Fig. 2.

Relationship between histological steatosis fraction and PDFF.

(A) Association between histological steatosis and PDFF in the pooled dataset of nine studies (N = 597). The best-fit line was determined using a linear mixed model, with study effects considered as random effects. Both variables underwent square root transformation before model fitting, and the curve fit was then back-transformed for display. The shaded area around the curve denotes 95% CI. The colour-shaded background of the plot illustrates division of the x-axis into histological steatosis grades S0–S3 (S0: <5%; S1: 5–33%; S2: 34–66%; S3: >66%). The horizontal dotted green lines denote optimal rule-in thresholds for PDFF to predict dichotomised steatosis grades at 90% specificity (see Table 2). The dashed black lines are drawn for illustrative purposes. (B) Bland–Altman plots showing the absolute differences and (C) ratios between histological steatosis and PDFF, as a function of histological steatosis. The best-fit lines were determined using linear regression, and variables in (C) underwent logarithmic transformation before model fitting (the curve fit was then back-transformed for display). (D) Representative liver histological sections of three individuals in the Helsinki MRS-PDFF cohort. Above each image, the corresponding pathologist-reported histological steatosis fraction and PDFF are shown. Histological sections of formalin-fixed and paraffin-embedded liver biopsies underwent Herovici staining and digitisation using Pannoramic Scan 150 (3DHISTECH Ltd.; Budapest, Hungary). The images were acquired at 10 × magnification. PDFF, proton density fat fraction.