Skip to main content

Magnetic resonance imaging platform for probing fat microstructure

Periodic Reporting for period 4 - ProFatMRI (Magnetic resonance imaging platform for probing fat microstructure)

Reporting period: 2020-10-01 to 2021-07-31

Metabolic syndrome and osteoporosis are the two metabolic diseases with the highest and most rapidly growing prevalence. Both the metabolic syndrome and osteoporosis are being transformed into a major global health concern with severe socioeconomic impact. The clinical management of the patients with the two metabolic diseases faces different but severe challenges. On the one hand, metabolic syndrome can be diagnosed with established biomarkers, but the selection of optimal prevention strategies for each individual patient is still problematic. For example, it is unclear why certain lifestyle interventions work in some obese subjects and fail in others. On the other hand, osteoporosis can be treated using known medication, but its current early diagnosis remains insufficient. Specifically, the traditional diagnostic biomarker for osteoporosis, the bone mineral density (BMD), has a compromised precision on predicting fractures. The two metabolic diseases have been linked through the role of fat. Fat is central to their incidence and progression, and the probing of fat cellular properties can provide groundbreaking solutions for overcoming the existing challenges in the diseases early diagnosis and prevention.
In metabolic syndrome, there is evidence supporting a role of brown fat in preventing the disease. Fat can be white or brown. White fat stores excess energy, whereas brown fat converts energy into heat. Excessive white fat in different depots is a characteristic of obesity and insulin resistance. Recent studies on human subjects have shown that brown fat prevalence negatively correlates with BMI and visceral fat volume and that brown fat metabolic activity is lower in diabetics. Therefore, there is evidence supporting a role of brown fat in preventing metabolic syndrome. However, there is no established non-invasive tool to reliably measure brown fat. Brown fat has distinctly different microstructure than white fat: brown fat cells are smaller and enclose more mitochondria than white fat cells.
In osteoporosis, there is evidence supporting a role of marrow fat, in combination with bone mineral density, for monitoring fracture risk. Bone marrow includes both fat and hematopoietic red blood cells and it is not considered any more as a simple filler of the bone cavities. Multiple previous works have shown a negative association between marrow fat content and BMD. It was also recently reported that higher marrow fat content is associated with prevalent vertebral fracture, even after adjustment for BMD. There is therefore evidence supporting a role of bone marrow fat, in combination with BMD, in monitoring fracture risk. Bone marrow fat microstructure is affected by both fat cell-specific microstructure (fat cell size and number) and microenvironment (mineral density). However, there is no non-invasive biomarker to measure marrow fat cellular changes in osteoporosis.
Magnetic resonance imaging (MRI) is the ideal modality for non-invasively measuring fat throughout the body. MRI can also provide unique surrogate biomarkers of tissue microstructure, relying on diffusion and magnetic susceptibility effects. In order to differentiate brown from white fat and characterize the relationship between bone mineral and marrow fat cells, the employed MR methodology needs a technical breakthrough, shifting from the state-of-the-art water-centered paradigm to a fat-centered microstructural MRI paradigm. Namely, methods need to be developed focusing entirely on the protons of the fat molecules. ProFatMRI describes an innovative research program that aims to develop and ex vivo validate diffusion and susceptibility MRI biomarkers of fat microstructure, and in vivo apply them at clinical MRI systems.
The resulting technologies will provide novel ways for selecting optimal individualized prevention strategies in metabolic syndrome and for achieving reliable risk fracture prediction in osteoporosis.
We developed and ex vivo validated fat fraction, diffusion, relaxometry, ultra-short echo time (UTE) and susceptibility magnetic resonance imaging (MRI) biomarkers of fat microstructure, and translated them for in vivo applications at clinical MRI systems.

Aim A: Fat-specific diffusion MRI
We developed diffusion-weighted (DW) single-voxel magnetic resonance spectroscopy (MRS) and diffusion-weighted MR imaging techniques targeted for measuring the diffusion properties of fat.

Subaim A1: Fat-specific DW single-voxel MRS
A DW-MRS sequence using STEAM (stimulated echo acquisition model) voxel localization was developed and combined with bipolar gradient readouts to compensate for eddy current effects. An analytical framework of the DW signal dependence on the size of diffusion restriction for spherical lipid droplets was established. To validate the method, oil-in-water emulsions were prepared and examined using DW-MRS, laser deflection and light microscopy. In phantoms, a good correlation between the measured droplet sizes by DW-MRS, laser deflection and microscopy measurements was obtained [https://doi.org/10.1002/mrm.27651]. The in vivo feasibility of the method was studied in the tibia bone marrow of volunteers (region minimally affected by motion) to test the method repeatability and characterize microstructural differences at different marrow locations. The coefficient of variation in bone marrow adipocyte size estimation stayed below 15% and the adipocyte diameter measured distally was smaller than the adipocyte diameter measured proximally, in agreement with previous literature [https://doi.org/10.1002/mrm.27651].
MR table vibrations were an important challenge in high b-value DW-MRS. The problem was in part solved by building an anti-vibration table for scanning small phantoms (with limited table loading). A new technique, employing an additional gradient to reduce the effect of table vibration-induced signal loss, was also developed for in vivo applications [https://doi.org/10.1002/mrm.28128]. The evaluation of the vibration effect reduction technique was validated using laser interferometry measurements, showing significant improvements in measuring diffusion properties of lipids [https://doi.org/10.1002/mrm.28128].

Subaim A2: Fat-specific DW imaging
Single-voxel DW-MRS enabled the quantification of the diffusion properties of the water and fat peaks, but averaged within a single-voxel. We also developed methods for spatially resolving the diffusion properties of fat using DW imaging. Given the need to sensitize slow diffusion processes with strong diffusion encoding and long diffusion times, compensation of motion-induced phase error effects was necessary. We therefore adopted a single-shot 2D spatial encoding design. Given the overall strong sensitivity of single-shot echo planar imaging (EPI) to the chemical shift of the different fat peaks, we decided to adopt a turbo spin echo (TSE) imaging approach using magnitude stabilizing gradients [https://doi.org/10.1002/mrm.28755]. The developed technique used a flexible diffusion preparation module allowing for multiple b-values and mixing times. The developed DW-TSE technique enabled the mapping of spatially resolved lipid droplet sizes. We followed a strategy of directly validating the DW-TSE lipid droplet size mapping results. Specifically, the measurement of the lipid droplet size based on DW-TSE was validated against microscopy in water-fat emulsions and was validated against microscopy in ex vivo human adipose tissue samples [https://doi.org/10.1002/mrm.28755].

Aim B: Fat-specific magnetic susceptibility mapping
We developed methods for measuring the magnetic susceptibility of fatty tissues.

Subaim B1: Fat-specific boundary-based susceptibility estimation
The first step required for measuring magnetic susceptibility was the extraction of the field map. In body applications, a chemical shift encoding-based water-fat separation should be employed for the field map extraction. A critical aspect in performing accurate field-map estimation in tissues containing water and fat was the analysis of effects confounding the water-fat separation. We therefore built a generalized framework for performing water–fat separation under different water–fat signal models [https://qims.amegroups.com/article/view/38368/29074]. The extraction of the field map remains however a non-convex optimization problem with multiple local minima resulting in field map jumps in regions with large B0 inhomogeneities and low signal-to-noise-ratio (SNR). We developed methods improving the robustness of existing chemical shift encoding-based water-fat separation methods by incorporating a priori information of the magnetic field distortions in water–fat separation [https://doi.org/10.1002/mrm.27097] and developed a variable-layer single-min-cut graph-cut algorithm for robust field-mapping in water-fat regions [https://doi.org/10.1002/mrm.28515].

Subaim B2: Fat-specific quantitative susceptibility mapping (QSM)
We developed QSM techniques specifically targeted for the needs of bone imaging. We built QSM algorithms first using traditional approaches performing background field removal and dipole inversion (using L1 or L2 regularization techniques) in two steps [https://doi.org/10.1002/mrm.27531] and later approaches using a water-fat total field inversion methodology (combining background field removal and susceptibility inversion in a single step) [https://doi.org/10.1002/mrm.28903]. Trabecular bone density mapping using QSM was shown for the first time in the calcaneus and was in agreement with bone density measurements from high-resolution MRI and computed tomography (CT) measurements [https://doi.org/10.1002/mrm.27531].
To characterize the effect of bone mineral signal in the gradient echo imaging used for susceptibility mapping, we converted all our multi-echo gradient echo Cartesian sampling sequences to radial sampling sequences and enhanced them with sampling of ultra-short TE echoes (UTEs). UTE imaging was in general prone to artifacts induced by k-space trajectory distortions. A novel method relying on the correction of UTE k-space trajectories based on the gradient impulse response function (GIRF) was developed and applied in the lumbar spine of patients with bone fractures [https://doi.org/10.1002/mrm.28566]. A novel methodology was finally developed for performing water-fat separation of data including only UTE signals in the spine [Kronthaler, Magn Reson Med, in press].

Aim C: Tissue-specific translation and in vivo application
We translated the developed imaging methodologies to the specific needs of imaging two highly relevant regions for the metabolism/nutrition and osteoporosis research: the supraclavicular fat region and the lumbar spine, respectively.

Subaim C1: Differentiating white and brown fat
The initial application of high b-value DW-MRS in the human supraclavicular fat showed a very large sensitivity to physiological motion induced by pulsating vasculature and respiratory motion. We therefore developed a flow-compensated DW-MRS technique and combined it with cardiac and respiratory triggering to mitigate such motion effects. However, the attempted approaches to compensate for motion effects using motion-compensated pulse sequence designs, triggered acquisitions and post-processing did not give satisfactory results in robustly sensitizing the diffusion properties of lipids in the supraclavicular fat across subjects. Low b-value DW-MRS measurements targeting the diffusion properties of the supraclavicular adipose tissue water component as a surrogate marker of brown fat microstructure was evaluated and showed a tendency with lower water ADC values in subjects with larger cold-induced thermogenesis [Wu ISMRM 2020 & manuscript under review].
We applied water-fat imaging methods in the supraclavicular and gluteal fat regions in order to assess the ability of the proton density fat fraction (PDFF) and T2* mapping in differentiating white and brown fat. We applied PDFF mapping in a cohort of 110 subjects and we have been able to show that both supraclavicular PDFF and gluteal PDFF relate to anthropometric and MRI obesity markers [https://doi.org/10.1038/ijo.2017.194] and that supraclavicular PDFF is related to thyroid hormone values [https://doi.org/10.1159/000507294]. We additionally showed that T2* mapping using a 20-echo acquisition enables a differentiation of supraclavicular and gluteal adipose tissue [https://doi.org/10.1002/jmri.26661]. We finally evaluated PDFF against cold induced thermogenesis metrics and showed an association of cold induced thermogenesis with the difference in PDFF between supraclavicular and gluteal regions in normal weight subjects [Held under review].

Subaim C2: Characterizing the relationship between bone mineral and marrow fat cells
The relationship between bone mineral matrix and marrow adiposity has recently attracted significant attention in the bone research community, as highlighted in two review papers published by our group [doi: 10.3389/fendo.2016.00074 & doi: 10.1002/jmri.25769]. We translated our bone marrow PDFF mapping, QSM and UTE frameworks in the spine of both volunteers and patients.
In volunteers, we extensively applied MRS methods in the assessment of vertebral bone marrow fat, addressing the feasibility of extracting bone marrow fat unsaturation [https://doi.org/10.1002/mrm.28453] and measuring bone marrow water T2 relaxation times [https://doi.org/10.1002/nbm.4439]. We additionally applied water-fat imaging methods in the vertebral bone marrow and reported on an anatomical variation of age-related changes in vertebral bone marrow PDFF with most pronounced changes at lower lumbar vertebral levels in both sexes [https://doi.org/10.3389/fendo.2018.00141].
In patients, we investigated the effect of aromatase inhibitor intake on the vertebral bone marrow fat fraction of breast cancer patients [https://doi.org/10.1186/s12891-019-2916-2]. We additionally showed that UTE enables the assessment of bone changes in patients with fractures in comparison to CT [https://doi.org/10.1002/mrm.28566] and that QSM enables the assessment of bone changes in patients with metastatic bone disease (differentiation between osteolytic versus osteoblastic lesions) in comparison to CT [https://doi.org/10.1002/mrm.28903].
1. The development of a diffusion-weighted MRS technique [https://doi.org/10.1002/mrm.27651] and a diffusion-weighted MRI technique [https://doi.org/10.1002/mrm.28755] allows for the first time to measure the lipid droplet size using the gradient hardware of a clinical MRI scanner.
2. The development of a variable-layer single-min-cut graph-cut algorithm enables robust field-mapping in water-fat regions across body parts [https://doi.org/10.1002/mrm.28515].
3. The development of water-fat total field inversion techniques enables robust body QSM in the presence of strong susceptibility inclusions and strong background fields across tissues in body and musculoskeletal imaging [https://doi.org/10.1002/mrm.28903]..
4. The development of UTE techniques that are robust against gradient hardware imperfections [https://doi.org/10.1002/mrm.28566] and can perform water-fat separation based on a single-TE [Kronthaler, Magn Reson Med, in press] improves the quality of UTE imaging in challenging anatomical locations (e.g. spine).
5. The development of PDFF and T2*mapping techniques assists in the differentiation between white and brown adipose tissue in the human supraclavicular fat.
6. The development of novel UTE and QSM methods enables high quality radiation-free CT-like MR imaging of bone.