Feasibility and accuracy of quantification of fat content using iterative decomposition of water and fat with asymmetry and least squares estimation-quantitative fat imaging：a phantom study / 中华放射学杂志

ABSTRACT

Objective To validate the feasibility and accuracy of iterative decomposition of water and fat with asymmetry and least squares estimation-quantitative fat imaging (IDEAL-IQ) in fat quantification using fat-water model. Methods A homogeneous fat-water mixture model consisting of various known fat-fractions were described, and the fat fraction was 0.00, 0.01, 0.02, 0.04, 0.06, 0.08, 0.10, 0.14, 0.18, 0.22, 0.26, 0.30 g/ml respectively. A water-vaseline separated model was also described. IDEAL-IQ was performed. Thin slices were acquired for fat-water mixture model and repeated 3 days later. Nineteen slices of 14 mm-thick parallel to the water-vaseline boundary in 1 mm steps from vaseline to water were acquired. The fat-fractions in 11 slices of fat-water mixture model were measured on FatFrac images. Accuracy was assessed through single sample t test or Kolmogorov-Sirmov test. Measured fat-fractions of the same known fat-fraction were assessed through independent samples t test between two scan times. Linear regression was used to assess the relationship between known fat-fractions and measured fat-fractions. Slices containing the water-vaseline boundary were measured with ROI in the middle of the FatFrac images. The relationship between measured fat-fractions and locations of scanning was exploded using curve fitting. Results (1) Fat-water mixture model no significant difference(P>0.05) was found between measured fat-fractions and known fat-fractions when it was 0.00, 0.02, 0.06 and 0.08 g/ml with the measured fat-fractions 0.60%, (2.30 ± 0.60)%, (5.76 ± 1.40)%, (7.62 ± 1.40)% respectively for the first time. No significant difference(P>0.05) was found between measured fat-fractions and known fat-fractions when it was 0.00, 0.02, 0.10 g/ml with the measured fat-fractions 0.04%, (2.32 ± 0.60)%, (9.41 ± 1.00)%respectively for the second time. Measured fat-fraction was inlinear relation with known fat-fractionY=0.898X+0.224, r2=0.993, P<0.01, F=36 129.548.(2) Water-vaseline separated model measured fat-fraction increased as scanning location changed, Y=0.045X2-0.499X-4.474, r2=0.978, P<0.05, F=350.623.Conclusions IDEAL-IQ can be used to quantify fat content with good repeatability and can accurately assess the actual fat content from the linearrelationship.