B.Leporq
1, H.Ratiney
1, F. Pilleul
1,2,O.Beuf
1--
Article original soumis dans le journal
European Radiology
1Université de Lyon; CREATIS; CNRS UMR 5220; Inserm U1044; INSA-Lyon; UCBL; Villeurbanne, France
ABSTRACT
Aim was to validate a magnitude-based method, transposable on any imaging system and for all current clinical fields (1.5 T and 3.0 T) for Fat Volume Fraction (FVF) quantification in the liver without dominant component ambiguity problem.
MR imaging was performed at 1.5 and 3.0T using a multiple-angle multiple-gradient echo acquisition. A quantification algorithm correcting for relaxation time effects using a disjointed estimation of T1 and T2* of fat and water and accounting for the NMR spectrum of fat was developed. Validations were performed at 1.5 and 3.0T on fat-water emulsion and compared with 1H-MRS, then, in-vivo, in a prospective comparative study with histology.
Phantom study showed a good agreement between MRI method and MRS. MR-estimated FVF and histological results were strongly correlated and FVF allowed the diagnosis of a mild (cut-off = 5.5 %) and a moderate steatosis (cut-off = 15.2 %) with a sensitivity/specificity of 100%.
FVF could be a relevant biomarker for the clinical follow-up of patient with or at risk of NAFLD and to follow steatosis in patients with others chronic liver diseases. Because of disjointed fat and water relaxation times estimates, this method could potentially detect iron with or without fat as well as cirrhosis.
Chapitre.3: Developpements réalisés Quantification de la stéatose
198 INTRODUCTION
In the past decade, the incidence rise of obesity, particularly of childhood obesity [1,2], diabetes and lipid metabolism disorders involved an epidemic increase of Non-Alcoholic Fatty Liver Diseases (NAFLD) prevalence which was estimated at 33.6 % in the USA [3] and between 20 and 30 % in Europe [4,5]. In Western countries, NAFLD is the most common cause of chronic liver disease [6]. In about 20 % of cases, NAFLD progress to its aggressive form known as Non-Alcoholic SteatoHepatitis (NASH) and characterized by inflammation and fibrosis in addition to steatosis. Many of these patients may progress to cirrhosis, the end-stage of liver fibrosis, and complications such as hepatic decompensation, portal hypertension and hepatocellular carcinoma (HCC) are linked to growing public health concerns. A study has shown that the risk of HCC was high in patients with NAFLD and that 7% of the patients with NAFLD-related cirrhosis developed HCC over a 10 year period [7]. HCC has been also reported in patients with NAFLD without fibrosis or cirrhosis, suggesting the direct carcinogenic effect of hepatic fat [8]. Moreover, it has been shown that steatosis may reduce the efficiency of antiviral therapy and accelerate disease progression in patient with C hepatitis [9]. While histology after liver biopsy is the gold standard for liver steatosis assessment, inherent risk with a recognized morbidity and mortality renders this method unsuitable and problematic for longitudinal clinical monitoring on children [10,11]. Furthermore, liver biopsies have other limitations such as cost, inter-observer variability and sampling errors [12,13]. For these reasons, non-invasive and cost effective quantification methods are needed for an accurate quantification of liver fat content.
Magnetic Resonance Imaging (MRI) is a promising imaging modality for the assessment of fat content, thanks to its sensitivity to the presence of fat. This arises due to difference in chemical shift between fat and water proton. Proton (1H) Magnetic Resonance Spectroscopy (MRS) is considered as the gold standard amongst all imaging modalities [14] and its specificity/sensitivity has been demonstrated to be higher than other imaging modalities such as X-ray CT and ultrasound [15]. Nevertheless, MRS may be prone to sampling error because quantification is typically realized on a single voxel. A variety of non-invasive quantification methods using MRI have been proposed for liver fat content quantification. The two-point Dixon method was first proposed [16-21]. These methods were limited by the inability to correct the confounding factors including the T2* decay, the T1 related-bias and the complex NMR spectrum of fat which may impair fat content quantification [22]. Other methods such as methods based on low flip angle three-point Dixon acquisitions [23-24] have been developed to both take into account the T2*-decay between two states (In Phase (IP) and out of phase (OP)) and to reduce T1 saturation effect. These methods are correcting for T2*-decay by using a single T2* estimation including both water and fat, that may induce error when the difference of T2* between fat and water is significant [24]. Multiple gradient echo acquisition with dual T2* calculation was reported but this method does not correct for T1-related bias and multiple NMR components of fat [25]. Then, low-flip angle multipoint method using IDEAL fat-water separation algorithm was introduced, taking into account the NMR spectrum of fat in addition to the other confounding factors
[26-28]. Recently, a complex-based multipoint method was proposed using IDEAL fat-water separation algorithm to solve the dominant component ambiguities problem and thus offering a fat quantification over a full dynamic range of 0-100 % [29-31]. This latter technique, available today only on GEHC MR systems, allows an accurate estimation of the fat fraction and is still currently under investigation.
The aim of this study was to validate a magnitude-based multipoint method, easily transposable on any imaging system and for all current clinical fields for Fat Volume Fraction (FVF) quantification in the liver without dominant component ambiguity. Based on a multiple-angle and multiple gradient echo acquisitions, we present a method correcting for T2* and T1 effect using a T2*- and T1-values estimation for fat and water components and accounting for the complex NMR spectrum using a precalibration procedure and resolving the dominant component ambiguity for a FVF quantification on a dynamic range of 0 to 100 %. Validations of the describe method were then performed (i) on fat-water emulsion phantom at 1.5 and 3.0 T and compared with 1H MRS; (ii) in-vivo at 1.5 T and 3.0 T in a prospective comparative study with histology in a group of 28 patients with chronic liver diseases.
MATERIAL AND METHODS MRI FVF quantification algorithm
This algorithm was developed on Matlab r2010a (The MathWorks, Natick, MA, USA). It is composed of fourth step.
In a first step, separation of fat and water transversal magnetization (M0xyf and M0xyw) was repeated for each weighting (angles) including the T2* estimation of both components from the multiple echoes and images. This procedure was achieved using a 4-parameters interference bi-exponential model integrating the modeling of five fat resonances (at 0.9, (1.3 + 1.6), (2.1 + 2.25), 4.2 and 5.3 ppm) as described by Hamilton et al [32]. The 2.75 ppm CH2 peak was neglected because it contributes to only 0.6% of the total fat and is rarely seen in-vivo. The following equation describes the model used:
Sሺሻ ൌ ቆۻܠܡܟൈ ି܂ܟుכ ቆۻܠܡൈ ି ు
܂כൈ σହ ୬
୬ୀଵ ଶ୧ቇቇ
Eq.1 S(TE) represents the signal intensity according to the echo time, T2*w and T2*f are the estimated T2* of water and fat component respectively. Cn are coefficients representing the part of signal given by the modeled fat resonance n over the total fat signal with C1 = 0.047, C2 = 0.039, C3 = 0.12, C4 = 0.706 and C5 = 0.088. fn corresponds to the chemical shift between water and each modeled fat resonance n in Hz with (f1 = -64.6 Hz, f2 = 63.8 Hz, f3 = 332 Hz, f4 = 432 Hz, f5 = 485 Hz at 3.0T and f1 = -32.3 Hz, f2 = 31.9 Hz, f3 = 166 Hz, f4 = 216 Hz, f5 = 242 Hz at 1.5T).