Liver functional radiomics of gadoxetic acid-enhanced magnetic resonance imaging: A proof-of-concept study

Objectives: This study aimed to identify reproducible and repeatable radiomics features (RF) of gadoxetic acid-enhanced MRI that correlate with liver function. Methods: This study included 10 patients with colorectal liver metastasis who received a right hepatectomy and were prospectively recruited. Before surgery, all patients underwent indocyanine green (ICG) retention test, hepatobiliary scintigraphy (HBS) exam and gadoxetic acid-enhanced MRI. RF were extracted from the non-tumoral liver parenchyma at the MRI images before contrast injection and the hepatobiliary phase. Delta-RF were calculated by RFs of hepatobiliary phase subtract those before contrast injection. ICG-R15, HBS results, and kinetic growth rate (KGR) of the future liver remnant volume were used as quantitative indices of liver function. Reproducible and repeatable RF were identified via inter- and intraclass correlation analysis with a coefficient threshold of 0.80. Correlations between RF/delta-RF and the three liver function indices (ICG-R15, HBS and KGR) were evaluated by Spearman correlation analysis. Results: One hundred and seven RFs were extracted from each MRI phase (214 in total and 107 derived delta RFs). With inter- and intraclass correlation analysis, 60 reproducible and repeatable RFs were identified. After removing RFs with higher correlation coefficients, 12 RF from images before contrast injection, 23 RF from hepatobiliary phase, 29 delta-RF were identified significantly correlated with the liver function metrics (ICG-R15, HBS and KGR). Conclusions: A set of reproducible and repeatable radiomics features of gadoxetic acidenhanced MRI which were associated with liver function were identified. These features can be adopted for future liver function-related radiomics research.