Aims: To evaluate the accuracy of different techniques of MRI steatosis quantification, based on histological grading and quantification of liver steatosis.
Patients and methods: Twenty-three patients (21 with nonalcoholic fatty liver disease and two controls) were included. Steatosis was evaluated in liver specimens using histological grading (five grades) and steatosis area (% of liver surface) was computed using an inhouse automated image analysis. The following five MRI quantification techniques were performed: two-point Dixon, three-point Dixon, DUAL, spin echo method and a new technique called multi-echo gradient-echo (MFGRE). Interobserver (two observers) and intersite (three different liver sites) agreements were evaluated for the two best-performing methods.
Results: Steatosis area was correlated with steatosis grade: Rs (Spearman coefficient) = 0.82, P value of less than 0.001. The steatosis area was significantly different between S0-S2 and S3-S4 grades: 4.2 + or - 2.4 versus 16.4 + or - 8.9% (P< 0.001). Correlations between the MRI techniques and steatosis area (or grading) were: MFGRE, Rs = 0.72 (0.78); spin echo method, Rs = 0.72 (0.76); DUAL, Rs =0.71 (0.76); two-point Dixon, Rs = 0.71 (0.75); three-point Dixon, Rs = 0.67 (0.77). Interobserver (Ric = 0.99) and intersite (Ric = 0.97) agreements were excellent for the liver steatosis measurement by MFGRE. The noninvasive diagnosis of the steatosis area was improved by adding blood markers like ALT and triglycerides to MFGRE (aR2: 0.805).
Conclusion: MRI, and in particular the MFGRE method, provides accurate and automatic quantification for the noninvasive evaluation of liver steatosis, either as a single measurement or in combination with blood variables.