Giuliano Scattolin MD, Presenter: Nothing to Disclose

Dow-Mu Koh MD, FRCR, Abstract Co-Author: Nothing to Disclose

Giovanni Morana MD, Abstract Co-Author: Nothing to Disclose

David Watkins MBBS, MRCP, Abstract Co-Author: Nothing to Disclose

Angela Mary Riddell MBBS, Abstract Co-Author: Nothing to Disclose

Peritoneal disease contraindicates surgery in patients with gastroesophageal, gastric and pancreatic cancers, but the diagnostic performance of CT in this context is unknown. We compare the CT diagnostic performance between experienced and inexperienced radiologists for peritoneal disease detection in primary upper abdominal malignancies.

In this IRB approved study, the pre-operative CT images of 110 patients with gastroesophageal (n=41), gastric (n=49) and pancreatic (n=10) cancers who underwent laparoscopy were retrospectively reviewed. Portovenous phase CT was performed within 3 weeks prior to laparoscopy using multichannel CT and 1-3 mm multiplanar image reconstruction. Images were reviewed by an expert (>10 years experience) and inexperienced radiologist for presence, size and location of peritoneal nodules; peritoneal stranding, peritoneal thickening, cystic disease and ascites. CT images were scored using set criteria on a 4-point scale for likelihood of peritoneal disease. Reader diagnostic performance was compared by ROC analysis and observer agreement by kappa statistics. CT features were compared between patients with and without peritoneal disease using Fisher’s exact test.

97 were laparoscopically negative for peritoneal disease and 13 positive. For experienced radiologist, CT score >2 showed 77% (95%CI: 46-95%) sensitivity and 58% (47-68%) specificity for detecting peritoneal disease. For inexperience radiologist, there was 71% (42-92%) sensitivity and 52% (42-62%) specificity. By ROC analysis, there was no difference in diagnostic performance between readers (Az 0.69 vs 0.61, p=0.23) with moderate interobserver agreement (kappa = 0.47). For experienced reader, peritoneal nodules > 5 mm in gastrohepatic ligament were more frequently observed in patients with peritoneal disease (Fisher’s exact test p=0.005), but other CT features were non-discriminatory in both readers.

In patients with primary upper abdominal malignancies, reader experience shows little advantage for detecting peritoneal disease. The diagnostic performance of CT is modest in both experienced and inexperienced readers.

Peritoneal disease detection in upper abdominal cancers is limited using CT and appears uninfluenced by reader experience. Other imaging techniques (e.g. DWI) should be investigated to improve diagnostic performance.

Scattolin, G, Koh, D, Morana, G, Watkins, D, Riddell, A, Detection of Peritoneal Disease Using Multidetector CT in Upper Abdominal Malignancies: Diagnostic Performance in Experienced and Inexperienced Readers.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14017506.html