PurposeTo assess the performance of FLIS in predicting adverse outcomes, namely post-hepatectomy liver failure (PHLF) and death, in patients who underwent liver surgery for malignancies. MethodsAll consecutive patients who underwent liver resection and 1.5 T gadoxetic acid MR were enrolled. PHLF and overall survival (OS) were collected. Two radiologists with 18 and 8 years of experience in abdominal imaging, blinded to clinical data, evaluated all images. Radiologists evaluated liver parenchymal enhancement (EnQS), biliary contrast excretion (ExQS), and signal intensity of the portal vein relative to the liver parenchyma (PVsQs). Reliability analysis was computed with Cohen’s Kappa. Cox regression analysis was calculated to determine which factors are associated with PHLF and OS. Area Under the Receiver Operating Characteristic curve (AUROC) was computed. Results150 patients were enrolled, 58 (38.7 %) in the HCC group and 92 (61.3 %) in the non-HCC group. The reliability analysis between the two readers was almost perfect (κ = 0.998). The multivariate Cox analysis showed that only post-surgical blood transfusions and major resection were associated with adverse events [HR=8.96 (7.98–9.88), p = 0.034, and HR=0.99 (0.781–1.121), p = 0.032, respectively] in the whole population. In the HCC group, the multivariable Cox analysis showed that blood transfusions, major resection and FLIS were associated with adverse outcomes [HR=13.133 (2.988–55.142), p = 0.009, HR=0.987 (0.244–1.987), p = 0.021, and HR=1.891 (1.772–3.471), p = 0.039]. The FLIS AUROC to predict adverse outcomes was 0.660 (95 %CIs = 0.484–0.836), with 87 % sensitivity and 33.3 % specificity (81.1–94.4 and 22.1–42.1). ConclusionsFLIS can be considered a promising tool to preoperative depict patients at risk of PHLF and death.