Acoustic radiation force impulse quantification of spleen elasticity for assessing liver fibrosis

Purpose: The purpose of this study is to assess the correlation between liver fibrosis and spleen stiffness measured by ARFI in patients with chronic viral hepatitis (B or C) and to explore the possible complementary role of spleen and liver ARFI in grading liver fibrosis. Methods: 84 subjects (51 patients, 33 healthy volunteers) were enrolled. ARFI of the spleen and the liver was performed. Patients subsequently underwent liver biopsy for grading liver fibrosis according to Knodell scoring system. Multivariate logistic regression and decision tree analysis were adopted to test the relationship between spleen and liver stiffness (independent variables) and liver fibrosis (F1 < vs. a parts per thousand yenF3). Leave-One-Out Cross-Validation was used for validating the predictive classification models. Area under the ROC curve (AUROCC) was used as accuracy metric. Results: Spleen ARFI was able to discriminate early (F1) from severe (a parts per thousand yenF3) liver fibrosis with an optimal cut-off of 3.05 m/s: AUROCC 0.807, cross-validated AUROCC 0.614. Liver ARFI was superior to spleen ARFI, using a cut-off of 2.11 m/s: AUROCC 0.879, cross-validated AUROCC 0.672. Neither spleen nor liver ARFI was able to differentiate healthy volunteers from F1 patients. Odds ratios derived from logistic regression were 23.1 and 9.9 for liver and spleen ARFI, respectively; resulting AUROCC was 0.905 (cross-validated 0.848). A decision tree considering the sequential use of liver and spleen ARFI with cut-off of 2.14 and 3.39 m/s, respectively, resulted in AUROCC of 0.903 (cross-validated 0.7). Conclusions: Spleen ARFI has the potential to discriminate early from severe liver fibrosis. Spleen and liver ARFI, when combined, show a better discriminative power than liver ARFI alone.