Measurement of spleen fat on MRI-proton density fat fraction arises from reconstruction of noise

• DOI: 10.1007/s00261-019-02079-z
• Publication Date: 2019-06-06T09:02:36Z
• Authors: Cheng William Hong; Gavin Hamilton; Catherine Hooker; Charlie C. Park; Calvin Andrew Tran; Walter C. Henderson; Jonathan C. Hooker; Soudabeh Fazeli D...; Jeffrey B. Schwimmer; Scott B. Reeder; Claude B. Sirlin
• Keywords: Artifactual; Adult; Male; Magnetic Resonance Spectroscopy; Adolescent; Signal-To-Noise Ratio; Imaging; Fat quantification; 03 medical and health sciences; Computer-Assisted; Imaging, Three-Dimensional; 0302 clinical medicine; Clinical Research; Image Interpretation, Computer-Assisted; Humans; Prospective Studies; Child; Image Interpretation; Spectroscopy; screening and diagnosis; Middle Aged; Magnetic Resonance Imaging; 4.1 Discovery and preclinical testing of markers and technologies; 3. Good health; Detection; Adipose Tissue; CSE-MRI; Three-Dimensional; Biomedical Imaging; Female; Algorithms; Spleen

This study compares splenic proton density fat fraction (PDFF) measured using confounder-corrected chemical shift-encoded (CSE)-MRI to magnetic resonance spectroscopy (MRS) in human patients at 3T.This was a prospectively designed ancillary study to various previously described single-center studies performed in adults and children with known or suspected nonalcoholic fatty liver disease. Patients underwent magnitude-based MRI (MRI-M), complex-based MRI (MRI-C), high signal-to-noise variants (Hi-SNR MRI-M and Hi-SNR MRI-C), and MRS at 3T for spleen PDFF estimation. PDFF from CSE-MRI methods were compared to MRS-PDFF using Wilcoxon signed-rank tests. Demographics were summarized descriptively. Spearman's rank correlations were computed pairwise between CSE-MRI methods. Individual patient measurements were plotted for qualitative assessment. A significance level of 0.05 was used.Forty-seven patients (20 female, 27 male) including 12 adults (median 55 years old) and 35 children (median 12 years old). Median PDFF estimated by MRS, MRI-M, Hi-SNR MRI-M, MRI-C, and Hi-SNR MRI-C was 1.0, 2.3, 1.9, 2.2, and 2.0%. The four CSE-MRI methods estimated statistically significant higher spleen PDFF values compared to MRS (p < 0.0001 for all). Pairwise associations in spleen PDFF values measured by different CSE-MRI methods were weak, with the highest Spearman's rank correlations being 0.295 between MRI-M and Hi-SNR MRI-M; none were significant after correction for multiple comparisons. No qualitative relationship was observed between PDFF measurements among the various methods.Overestimation of PDFF by CSE-MRI compared to MRS and poor agreement between related CSE-MRI methods suggest that non-zero PDFF values in human spleen are artifactual.