A pilot study of ex-vivo MRI-PDFF of donor livers for assessment of steatosis and predicting early graft dysfunction

Document Type


Publication Date


Publication Title

PLoS One


BACKGROUND: The utility of ex vivo Magnetic resonance imaging proton density fat fraction (MRI-PDFF) in donor liver fat quantification is unknown.

PURPOSE: To evaluate the diagnostic accuracy and utility in predicting early allograft dysfunction (EAD) of ex vivo MRI-PDFF measurement of fat in deceased donor livers using histology as the gold standard.

METHODS: We performed Ex vivo, 1.5 Tesla MRI-PDFF on 33 human deceased donor livers before implantation, enroute to the operating room. After the exclusion of 4 images (technical errors), 29 MRI images were evaluable. Histology was evaluable in 27 of 29 patients. EAD was defined as a peak value of aminotransferase >2000 IU/mL during the first week or an INR of ≥1.6 or bilirubin ≥10 mg/dL at day 7.

RESULTS: MRI-PDFF values showed a strong positive correlation (Pearson's correlation coefficient) when histology (macro-steatosis) was included (r = 0.78, 95% confidence interval 0.57-0.89, p<0.0001). The correlation appeared much stronger when macro plus micro-steatosis were included (r = 0.87, 95% confidence interval 0.72-0.94, p<0.0001). EAD was noted in 7(25%) subjects. AUC (Area Under the Curve) for macro steatosis (histology) predicted EAD in 73% (95% CI: 48-99), micro plus macro steatosis in 76% (95% CI: 49-100). AUC for PDFF values predicted EAD in 67(35-98). Comparison of the ROC curves in a multivariate model revealed, adding MRI PDFF values to macro steatosis increased the ability of the model in predicting EAD (AUC: 79%, 95% CI: 59-99), and addition of macro plus micro steatosis based on histology predicted EAD even better (AUC: 90%: 79-100, P = 0.054).

CONCLUSION: In this pilot study, MRI-PDFF imaging showed potential utility in quantifying hepatic steatosis ex-vivo donor liver evaluation and the ability to predict EAD related to severe allograft steatosis in the recipient.

PubMed ID