Fully automated measurement of total adipose tissue volume using quantitative chemical shift MRI: Phantom Validation

INTRODUCTION. Accurate identification and quantification of total adipose tissue (TAT) volume is a key first step for segmentation and measurement of visceral adipose tissue (VAT) and subcutaneous adipose tissue (SCAT), which are critical metrics in diagnosis and treatment of obesity-related diabetes, cardiovascular disease, and metabolic syndrome (1-3). Anthropometric measurements of waist circumference, waist-hip ratio, and body mass index (BMI) are widely used clinically to indirectly characterize TAT, VAT and SCAT, but are highly prone to systematic error (4-5) and correlate poorly with actual adipose tissue volumes (6-7). Qualitative segmentation using empirical signal thresholds and manual segmentation of adipose tissue on T1-weighted MRI is considered the reference standard for direct VAT measurement, but is prohibitively timeconsuming for clinical use. Qualitative manual segmentation is also subject to partial volume effects at fat-water and fat-void interfaces, potentially leading to significant errors and poor repeatability in TAT/VAT/SCAT estimation. Chemical shift-based fat/water MRI methods are more accurate than T1-weighted MRI for visualizing adipose tissue (8-9) and potentially permit more rapid adipose tissue segmentation (10-11) by applying a simple fat-fraction threshold. However, the quantitative accuracy of chemical shift methods is confounded by relaxation effects (12-14) and spectral complexity of fat (13, 15), resulting in significant errors in fat-fraction values (16-17). Also, to avoid partial volume effects at signal boundaries, the fat-fraction threshold for adipose tissue is typically defined as 50%, implicitly assuming a maximum fat fraction (ηMAX) of 100%, but in vivo adipose tissue also contains organelles, blood vessels, and water components which result in a true ηMAX < 100%. Therefore, ηMAX/2 is a more physiologically meaningful choice for adipose tissue thresholding, which can be directly measured from quantitative fat-fraction maps. The purpose of this work is to describe a quantitative chemical shift-based fat/water MRI method for fully automated estimation of ηMAX and volume of TAT. To assess the robustness of the TAT volume measurement with respect to partial volume effects, we employ a series of oil phantoms with varying volume and surface area complexity, using agar gel, glass rods, and empty plastic vials.