic mediastinal hemorrhage occurring after extrapericardial thoracic aorta rupture. MATERIALS AND METHODS: Twenty-seven consecutive chest radiographs obtained at admission of patients with hemorrhage from ruptured thoracic aorta aneurysms, aortic dissections, or penetrating aortic ulcers were randomized with radiographs of 23 subjects with nonruptured thoracic aorta aneurysms, 20 subjects with nonruptured dissections, and 20 control subjects. Diagnoses were established by interpreting CT scans, MR images, and findings at surgery or autopsy or both. A retrospective review was performed by three independent radiologists who were unaware of patients' diagnoses. Observers assessed 20 parameters on each of these 90 radiographs and summarized their findings with final diagnoses. The 20 parameters were analyzed with logistic regression and rank correlation to determine the best predictors of hemorrhage. RESULTS: Logistic regression analysis showed a combination of obscuration or convexity of the aorticopulmonary window and a displaced left paraspinal interface to be the most useful predictor of hemorrhage (p < .05). Rank correlation analysis indicated obscuration or convexity of the aorticopulmonary window; a displaced left paraspinal interface; enlarged aortic knob width; enlarged thoracic aorta size; an enlarged, obscured, or irregular aortic margin; and left pleural or extrapleural space fluid were potential individual predictors of hemorrhage (p < .05). Observer sensitivities for recognizing hemorrhage were 30-59% and specificities were 83-91%. Sensitivities for distinguishing an abnormal (n = 70) from a normal (n = 20) mediastinum were 79-90% and specificities were 65-90%. CONCLUSION: Obscuration or convexity of the aorticopulmonary window and a displaced left paraspinal interface on radiographs may indicate mediastinal hemorrhage. Further imaging is required to establish a definitive diagnosis.
