Purpose: To determine the added value of multiplanar reformation (MPR) images combined with computed tomographic (CT) images in staging of T4 gastric cancers.
Materials and Methods: The institutional review board approved this retrospective study and waived informed consent. One hundred forty-nine consecutive patients (99 men [age range, 33-85 years; mean age, 63.1 years] and 50 women [age range, 30-85 years; mean age, 57.1 years]; age range, 30-85 years; mean age, 61.1 years) with pathologically and/or surgically confirmed T3 (n = 110) or T4 tumors (n = 39), with invasion of the transverse colon or mesocolon (n = 23), pancreas (n = 16), diaphragm (n =4), liver (n = 3), and/or spleen (n = 1) were included. Two experienced radiologists independently reviewed 4-mm-thick transverse CT images and, 10 weeks later, reviewed both transverse CT and additional coronal and sagittal MPR images to assess tumor invasion in each of the aforementioned five adjacent organs. Receiver operating characteristic (ROC) analyses and weighted k statistics were obtained to evaluate reader performance and agreement in identifying tumor invasion.
Results: With addition of MPR images, the area under the ROC curve (AUC) led to a significant increase in the prediction of adjacent organ invasion in per-organ analyses (0.88 vs 0.95 for reader 1 [P = .01], 0.88 vs 0.98 for reader 2 [P < .001]) and identification of T4 tumors in per-patient analyses (0.79 vs 0.91 for reader 1 [P = .006], 0.78 vs 0.96 for reader 2 [P < .001]). In the five analyzed organs, AUC was significantly increased in regard to invasion of the transverse colon or mesocolon and pancreas; separate analysis was not performed for the diaphragm, liver, and spleen because of the small number of true invasions, lnterob-server agreement increased in per-organ ( = 0.43 vs 0.67) and per-patient ( = 0.42 vs 0.62) analyses.
Conclusions: Adding MPR images to transverse CT images improves the capability for distinguishing T3 from T4 gastric cancer and prediction of adjacent organ invasion.