Proposal of objective morphological classification system for hepatocellular carcinoma using preoperative multiphase computed tomography

Background To establish a preoperative morphological classification system for hepatocellular carcinoma using multiphase computed tomography. Methods All consecutive patients who were diagnosed with hepatocellular carcinoma between 2004 and 2009 were enrolled, for a total of 232 patients. The concavity and convexity of each outer contour of hepatocellular carcinoma acquired from multiphase computed tomography were analyzed, and the area and depth of each indentation were quantified. The indentation area to tumor area ratio (s:S) and the s: S ratio multiplied by the indentation depth to indentation base ratio (s: S x d:t) were used as feature values reflecting the individual shapes. Results Using a hierarchical cluster analysis, the shapes were classified into three groups: Type I (smooth: n = 158), Type II (jagged: n = 63), and Type III (rough: n = 11). The 5-year survival rates for Types I, II, and III were 64, 53, and 0 %, respectively (I vs. II, P = 0.038; I vs. III, P = 0.001; II vs. III, P = 0.002). The 5-year disease-free survival rates for Types I, II, and III were 27, 23, and 0 %, respectively (I vs. III, P = 0.0003 and II vs. III, P = 0.008). Microscopic portal venous invasion was significantly more likely with Type III than with Type I or II (P<0.001 and P = 0.001, respectively). Conclusions The newly developed semiautomatic computed tomography-based morphological classification system appears to provide a promising additional criterion for the prognostic categorization of patients with hepatocellular carcinoma.