New exponential functions based on CT density to estimate the percentage of liver that is fat.

OBJECTIVE: In fluorine-18 fludeoxyglucose positron emission tomography/CT, hepatic standardized uptake value (SUV) is reduced through "signal dilution" by hepatic fat. The maximum SUV (SUVmax ) is less affected than the mean SUV (SUVmean ), therefore SUVmax /SUVmean correlates with hepatic fat. The SUV can be corrected for signal dilution using an equation relating CT density (CTD) to %fat. The objective was to exploit the relationship between SUV indices and CTD to assess the validity of two previously published equations (one linear and one sigmoid) for estimating %fat from CTD and two new exponential equations.

METHODS: Study population comprised 465 patients having routine fluorine-18 fludeoxyglucose positron emission tomography/CT. The SUVmax , SUVmean and CTD were measured from a 3-cm-diameter region of interest over the liver. The exponential equations assumed that 100% fat corresponds to CTD of -50 or -100 HU. The proportion of liver occupied by fat (PF ) was estimated from all four equations. Then fat-corrected SUVmean is SUVmean /(1 - PF ). The ideal equation should give SUVmean approaching but not exceeding SUVmax and give fat-corrected SUVmean /SUVmax that shows no correlation with CTD.

RESULTS: The linear equation failed at CTD values exceeding 55.8 HU because it gave negative PF values. Moreover, fat-corrected SUVmean /SUVmax still correlated with CTD. The sigmoid equation grossly overcorrected SUVmean at low CTD. The exponential equations abolished the correlation between fat-corrected SUVmean /SUVmax and CTD.

CONCLUSION: The sigmoid equation is unsuitable for estimating %fat from CTD. The linear equation performed well, but the exponential equation assuming that 100% fat corresponds to -50 HU performed best. Advances in knowledge: Improved (exponential) equations to estimate hepatic fat from hepatic CTD.