PURPOSE: To evaluate the performance of radiologists in the diagnosis of internal hernia with specific computed tomographic (CT) findings. MATERIALS AND METHODS: Abdominal CT scans obtained in 42 patients were retrospectively reviewed by three radiologists. The case group consisted of 18 patients with internal hernias (two paraduodenal, 16 transmesenteric); the comparison group was 24 patients with no internal hernia. Images were reviewed in a random and blinded fashion. Individual and group performance was evaluated with receiver operating characteristic (ROC) analysis, and interobserver agreement was measured with Cronbach coefficient alpha. Individual CT signs relevant as predictors of transmesenteric hernia were identified with logistic regression analysis and ranked by their odds ratio and P values. RESULTS: Both paraduodenal hernias were diagnosed by all readers on the basis of CT signs, including a retrogastric saclike mass of small-bowel loops. Diagnosis of transmesenteric hernia was more difficult and variable, with an average accuracy of area under the ROC curve (A(z)) of 77%, sensitivity of 63%, and specificity of 76%. CT signs of transmesenteric hernia were recognized consistently (Cronbach coefficient alpha >or= 0.80) and included a cluster of dilated small-bowel segments and stretching and displacement of mesenteric vessels. Coexisting volvulus and ischemia were diagnosed with low sensitivity (46% and 43%, respectively) but high specificity (96% and 98%, respectively). CONCLUSION: Diagnosis of internal hernia with CT remains difficult. Special attention should be given to the clustering of bowel loops, the mesenteric vessels, and signs of small-bowel obstruction.
PURPOSE: To evaluate the performance of radiologists in the diagnosis of internal hernia with specific computed tomographic (CT) findings. MATERIALS AND METHODS: Abdominal CT scans obtained in 42 patients were retrospectively reviewed by three radiologists. The case group consisted of 18 patients with internal hernias (two paraduodenal, 16 transmesenteric); the comparison group was 24 patients with no internal hernia. Images were reviewed in a random and blinded fashion. Individual and group performance was evaluated with receiver operating characteristic (ROC) analysis, and interobserver agreement was measured with Cronbach coefficient alpha. Individual CT signs relevant as predictors of transmesenteric hernia were identified with logistic regression analysis and ranked by their odds ratio and P values. RESULTS: Both paraduodenal hernias were diagnosed by all readers on the basis of CT signs, including a retrogastric saclike mass of small-bowel loops. Diagnosis of transmesenteric hernia was more difficult and variable, with an average accuracy of area under the ROC curve (A(z)) of 77%, sensitivity of 63%, and specificity of 76%. CT signs of transmesenteric hernia were recognized consistently (Cronbach coefficient alpha >or= 0.80) and included a cluster of dilated small-bowel segments and stretching and displacement of mesenteric vessels. Coexisting volvulus and ischemia were diagnosed with low sensitivity (46% and 43%, respectively) but high specificity (96% and 98%, respectively). CONCLUSION: Diagnosis of internal hernia with CT remains difficult. Special attention should be given to the clustering of bowel loops, the mesenteric vessels, and signs of small-bowel obstruction.