SEARCH

SEARCH BY CITATION

Keywords:

  • water–fat imaging;
  • fat–water separation;
  • Dixon;
  • chemical-shift;
  • fat quantification

Abstract

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Approximately 130 attendees convened on February 19–22, 2012 for the first ISMRM-sponsored workshop on water–fat imaging. The motivation to host this meeting was driven by the increasing number of research publications on this topic over the past decade. The scientific program included an historical perspective and a discussion of the clinical relevance of water–fat MRI, a technical description of multiecho pulse sequences, a review of data acquisition and reconstruction algorithms, a summary of the confounding factors that influence quantitative fat measurements and the importance of MRI-based biomarkers, a description of applications in the heart, liver, pancreas, abdomen, spine, pelvis, and muscles, an overview of the implications of fat in diabetes and obesity, a discussion on MR spectroscopy, a review of childhood obesity, the efficacy of lifestyle interventional studies, and the role of brown adipose tissue, and an outlook on federal funding opportunities from the National Institutes of Health. Magn Reson Med, 2012. © 2012 Wiley Periodicals, Inc.

Research in water–fat MRI has increased significantly in recent years with the development of robust qualitative and quantitative MRI techniques that utilize chemical-shift principles to separate, identify, and measure water and fat signals. The motivation has been catalyzed partly by the rising prevalence of obesity and type 2 diabetes (T2D) and the need to study body and organ fat deposition in relation to comorbidities.

This was the first ISMRM workshop on water–fat MRI and was held at the Hotel Maya in Long Beach, California, from February 19 to 22, 2012. The meeting brought together the international community to exchange insights, progress, and applications steadily gained over the past years and to stimulate dialog in identifying future research directions. Participants traveled from Canada, France, Germany, Italy, the Netherlands, Norway, Sweden, Switzerland, United Kingdom, China, Japan, South Korea, and the United States. The program involved 25 lectures that provided summaries and focused themes and described innovative advances and applications. A poster session facilitated in-depth discussions and interactions among attendees. More than 60 abstracts were submitted: 52.4% focused on data reconstruction and pulse sequences; 15.9% on skeletal muscle fat; 14.3% on abdominal adiposity, whole-body composition, and image segmentation; and 17.4% on fat in organs.

The program broadly attracted trainees, engineers, and investigators from academia, government, and industry. On the first day, the program focused on pulse sequences and performance limits, mathematical theory, data acquisition and reconstruction algorithms, considerations in signal-to-noise ratio, confounders that influence quantitative fat measurements as a biomarker, and validation techniques. On the second day, the theme transitioned to diagnostic applications in the heart, liver, pancreas, abdomen, spine, pelvis, and muscles. This was followed by an overview of diabetes and obesity, a session on MR spectroscopy (MRS), and an outlook on future research opportunities. On the third day, the discussion focused on childhood obesity, lifestyle interventions, and brown adipose tissue (BAT). This article summarizes the scientific highlights, and the future needs and trends identified from presentations of the 25 invited speakers.

HISTORICAL AND CLINICAL PERSPECTIVES

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

It was an honor to have W. Thomas Dixon (GE Global Research) as the opening keynote speaker and set the tone for this workshop. He described how the idea of “simple spectroscopic imaging” came to him on a dark stormy afternoon in 1983. He entertained the audience with the peer-review process of his now seminal Radiology paper (1), noting the 94 days between initial submission and the revision request, and the additional 44 days between revision submission and acceptance. There were two hurdles: the lack of nomenclature at the time distinguishing radiofrequency-recalled and gradient-recalled echoes, and the general editorial disbelief of his proposed method. Nonetheless, Dr. Dixon recognized two other groups that independently arrived at similar observations (2, 3).

Gary Glover (Stanford University) gave a tour on the historical developments in water–fat MRI. He began with T1-based Short-Tau Inversion Recovery methods and frequency-selective excitation/saturation techniques and progressed to two- and three-echo symmetric and asymmetric sampling approaches to mitigate B0 inhomogeneity and address phase unwrapping with strategic echo time choices and innovative postprocessing algorithms (4–9). He summarized early efforts to account for signal attenuation, spectral broadening, and the multipeak spectrum of fat (10), to estimate the effective signal-to-noise ratio and number of signal averages (NSAs), and to reduce scan time with interleaved and multiecho acquisitions (11, 12). Dr. Glover concluded with a review of N-point schemes. It was informative to see the logical progression in methodological advances over the past 30 years.

Shahid Hussain (University of Nebraska Medical Center) demonstrated the importance of traditional inversion-recovery (Short-Tau Inversion Recovery) and frequency-selective fat suppression (FATSAT, SPECIAL, SPIR, and CHESS) techniques, focusing on liver applications in adults. He discussed the clinical utility of fat suppression and its function in improving lesion conspicuity on T1- and T2-weighted images, expanding the tissue contrast dynamic range, reducing ghost artifacts arising from subcutaneous fat and respiratory motion, and minimizing chemical-shift in diffusion-weighting echo-planar imaging sequences (13). To emphasize the role of fat suppression/detection in abdominal MRI, Dr. Hussain illustrated examples of fat containing lesions that can be diagnosed based on their unique appearances, including hepatocellular adenomas versus fat-containing hepatocellular carcinomas, hepatic lipomas versus hemangiomas, neuroendocrine metastases versus multifocal fat-containing hepatocellular carcinomas, clear-cell type renal cell carcinomas versus angiomyolipoma, and normal pancreas versus acute focal pancreatitis (14). These classical MRI techniques continue to play a central role in daily clinical workflow, despite their sensitivity to B0 and B1 inhomogeneities and the continuing advances of water–fat MRI.

Jeffrey Schwimmer (University of California, San Diego, and Rady Children's Hospital) gave a thought-provoking presentation on pediatric fat imaging. He highlighted the prevalence of childhood obesity and its exponential growth from approximately 2 to 12 million cases between 1980 and 2010. Obese adolescents have a higher mortality rate by the age of 50 in comparison to those with normal weight, with liver and heart diseases, infections, cancer, and diabetes as the predominant causes of death. He focused on visceral adipose tissue (VAT) and emphasized that it is not only the amount of VAT but also the distribution and the lipid particle size within the adipocytes that influence metabolic disease risk. He particularly noted the association between VAT and sleep apnea (15). He summarized the lack of guidelines for the utilization of MRI to measure fat in pediatrics and offered one putative explanation. There is a wide range of fat values in the literature, with dependencies on gender (16), age, and ethnicity. The spectrum is further widened by the use of different MR techniques, including single-slice and multislice imaging, and spectroscopy. Consequently, it is difficult to ascertain in a systematic manner appropriate stratification values for routine clinical diagnosis. He discussed the prevalence of fatty liver in children (17) and argued that while MRI is useful in determining hepatic steatosis, it does not yet provide a myriad of other useful disease-related information that is available from biopsy. Lastly, Dr. Schwimmer described the motivation to standardize quantitative MRI fat measurements. He challenged the community to develop a framework such that an accurate diagnosis of disease, an understanding of disease severity and progression over time, a meaningful measure of intervention and therapeutic efficacy, and a reliable predictor of clinical outcomes can be determined from nominal numbers.

TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Qing-San Xiang (University of British Columbia) comprehensively described one-, two-, and three-echo water–fat MRI methods, utilizing a framework of two rotating vectors to demonstrate each technique and how factors such as time-dependent phase errors from B0 inhomogeneity and time-independent ones from receiver coils affected the signal model. With one-echo strategies, he showed that by judiciously selecting the echo time (TE) such that water and fat vectors are orthogonal, one can separate water and fat into the real- and imaginary-axis of the resultant signal, provided phase errors are removed by calibration or postprocessing (18–21). Next, he discussed two-echo methods with symmetric sampling that yield in- and opposed-phase data, and subsequent determination of the major “big” and minor “small” components, effectively separating water and fat in each voxel. However, identification of water and fat can be ambiguous (e.g., water–fat swaps) due to B0 inhomogeneity. He then demonstrated how asymmetric “partially-opposed-phase” (22) sampling can overcome this ambiguity to achieve consistent water–fat separation and identification (23, 24). He concluded with three-echo methods, summarizing symmetric and asymmetric strategies (25) and early efforts in Tmath image mapping (26), multispectral-peak ethanol imaging, and multicomponent (water/fat/silicone) applications (27). Dr. Xiang concluded that asymmetrical sampling and proper handling of phase are critical to robust water–fat MRI.

Holger Eggers (Philips Research) discussed the benefits of water–fat MRI using three or more echoes. He emphasized that two-point methods are adequate for separating water and fat signals and are sufficient for qualitative use. Expanding on the classical signal model where water, fat, and B0 inhomogeneity are unknowns, Dr. Eggers demonstrated that using three or more echoes allows improved accuracy in water–fat separation and provides additional clinically relevant information. For example, the measurement of more echoes can facilitate increasing tolerance to eddy currents (23, 28), improving B0 inhomogeneity estimation, reducing water–fat swaps, enhancing signal-to-noise ratio, estimating Tmath image of water and fat components (29), and characterizing triglyceride properties such as degree of saturation (30–32). Sampling more echoes does not necessarily come at the expense of increased scan times. Rather, it allows more flexible TE and echo spacing choices. He noted that multiecho water–fat imaging is a subset of chemical-shift imaging, where resonances and relative amplitudes of spectral models of components to be separated are assumed a priori in the former but unconstrained in the latter.

Walter Block (University of Wisconsin-Madison) summarized flavors of water–fat MRI pulse sequences and B0 fieldmap estimation techniques. He reviewed gradient echo (33), spin echo (34), and hybrid designs (35). He highlighted developments in magnetization transfer water–fat imaging and where the absence of magnetization transfer effects in fat is exploited (36). Dr. Block focused on balanced steady state free precession (SSFP) variants, demonstrating fat-suppression (37), water-excitation (38), phase sensitive approaches (39), integration with chemical-shift water–fat schemes (40, 41), fluctuating equilibrium magnetic resonance (42), linear combination approaches (43), and alternating pulse repetition time methods (44, 45) that shape the spectral frequency response to place the stop band over a majority of fat resonances (46). He discussed B0 fieldmap estimation, a critical component in water–fat MRI. Several methods to address a fundamental challenge in avoiding incorrect global minima fieldmap solutions that lead to swaps have been reviewed (47). In overcoming this ambiguity, field map smoothness assumptions have been exploited in region growing (48) and multiresolution approaches (49–52), direct algebraic formulations (53), and regularization schemes (54, 55).

Peter Börnert (Philips Research) emphasized the need to accelerate water–fat MRI, because the chemical-shift encoding process needs extra time, requiring usually the acquisition of multiple data at different TEs. A comprehensive overview of techniques to accelerate multiecho water–fat MRI was provided, summarizing advantages and drawbacks spanning partial-Fourier approaches (56), integration with parallel imaging (57), interleaved fly-back and nonfly-back schemes (58, 59), and advanced methods such as k-space-based water–fat decomposition (60), compressed sensing (61, 62), and compressed sensing with parallel imaging (63, 64). Furthermore, Dr. Börnert highlighted that non-Cartesian methods (60, 65–68) are also compatible with water–fat MRI and have interesting inherent benefits, including robustness to physiological and patient motion and the ability to acquire ultrashort and very long effective TEs (e.g., radial, spiral, and propeller sampling). Some water–fat approaches, especially the non-Cartesian variants, are still in the research phase, but the current combination of parallel imaging with efficient sampling has already facilitated rapid acquisition for several interesting applications in research and clinical practice. Additional work needs to be investigated in data reconstruction, not only to handle large amounts of data from multicoil, multiecho acquisitions, but also for advanced and iterative methods that require additional processing steps to compensate for chemical-shift effects during signal sampling and potential off-resonance blurring.

QUANTITATIVE WATER–FAT MRI

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Claude Sirlin (University of California, San Diego) gave a prelude to quantitative water–fat MRI, emphasizing fat as an objective image-based indicator of biological, pathological, or pathogenic process and a surrogate for a clinical endpoint—an imaging biomarker. He illustrated the concepts of accuracy and precision (69, 70). Accuracy is the correctness of a biomarker measurement in comparison to a reference standard and can be assessed with receiver operating curves and true/false positives/negatives if the metric is dichotomous or with correlation and regression if the measure is continuous or ordinal. Precision is the consistency (and variability) of the biomarker. No reference is needed in assessing precision. Precision is affected by technical components including repeatability (within a site, within examinations, between examinations), reproducibility (between sites, between manufacturers, magnetic field strength, and software platforms), and robustness (invariance to changes in pulse sequence parameters and operator). It is also affected by biological components, including temporal variability (within day, between days, in response to physiological changes) and spatial variability in cross-sectional and longitudinal studies. While accuracy is critical in single-site studies, precision is important in multisite studies, and standardization efforts are needed to properly integrate a biomarker into clinical trials and patient practice.

Catherine Hines (Merck) compared measures of fat fraction from histology, chemical extraction, and MRI/S. MRI/S measures a signal fat fraction. Once J-coupling, T1 and T2 relaxation, noise, phase errors, multiple fat peaks, and other confounders are considered, the resultant metric is a proton-density fat fraction (PDFF), a ratio of unconfounded fat signal to the sum of the unconfounded fat and water signals (71). The PDFF only represents MR-visible signals. Protons originating from macromolecules and solids are MR-invisible and not reflected. She described biopsy/histology grading of fat in the diagnosis of organ steatosis (72, 73), where numerical scores reflect increasing proportions of fat-involved cells. Next, three chemical processes that yield extraction-based fat fractions were illustrated, and Dr. Hines cautioned that these techniques destroy the sample. They can be categorized as: measurement of total lipid mass (Folch method) (74), measurement of triglyceride mass (colorimetric assays) (75), and measurement of triglyceride composition in terms of different fatty acid concentrations (gas chromatography) (76). When normalized to the sample mass, the endpoints are total lipid, triglyceride, and individual fatty acid mass percents, respectively. While a correlation between PDFF and extraction-based fat fraction is expected, an identity association is nontrivial (77), because extraction-based fat fraction includes both MR-visible and invisible components. She concluded her talk with recipes for constructing water–fat emulsions, emphasizing antimicrobial agent (sodium azide), emulsifiers and surfactants (sodium dodecyl sulfate), and agar ingredients. She highlighted D2O (deuterium heavy water) to mimic MR-invisible components.

Mark Bydder (University of California, San Diego) described confounders that affect the accuracy of PDFF, including T1, Tmath image, and multiple fat peaks (78–80). He classified these factors as operational confounders, in the sense that current MR methods for estimating PDFF may also measure other properties of the tissue as well. Confounders can be addressed by either minimizing their effect or postprocessing correction strategies. T1 bias can be mitigated with low flip angles and long pulse repetition times in gradient echo sequences, or corrected using predetermined calibration values. He cautioned that B1 inhomogeneity and variations in T1 between different fat peaks (81) and in water- or fat-dominant tissues (82) should be considered. He reviewed algorithms with single or dual Tmath image terms (29) and demonstrated analytical solutions that modeled the degree of fat fraction over- and under-estimation in two-echo water–fat MRI. Monte-Carlo simulations are needed to describe the effects across arbitrary parameters. He also noted that robustness of the estimated PDFF to changes in T1 and Tmath image can be tested using contrast agents (83). He compared methods that used predefined fat resonances and amplitudes and recent efforts to model peak amplitudes as unknown parameters, describing mathematically each fat resonance as a function of chain length and the number of (double) C[DOUBLE BOND]C and (methylene-interrupted double) C[DOUBLE BOND]C[BOND]C[DOUBLE BOND]C bonds (30–32). Dr. Bydder acknowledged J-coupling, the initial phases of water and fat after radiofrequency excitation (84), temperature, and eddy currents as additional confounders, and commented that some confounders are in fact physical properties of the underlying tissue that may themselves find utility as imaging biomarkers.

Angel Pineda (California State University, Fullerton) discussed noise, error propagation, Cramér-Rao Bound, Fisher Information Matrix, and NSA in water–fat MRI, focusing on 3-echo acquisitions (29, 85–87). He emphasized that the reconstructed image quality of any water–fat MRI algorithm depends on the choice of TEs, echo spacings, and the inherent gaussian noise in the raw data. Dr. Pineda and his collaborators reported that regardless of the underlying pulse sequence, the NSA is a measure of the noise efficiency of an acquisition and varies from 0 to N, where N is the number of echoes used to estimate the water and fat. When NSA is close to zero, parameter estimates become unstable and error-prone. The NSA was originally defined for linear systems where only water and fat were the unknowns but has since been generalized to the Cramér-Rao Bound definition to integrate nonlinear parameter estimates, including object phase, B0 inhomogeneity, and Tmath image. The Cramér-Rao Bound analysis allows one to determine how uncertainties in parameters propagate and to what extent they impact the confidence in water and fat separation. In turn, 2D NSA plots reflect the most optimal, and at times nontrivial, choices in TEs that ensure robust water and fat decomposition with minimal susceptibility to magnet and gradient imperfections. He demonstrated how Cramér-Rao Bound and NSA analysis can also quantify improvements in noise performance when constraints such as B0 fieldmap smoothness are incorporated. Lastly, he showed a sample NSA map generated from the Matlab Toolbox (see below) highlighting the spatial distribution of the variance across the image.

BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Peter Kellman (National Institutes of Health) provided an in-depth look at water–fat MRI in the heart (88, 89) and identified two motivations. First, it facilitates improved visualization of anatomical structures in the heart and provides characterization of myocardial tissue with positive contrast for fat. He showed examples of intramyocardial fat and fibrofatty infiltration, lipid containing masses (lipomas), and epicardial, mediastinal, and pericardial fat surrounding the heart. Second, water–fat MRI can mitigate artifacts and reduce ambiguities that commonly arise from fat. He demonstrated examples where elimination of bright epicardial fat signals improved visualization of the myocardium, detection of subepicardial gadolinium enhancement in nonischemic cardiomyopathies, differentiation of myocardial fibrosis from fibrofatty infiltration, and conspicuity of the parietal pericardium and coronary arteries. He also showed examples where fat signals can obscure detail of the thin myocardial wall due to voxel shifts and signal cancellations from chemical-shift, and where fat infiltration in the myocardium can mimic late gadolinium enhancement and false interpretation of myocardial infarction. Dr. Kellman described electrocardiogram (ECG)-triggered and segmented 2D multislice breath-held cine acquisitions, free-breathing and navigated 3D approaches, and single-shot 2D sequences. He also demonstrated magnetization preparations (dark blood double inversion recovery or inversion recovery for late enhancement). Improved confidence in the interpretation of complicated structures and the potential diagnostic value of identifying fat in the myocardium are top benefits of cardiac water–fat MRI.

Russell Low (Sharp and Children's MRI Center) continued with examples of water–fat MRI in the abdomen, pelvis, and spine. The key element of Dr. Low's presentation was the superior diagnostic image quality achieved with water–fat MRI, where results exhibited homogeneous fat suppression while clinically relevant T1 and T2 tissue contrasts were preserved (90). He compared water–fat MRI to conventional fat-suppressed counterparts and noted B0 inhomogeneity as the main culript of artifacts that can mimic or mask disease and degrade conspicuity of benign and malignant findings. In the abdomen, pelvis, and spine where volume coverage is large, frequency-selective fat saturation strategies can become increasingly sensitive to B0 inhomogeneity, Short-Tau Inversion Recovery methods can become less reliable in suppressing fat signals, and fat-related artifacts can be exacerbated, especially in obese patients. For example, he illustrated a subtle peritoneal tumor in the abdomen that is located in the periphery of the abdominal cavity and where inhomogeneous fat suppression can obscure visualization. He showed examples of renal angiomyolipoma, liposarcoma, and ovarian dermoid that were better visualized with water–fat MRI. Lastly, he commented that a single Dixon acquisition can replace conventional protocols utilizing multiple scan to achieve similar tissue contrast with and without fat suppression (91). The improved efficiency is critical in claustrophobic and pediatric patients.

Scott Reeder (University of Wisconsin-Madison) thoroughly reviewed fat quantification in hepatic steatosis (92). He described hepatic steatosis as abnormal and excessive intracellular accumulation of fat in hepatocytes. Long considered an incidental consequence of other conditions such as diabetes, obesity, and cardiovascular diseases, hepatic steatosis is now recognized as having a causative role in important hepatic and systemic disorders, including nonalcoholic fatty liver disease, fibrosis, and cirrhosis, hepatitis C, and hepatocellular carcinoma. He emphasized the benefits of a MRI-based biomarker for liver fat in contrast to percutaneous biopsy, which is invasive and highly sensitive to sampling variability (93). Accurate measurement of steatosis with MRI permits frequent evaluation of the entire organ with a greatly improved safety profile and reduced cost. It also facilitates monitoring of disease progression and efficacy of treatments in reversing hepatic steatosis. A comparison between magnitude- and complex-based methods in multiecho water–fat MRI was made. The importance of addressing signal confounders was reiterated. These corrections are critical in standardizing PDFF across MRI platforms as an independent biomarker of hepatic steatosis. Dr. Reeder concluded that developments and validations in PDFF and Tmath image-based iron quantification approaches are still needed to mature these metrics into broadly accepted biomarkers.

DIABETES AND OBESITY

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

John Wood (Children's Hospital Los Angeles) focused on iron overload, which occurs in hereditary hemochromatosis, porphyrias, liver diseases, the metabolic syndrome, thalassemia, sickle-cell anemia, and myelodysplastic syndrome. Iron toxicity leads to dysfunctions of the pituitary, thyroid and parathyroid glands, and organs such as the liver, pancreas, and heart. It appears to also impact atherosclerosis, endothelial dysfunction, and osteoporosis. He described the pathophysiology of iron overload, describing its transferrin-mediated uptake into organs. He explained the liver as a high-capacity iron reservoir with low toxicity effects and characterized the pancreas and heart oppositely. Ferritin is responsible for shuttling iron into intracellular lysosomes that protect cells from oxidative effects (94). He highlighted works that correlated Tmath image with iron concentration determined from liver biopsies (95, 96) and pointed to fat-mediated modulations when signal intensity was plotted against TE. The magnitude of the residuals remaining after monoexponential data fitting reflected the level of lipid present, which in turn correlated positively with glucose dysregulation measures (97, 98). Dr. Wood described the relationship between iron in the pancreas and the heart (99). He illustrated subjects with normal glucose tolerance who had higher Tmath image values than those with impaired glucose tolerance and diabetes. There was a pattern where progressive iron in the pancreas preceded iron overload in the heart (100). He noted that iron and fat content, which can be simultaneously measured by water–fat MRI, are both critical biomarkers in diabetes.

Richard Bergman (Cedars-Sinai Medical Center), Editor-in-Chief of Obesity, presented on the pathogenesis of diabetes and obesity. He summarized putative causes of obesity, including high-calorie food and exercise-limited environments, temperature regulation, psychotropic agents, reproductive selection, infectious agents, and sleep deprivation (101), and the negative association between visceral adiposity and insulin sensitivity (102). He described a canine model to study the complex interplay between organ steatosis, body adiposity, insulin regulation, diabetes and beta cell failure, and risk for the metabolic syndrome and highlighted the critical role MRI has played in revealing these relationships. He described a series of events following a six-week high-fat diet regimen, which included notable subcutaneous and VAT accumulation, an expected rise in body weight, an increase in visceral adipocyte size, and elevated fat storage in the liver. The latter in turn leads to a decrease in hepatic insulin clearance and hyperinsulinemia, resulting in insulin resistance at skeletal muscle sites and a rise in nocturnal free fatty acid levels. Dr. Bergman noted that normal functioning beta cells of the pancreatic islets can compensate for insulin resistance and prevent hyperglycemia. Lastly, the relationship between insulin sensitivity and insulin secretion (e.g., disposition index) was described (103), and the metric remains the strongest independent predictor of T2D (104).

Michael Goran (University of Southern California), Editor of Pediatric Obesity, presented on differences in subcutaneous adipose tissue/VAT and hepatic and pancreatic fat between Hispanics and African Americans. He cited disparities between the two groups and cautioned that the underlying mechanisms remain elusive. While VAT is linked to obesity and metabolic dysfunction, this hypothesis leads to a paradox (105), because African Americans have lower VAT and higher subcutaneous adipose tissue compared to Hispanics (106). Hispanics also have higher hepatic and pancreatic fat (107). However, Hispanics and African Americans are equally at risk for obesity and metabolic diseases. Part of this disparity can be attributed to the PNPLA3 allele, a polymorphism that is more prevalent in Hispanics (108, 109). Dr. Goran showed that pancreatic fat is correlated with VAT and hepatic fat but is not related to insulin resistance and beta-cell function (110). Pancreatic fat is higher in prediabetic African Americans than those with normal glucose tolerance but not in Hispanics. He alluded to adipocyte cell size and macrophage inflammation as plausible explanations. When cell size increases, it indicates the inability of pre-existing adipocytes to expand further in number to accommodate additional triglycerides. Larger adipocytes are associated with VAT and hepatic fat, and macrophage accumulation is associated with higher fasting insulin levels and reduced beta-cell function (111). He urged the development of MRI methods to facilitate inflammation and cell size measurements. He emphasized the desire to image adipokine and cytokine release and to dynamically track fat deposition during and after meal intake with MRI.

Jürgen Machann (University of Tübingen) discussed cross-sectional and interventional studies in obesity. He described the utilization of T1-weighted protocols to assess whole-body adiposity (112), single-voxel proton MRS for evaluating hepatic lipids (113, 114) and intramyocellular lipid/extramyocellular lipid (IMCL/EMCL) in the soleus and tibialis anterior muscles, spectral-spatial fat-selective techniques (115, 116), and automated and standardized image segmentations (117) to study the conjunction between adipose tissue/ectopic lipids and insulin resistance. Dr. Machann motivated his talk by highlighting that insulin resistance is reversible and that worsening toward overt T2D can be circumvented by diet and exercise. He described cross-sectional results from two studies comprising 550 subjects at increased risk for T2D as identified by high body-mass-index (BMI > 27 kg/m2), family history of T2D, impaired glucose tolerance, and/or gestational diabetes. Results showed that subjects with comparable BMI can exhibit very different subcutaneous adipose tissue and VAT distributions. Females have ∼15% more body fat than males in all fat depots except in VAT, where males have greater amounts (118). He highlighted that in females, less VAT and lower hepatic lipid and IMCL were associated with better insulin sensitivity and appeared to be a benign form of adiposity (119). He also described a longitudinal study involving 400 subjects enrolled in a 2-year program that included 30% reduced calorie intake from fat, increased fiber uptake, and 3 h of moderate weekly exercise. The study demonstrated that improved insulin sensitivity goes along with a reduction in VAT and hepatic lipid and that lower baseline values in VAT and hepatic lipid are predictors of a success in lifestyle intervention in respect of metabolic parameters (120).

Rosa Tamara Branca (University of North Carolina at Chapel Hill) described MRI of BAT, an organ involved in energy expenditure and thermogenesis (121–123). Although BAT physiology has been thoroughly studied in rodents, the role of human BAT remains in question, especially during childhood growth. Positron emission and computed tomography (PET/CT) is the current standard for BAT imaging. The approach, however, is hindered by radiation exposure, the ability to detect only metabolically active tissue, and sensitivity to environmental temperature. She described the use of chemical-shift water–fat MRI to distinguish BAT and white adipose tissue using fat fraction (124, 125). She then stressed that human BAT imaging is challenging, because in adult humans BAT is present only with scattered distribution and often mixed with white adipose tissue and muscle, and that partial volume effects can lead to misleading results. She explained that intermolecular zero-quantum coherence spectroscopy, with its sensitivity to the intravoxel distribution of water and fat spins, may overcome the partial volume effect and allow users to clearly differentiate white adipose tissue from BAT (126). Dr. Branca concluded by presenting T2-based techniques to assess BAT activity with blood oxygen level dependent MRI in mice, effectively exploiting the greater vascularity and rich mitochondria content of BAT (127).

MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Gavin Hamilton (University of California, San Diego) summarized single-voxel MRS and focused on three themes: the proton spectrum of triglycerides is complex with six or more peaks (81); the fat spectrum is shaped by both J-coupling (128) and contributions from distinct triglyceride proton moieties; the relative area of each fat peak is not fixed but changes depending on the type of triglyceride (129). He described Point Resolved Spectroscopy (PRESS) (130) and Stimulated Echo Acquisition Mode (STEAM) (131) techniques and noted that both are equally popular in the literature. While PRESS benefits from increased signal-to-noise ratio, STEAM is capable of shorter minimum TEs (132). Dr. Hamilton demystified the concept of J-coupling. He showed how J-coupling can lead to nonexponential T2 decay and incorrect fat peak quantification (133), that all fat peaks experience J-coupling, and that J-coupling effects vary between STEAM and PRESS, even at identical TE (134). He explained the MRUI software for MRS analysis (135), and cautioned that the modeling of complex spectra shapes may require multiple Gaussians or Lorentzian fits per fat peak. He discussed the impact of saturation bands in MRS fat quantification and reiterated that like its imaging counterparts, accurate peak-area estimates requires T1 and T2 relaxation. Lastly, he showed that MRS can likewise determine triglyceride carbon chain length, number of double bonds, and number of methylene-interrupted bonds (81) and can detect differences in these three parameters in vivo (136).

Lidia Szczepaniak (Cedars-Sinai Medical Center) next described the application of proton MRS for measurement of triglycerides within the pancreas and the heart (137–142). She motivated her talk by introducing a hypothesis on organ steatosis and lipotoxicity, which states that triglycerides overflow to ectopic sites due to inefficient fat storage by adipose tissue depots. Furthermore, chronic deposits of triglycerides within the cytosol of parenchymal cells lead to steatosis and organ dysfunction. She described strategies for the accurate prescription of MRS voxels in each organ. She emphasized the importance of respiratory and cardiac motion compensation and correction, discussed the lower dynamic range of fat fraction in the pancreas and the heart, and summarized normal and abnormal ranges of triglyceride levels. She also reviewed studies demonstrating the development of fatty pancreas and heart in stages that worsen with obesity and glucose tolerance impairment, and noted that steatosis in these organs establishes prior to the onset of diabetes and does not change significantly further with disease progression. Dr. Szczepaniak discussed how fatty organs are dysfunctional. For the pancreas, fat content tracks positively with the duration of obesity, compensatory insulin secretion, and disposition index, but not with insulin sensitivity. For the heart, she showed positive associations between organ failure and myocardial steatosis in the septum and reduced myocardial contractility in fatty hearts.

Chris Boesch (University of Bern) described fat quantification in skeletal muscles, focusing on the ability of MRS to differentiate IMCL and EMCL (143, 144). Whereas IMCL consists of spherical fat droplets and provide energy to muscles metabolism, EMCL lipids are arranged in parallel plates and exhibit characteristics similar to adipose tissue. Magnetic susceptibility causes the frequency difference in the [BOND]CH2[BOND] resonances between IMCL and EMCL (144, 145), and maximal difference occurs when the muscle fibers and the magnetic field are parallel. He emphasized that IMCL levels can fluctuate with insulin sensitivity, physical activity, diet, pharmacological effects, and others influences. In particular, obese, sedentary, and diabetic patients are characterized by higher IMCL levels (146) and lower insulin sensitivity (147, 148). However, in athletes, a paradox exists where both high IMCL and insulin sensitivity are observed (149–151). Dr. Boesch stressed the need for standardized IMCL measurements, because preceding diet and physical activity can strongly influence baseline levels (152). He touched on diffusion-weighted characterization of IMCL (153). As IMCL experiences hindered diffusion, one challenge has been the design of sequences that utilize strong diffusion gradients and are robust to phase errors from physiological motion. He outlined the potential of using other metabolites in the proton spectrum, such as acetylcarnitine (154, 155) to study generalized aspects of lipid metabolism. He emphasized that 13C and 31P MRS can provide additional information on substrate selection in muscle that may complement the knowledge on skeletal muscle lipid metabolism.

Fritz Schick (University of Tübingen) described peripheral yellow and vertebral red marrow. He demonstrated hematological diseases where the water and fat content in red marrow changes and how MRI/S can monitor the efficacy of therapeutic interventions (113, 156, 157). In contrast to yellow marrow, which contains predominantly fat, red marrow is characterized by similar proportions of water from hemopoietic cells, extracellular fluid (plasma), connective tissue, and fat (adipocytes). Red marrow consequently exhibits marked microscopic field inhomogeneities due to the bony trabecular structure and increased iron deposition (hemosiderin), leading to broader MRS resonances. Dr. Schick illustrated examples from acute myeloid leukemia patients, where red marrow is initially characterized by high cellularity and very little fat signal, because adipocytes are replaced by tumor cells. He then showed follow-up examples where effective chemotherapy led to a reduction of cellularity—a decrease in water signal and progressive increase in fat signal—due to reconstitution of adipocytes in the red marrow. He also showed the efficacy of stem cell transplantation in cases of myeloma, leukemia, lymphoma, neuroblastoma, and Ewing's sarcoma. In this procedure, red marrow is first ablated by high-dose chemotherapy and/or total body irradiation, destroying all hemopoietic cells in the process. Consequently, cellularity is extremely low. The red marrow is then repopulated with transfusion of autologous or allogenic stem cells, and the resultant increase in cellularity and reconstitution of normal marrow composition can be visualized with water–fat MRI, magnetization transfer (158), and diffusion pulse sequences. Magnetization transfer and diffusion contrast mechanisms exhibit higher sensitivity to changes in cellularity immediately following transplantation.

OUTLOOK

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Maren Laughlin (National Institutes of Health) provided a roadmap in fat research. She emphasized the need for more in-depth mechanistic studies of childhood obesity, the maternal effects of obesity to offspring, and the relations between obesity and its comorbidities, particularly in minorities. She identified funding opportunities to promote interdisciplinary collaborations (PAR-11-221), to develop interventions and treatments of overweight/obesity (PAR-09-124), to create educational opportunities in disease-oriented biomedical research (PAR-10-092), and to stimulate ongoing work in nonalcoholic steatohepatitis, bariatric surgery, chronic renal insufficiency, diabetes, and adipocyte biology (PAR-09-247). She underlined BAT, citing that its incidence and metabolic activity, and its response to pharmacological agents, temperature, environment, diet, and exercise, remain largely unknown. She stressed the need to investigate fat as a complex organ. Fat depots differ in physiology (e.g., endocrine functions, metabolism, regulation, and cell lineage), and the interplay between depots in diseases has not been thoroughly studied. She highlighted the need to assess the fat redistribution due to lifestyle interventions such as bariatric surgery or weight/loss gain due to diet/exercise, along with implications in metabolism and glucose regulation, and disease risks. Questions remain whether certain depots and fat patterns pose more health risks than others and whether certain characteristics (triglyceride vs. fatty free acids and saturated vs. unsaturated) are associated with healthy phenotypes. Lastly, Dr. Laughlin challenged the need to develop MRI methods for monitoring the progression of adipocyte macrophage inflammation and cell size, and whether these properties vary with chemical profiles of the stored fat, alterations in body composition, and metabolic diseases.

POSTER AWARDS AND MATLAB TOOLBOX

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Three top-scoring posters were recognized: “Prospectively Accelerated Water–Fat Separation Using Parallel Imaging and Compressed Sensing” by Samir Sharma (University of Southern California), “Fat–Water Separation and Tmath image Estimation Based on Discrete Whole-Image Optimization in 3D” by Johan Berglund (Uppsala University), and “Fast Field Map Estimation with Multi-Labeling Continuous Max-Flow” by Abraam Soliman (University of Western Ontario).

Diego Hernando (University of Wisconsin-Madison) spearheaded an initiative to develop a Matlab Toolbox for water–fat data reconstruction. The goals were to establish a resource for investigators to implement previously published algorithms without having to “reinvent the wheel,” and to provide an interface that utilized a common input/output structure. The toolbox consisted of:

  • Reconstruction algorithms for 2D/3D Cartesian and non-Cartesian (e.g., spiral, concentric rings) data, and undersampled compressed sensing trajectories were provided. Users can toggle and select single or multiple peak fat spectral modeling, the saturation degree of triglycerides, magnitude or complex data fitting, initial phase constraints, Tmath image fitting, and various B0 fieldmap estimation approaches. Postprocessing such as image deblurring and the computation of fat fraction and NSA maps were also included. Instructions and descriptions of parameters were provided.

  • Multiple data sets including 2D/3D, single/multicoil, and 2–15 echoes, from cardiac, abdomen, thigh, knee, ankle, and phantoms, acquired at 1.5 and 3.0 T, were provided. Code to generate synthetic data was also included.

The toolbox was distributed to workshop attendees and can be downloaded from the ISMRM website. Attendees were encouraged to “plug-n-play” the existing code, tweak and improve the algorithms, integrate the code building blocks to tailor to their specific applications, and provide feedback. Along with Diego Hernando, contributions came from Johan Berglund, Emily Bice, Mark Bydder, Mariya Doneva, Holger Eggers, Houchun Hu, Yun Jiang, Peter Kellman, Wenmiao Lu, Angel Pineda, Samir Sharma, Jeffrey Tsao, and Holden Wu. Investigators interested in gaining toolbox access and contributing to future versions, which are planned to be made available to the general ISMRM community, should contact Diego Hernando (dhernando@wisc.edu) or Houchun Hu (houchunh@usc.edu).

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Water–fat MRI methods are a critical toolset, and their utility is gaining momentum. Nearly 30 years after Dixon's seminal manuscript on “simple proton spectroscopic imaging,” new findings, novel methodologies, and innovative applications continue to progress. The water–fat community is strong and active. The workshop was successful in fostering collaborations and dialog, identifying challenges and opportunities, and promoting consensus in standardizing qualitative and quantitative water–fat techniques. As a result of the collective efforts of many talented scientists who have advanced this field, our knowledge of water–fat MRI has significantly expanded, the topic has greatly matured, and applications that can benefit from water–fat MRI have broadened. Accompanying this summary is a MRM Virtual Issue on the journal's website that lists water–fat MRI contributions in 2010 and 2011.

Acknowledgements

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

The organizers are grateful to the ISMRM staff for their assistance. The authors thank ISMRM's corporate members, and specifically Joanna Jobson (GE Healthcare), Gordon Herron (Philips Healthcare), and Stuart Clarkson (Toshiba America Medical Systems), for their sponsorship. They thank the speakers for their contributions and the attendees for their participation.

REFERENCES

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. HISTORICAL AND CLINICAL PERSPECTIVES
  4. TECHNIQUE AND RECONSTRUCTION CONSIDERATIONS
  5. QUANTITATIVE WATER–FAT MRI
  6. BODY AND CARDIAC APPLICATIONS: ORGANS, ABDOMEN, PELVIS, AND SPINE
  7. DIABETES AND OBESITY
  8. MR SPECTROSCOPY OF ORGANS, MUSCLES, AND BONE MARROW
  9. OUTLOOK
  10. POSTER AWARDS AND MATLAB TOOLBOX
  11. CONCLUSIONS
  12. Acknowledgements
  13. REFERENCES
  14. Supporting Information
FilenameFormatSizeDescription
MRM_24369_sm_SuppInfo.tif27671KTat water photo group

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.