Abstract: Objective To explore a new method for assessing fat content in nonalcoholic fatty liver disease(NAFLD) by combining imaging methods and serological indicators to construct a multiple linear regression prediction model. Methods A total of 220 patients were enrolled from January 2015 to December 2020 in the Radiology Department of The Hospital of Traditional Chinese Medicine affiliated with Shanghai University of Traditional Chinese Medicine, including 86 patients in the non-NAFLD group and 134 patients in the NAFLD group. Furthermore, the NAFLD group was subdivided into mild group, moderate group and severe group, according to the degree of fatty liver. All patients underwent CT and MRI scans at the same time and obtained serological examination including alanine aminotransferase (ALT), aspartate aminotransferase (AST), γ-glutamine transpeptidase (γ-GT), total cholesterol (TC), triglyceride (TG), high-density lipoprotein (HDL-C), and low-density lipoprotein (LDL-C). The correlation of serology, MRI in-phase and opposed-phase and diffusion-weighted imaging with different degrees of fatty liver was analyzed, and the prediction model of multiple linear regression was established. Results AST levels of the mild, moderate and severe groups were 32.78±23.81 U/L, 37.53±24.87 U/L and 46.10±28.62 U/L, respectively. ALT levels were 37.79±33.64 U/L, 43.60±22.60 U/L and 70.11±57.23 U/L, respectively. TG levels were (2.18±1.09), (2.47±1.75) and (3.00±2.60) mmol/L, respectively. GGT levels were 61.81±86.91 U/L, 108.22±157.38 U/Land 78.80±73.39 U/L, respectively. These serological indicators were significantly higher than the control group(P<0.05). HDL-C of the mild, moderate and severe groups were 1.19±0.35 U/L, 1.10±0.28 mmol/L and 1.11±0.31 mmol/L, respectively, significantly lower than that of the control group (1.32±0.30 mmol/L)(P<0.05). In the diagnosis of the mild group, the optimal cut-off value of HFF was 0.04, AUROC was 0.911 (95%CI 0.873-0.949), sensitivity was 88.1%, specificity was 81.4%; In the diagnosis of the moderate group, the optimal cut-off value was 0.11, AUROC was 0.880 (95%CI 0.815-0.945), sensitivity was 87.2%, specificity was 77.5%. In the diagnosis of the severe group, the optimal cut-off value was 0.18, AUROC was 0.978 (95%CI 0.960-0.996), sensitivity was 100%, and specificity was 88.9%. The final multiple linear regression equation was: liver/spleen density ratio =1.202-0.002AST-2.215HFF (R²=0.690). Conclusion In the initial diagnosis of fat content in patients with NAFLD, AST and HFF have the potential to substitute the ratio of liver/spleen CT value in predicting the fat content of NAFLD.

Key words: Nonalcoholic fatty liver disease, Serology, Imaging, Noninvasive diagnosis, Multiple linear regression