Development of a high-precision image-processing automatic measurement system for MRI visceral fat images acquired using a binomial RF-excitation pulse

Abstract

Development of a rapid and accurate method for visceral fat measurement is an important task, given the recent increase in the number of patients with metabolic syndrome. In this study, we optimized the Fast Low Angle Shot (FLASH) sequence using a binominal radiofrequency excitation pulse, in which the acquisition time is short, and measured changes in the amount of visceral fat in subjects after a period of wearing clothes with a fat-reducing effect during walking. We solved the reproducibility problem associated with the number of slices, and developed automatic measurement software for high-precision separation and extraction of abdominal visceral fat images. This software was developed using intensity correction with the coil position, derivation of a threshold by histogram analysis and fat separation by template matching for abdominal images. The cross-sectional area of a single slice varies for every acquisition due to visceral organ movement, but the relative error largely converged for seven slices. The measured amount of abdominal fat tended to be consistent with changes in the body fat and waist circumference of the subjects. The correlation coefficients between automatic extraction using the measurement software and manual extraction were 0.9978 for subcutaneous fat and 0.9972 for visceral fat, showing very strong positive correlations. The consistency rates were 0.9502±0.0167 for subcutaneous fat and 0.9395±0.0147 for visceral fat, and the shapes of the regions were also extracted very accurately. These results show that the magnetic resonance imaging acquisition method and image processing system developed in this study are beneficial for measurement of abdominal visceral fat. Therefore, this method may have a major role in future diagnosis of metabolic syndrome.

Introduction

The increased number of cases of metabolic syndrome is a cause of concern [1], [2], [3]. Metabolic syndrome is a condition in which insulin action is reduced due to visceral fat accumulation and risk factors for arteriosclerosis, including abnormal glucose metabolism (abnormal glucose tolerance, diabetes), abnormal lipid metabolism (hypertriglyceridemia, hypo-high-density lipoprotein cholesterolemia), and hypertension. Each factor independently increases the risk, and the incidence of arteriosclerosis is increased synergistically when multiple factors are present concomitantly. Visceral fat accumulation is an essential requirement for diagnosis of metabolic syndrome, along with the presence of two conditions among hyperglycemia, hypertension and hyperlipidemia [4], [5], [6], [7].

Since visceral fat accumulation is present in metabolic syndrome, accurate measurement of the level of visceral fat is important as a basis for diagnosis of the disorder. In current measurements (or estimations) of the amount of visceral fat, waist circumference is measured as a simple method [8], [9], and X-ray computed tomography (CT) and magnetic resonance imaging (MRI) are used for precise examination [10], [11], [12]. In X-ray CT, only a single cross-sectional view at the navel level is acquired to reduce X-ray exposure, but accurate, reproducible fat measurement is difficult using imaging of only a single slice because the visceral organs move while the patient is lying on a bed. MRI for abdominal imaging is advantageous since there is no X-ray exposure, but magnetic field homogeneity is likely to be disturbed by respiratory movement and air in the intestine [13], [14], [15]. Fat-selective MRI is commonly performed using the FLAIR (fluid-attenuated inversion recovery), Dixon, Chess (chemical shift selective saturation), and Binominal methods [16]. The scanning time is prolonged in FLAIR due to use of an inversion recovery (IR) pulse and in the Dixon method due to the requirement for fusion of multiple images. The Dixon method uses echo times or echo time shifts such that the phase difference of water and fat is at either 0 or a multiple of π at the echo times. The water suppression effect of the Dixon method is decreased in the abdominal region where a change of magnetic susceptibility is large and a magnetic field is inhomogeneous. In the Chess method, fat suppression decreases in regions in which the magnetic field is relatively heterogeneous or the magnetic susceptibility changes markedly, which is a major problem in abdominal imaging.

In this study, we optimized a sequence using a 1-2-1 binomial radiofrequency (RF) excitation pulse, in which the acquisition time is short and motion artifact is less influential, and measured changes in the amount of visceral fat after wearing of clothes with a fat-reducing effect during walking (M2). We acquired images of several cross-sections over the abdomen, rather than a single slice, and investigated the reproducibility of measurements by calculating the fat volume, rather than the area, from several slices (M1). We also developed software for high-speed high-precision visceral fat image-processing automatic measurement, thereby reducing the influence of coil sensitivity heterogeneity and increasing its applicability in clinical practice (M3).