Test-retest reliability of rapid whole body and compartmental fat volume quantification on a widebore 3T MR system in normal-weight, overweight, and obese subjects

Automated Deep Learning-Based Segmentation of Abdominal Adipose Tissue on Dixon MRI in Adolescents: A Prospective Population-Based Study

BACKGROUND. The prevalence of childhood obesity has increased significantly worldwide, highlighting a need for accurate noninvasive quantification of body fat distribution in children. OBJECTIVE. The purpose of this study was to develop and test an automated deep learning method for subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) segmentation using Dixon MRI acquisitions in adolescents. METHODS. This study was embedded within the Generation R Study, a prospective population-based cohort study in Rotterdam, The Netherlands. The current study included 2989 children (1432 boys, 1557 girls; mean age, 13.5 years) who underwent investigational whole-body Dixon MRI after reaching the age of 13 years during the follow-up phase of the Generation R Study. A 2D competitive dense fully convolutional neural network model (2D-CDFNet) was trained from scratch to segment abdominal SAT and VAT using Dixon MRI-based images. The model underwent training, validation, and testing in 62, eight, and 15 children, respectively, who were selected by stratified random sampling, with manual segmentations used as reference. Segmentation performance was assessed using the Dice similarity coefficient and volumetric similarity. Two observers independently performed subjective visual assessments of automated segmentations in 504 children, selected by stratified random sampling, with undersegmentation and oversegmentation scored on a scale of 0-3 (with a score of 3 denoting nearly perfect segmentation). For 2820 children for whom complete data were available, Spearman correlation coefficients were computed among MRI measurements and BMI and dual-energy x-ray absorptiometry (DEXA)-based measurements. The model used (gitlab.com/radiology/msk/genr/abdomen/cdfnet) is publicly available. RESULTS. In the test dataset, the mean Dice similarity coefficient and mean volu-metric similarity, respectively, were 0.94 ± 0.03 [SD] and 0.98 ± 0.01 [SD] for SAT and 0.85 ± 0.05 and 0.92 ± 0.04 for VAT. The two observers assigned a score of 3 for SAT in 94% and 93% for the undersegmentation proportion and in 99% and 99% for the oversegmentation proportion, and they assigned a score of 3 for VAT in 99% and 99% for the undersegmentation proportion and in 95% and 97% for the oversegmentation proportion. Correlations with SAT and VAT were 0.808 and 0.698 for BMI and 0.941 and 0.801 for DEXA-derived fat mass. CONCLUSION. We trained and evaluated the 2D-CDFNet model on Dixon MRI in adolescents. Quantitative and qualitative measures of automated SAT and VAT segmentations indicated strong model performance. CLINICAL IMPACT. The automated model may facilitate large-scale studies investigating abdominal fat distribution on MRI among adolescents as well as associations of fat distribution with clinical outcomes.

A Combined Region- and Pixel-Based Deep Learning Approach for Quantifying Abdominal Adipose Tissue in Adolescents Using Dixon Magnetic Resonance Imaging

BACKGROUND

The development of adipose tissue during adolescence may provide valuable insights into obesity-associated diseases. We propose an automated convolutional neural network (CNN) approach using Dixon-based magnetic resonance imaging (MRI) to quantity abdominal subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) in children and adolescents.

METHODS

474 abdominal Dixon MRI scans of 136 young healthy volunteers (aged 8-18) were included in this study. For each scan, an axial fat-only Dixon image located at the L2-L3 disc space and another image at the L4-L5 disc space were selected for quantification. For each image, an outer and an inner region around the abdomen wall, as well as SAT and VAT pixel masks, were generated by expert readers as reference standards. A standard U-Net CNN architecture was then used to train two models: one for region segmentation and one for fat pixel classification. The performance was evaluated using the dice similarity coefficient (DSC) with fivefold cross-validation, and by Pearson correlation and the Student's t-test against the reference standards.

RESULTS

For the DSC results, means and standard deviations of the outer region, inner region, SAT, and VAT comparisons were 0.974 ± 0.026, 0.997 ± 0.003, 0.981 ± 0.025, and 0.932 ± 0.047, respectively. Pearson coefficients were 1.000 for both outer and inner regions, and 1.000 and 0.982 for SAT and VAT comparisons, respectively (all p = NS).

CONCLUSION

These results show that our method not only provides excellent agreement with the reference SAT and VAT measurements, but also accurate abdominal wall region segmentation. The proposed combined region- and pixel-based CNN approach provides automated abdominal wall segmentation as well as SAT and VAT quantification with Dixon MRI and enables objective longitudinal assessment of adipose tissues in children during adolescence.

Whole-body Composition Profiling Using a Deep Learning Algorithm: Influence of Different Acquisition Parameters on Algorithm Performance and Robustness

OBJECTIVES

To develop, test, and validate a body composition profiling algorithm for automated segmentation of body compartments in whole-body magnetic resonance imaging (wbMRI) and to investigate the influence of different acquisition parameters on performance and robustness.

MATERIALS AND METHODS

A segmentation algorithm for subcutaneous and visceral adipose tissue (SCAT and VAT) and total muscle mass (TMM) was designed using a deep learning U-net architecture convolutional neuronal network. Twenty clinical wbMRI scans were manually segmented and used as training, validation, and test datasets. Segmentation performance was then tested on different data, including different magnetic resonance imaging protocols and scanners with and without use of contrast media. Test-retest reliability on 2 consecutive scans of 16 healthy volunteers each as well as impact of parameters slice thickness, matrix resolution, and different coil settings were investigated. Sorensen-Dice coefficient (DSC) was used to measure the algorithms' performance with manual segmentations as reference standards. Test-retest reliability and parameter effects were investigated comparing respective compartment volumes. Abdominal volumes were compared with published normative values.

RESULTS

Algorithm performance measured by DSC was 0.93 (SCAT) to 0.77 (VAT) using the test dataset. Dependent from the respective compartment, similar or slightly reduced performance was seen for other scanners and scan protocols (DSC ranging from 0.69-0.72 for VAT to 0.83-0.91 for SCAT). No significant differences in body composition profiling was seen on repetitive volunteer scans (P = 0.88-1) or after variation of protocol parameters (P = 0.07-1).

CONCLUSIONS

Body composition profiling from wbMRI by using a deep learning-based convolutional neuronal network algorithm for automated segmentation of body compartments is generally possible. First results indicate that robust and reproducible segmentations equally accurate to a manual expert may be expected also for a range of different acquisition parameters.

Reproducibility of semiautomated body composition segmentation of abdominal computed tomography: a multiobserver study

Background

Segmentation of computed tomography (CT) images provides quantitative data on body tissue composition, which may greatly impact the development and progression of diseases such as type 2 diabetes mellitus and cancer. We aimed to evaluate the inter- and intraobserver variation of semiautomated segmentation, to assess whether multiple observers may interchangeably perform this task.

Methods

Anonymised, unenhanced, single mid-abdominal CT images were acquired from 132 subjects from two previous studies. Semiautomated segmentation was performed using a proprietary software package. Abdominal muscle compartment (AMC), inter- and intramuscular adipose tissue (IMAT), visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) were identified according to pre-established attenuation ranges. The segmentation was performed by four observers: an oncology resident with extensive training and three radiographers with a 2-week training programme. To assess interobserver variation, segmentation of each CT image was performed individually by two or more observers. To assess intraobserver variation, three of the observers did repeated segmentations of the images. The distribution of variation between subjects, observers and random noise was estimated by a mixed effects model. Inter- and intraobserver correlation was assessed by intraclass correlation coefficient (ICC).

Results

For all four tissue compartments, the observer variations were far lower than random noise by factors ranging from 1.6 to 3.6 and those between subjects by factors ranging from 7.3 to 186.1. All interobserver ICC was ≥ 0.938, and all intraobserver ICC was ≥ 0.996.

Conclusions

Body composition segmentation showed a very low level of operator dependability. Multiple observers may interchangeably perform this task with highly reproducible results.

Electronic supplementary material

The online version of this article (10.1186/s41747-019-0122-5) contains supplementary material, which is available to authorized users.

Measuring liver fat fraction with complex-based chemical shift MRI: the effect of simplified sampling protocols on accuracy

BACKGROUND

The assessment of liver percentage fat fraction (%FF) using proton density fat fraction sequences is becoming increasingly accessible. Previous studies have tended to use multiple small ROIs that focus on Couinaud segments. In an effort to simplify day-to-day analysis, this study assesses the impact of using larger, elliptical ROIs focused on a single hepatic lobe. Additionally, we assess the impact of sampling fewer transhepatic slices when measuring %FF.

METHODS

Retrospective analysis of prospectively obtained images from 34 volunteers using an IDEAL IQ sequence. Two observers independently measured %FF using three different protocols: freehand whole-liver ROI (fh-ROI), elliptical-ROI on the right lobe (rt-ROI) and elliptical-ROI on the left lobe (lt-ROI).

RESULTS

Inter-observer reliability for all measurements techniques was 'excellent' (Spearman's rank correlation coefficients 0.81-0.98). There was a significant difference (Paired Wilcoxon Test: p < 0.001) between the median %FF obtained using fh-ROI when compared to the rt-ROI method, the maximum mean difference between the two techniques was 2.79% (95% CI). For all sampling methods a Kruskall-Wallis analysis demonstrated no significant difference in mean %FF when the number of slices sampled was reduced from 11 to 1. The mean coefficient of variance increased when more slices were sampled (3 slices = 0.1, 11 slices = 0.17, p < 0.001).

CONCLUSION

Simplified ROIs focused on one hepatic lobe provide %FF measurements that are unlikely to be sufficiently accurate for use in clinical practice. Freehand whole-liver ROIs should be used in preference. A single freehand ROI measurement taken at the level of the hepatic hilum yields a %FF that is representative of the mean whole liver % FF. Multiple slices are needed to measure heterogeneity.

MRI adipose tissue and muscle composition analysis-a review of automation techniques

MRI is becoming more frequently used in studies involving measurements of adipose tissue and volume and composition of skeletal muscles. The large amount of data generated by MRI calls for automated analysis methods. This review article presents a summary of automated and semi-automated techniques published between 2013 and 2017. Technical aspects and clinical applications for MRI-based adipose tissue and muscle composition analysis are discussed based on recently published studies. The conclusion is that very few clinical studies have used highly automated analysis methods, despite the rapidly increasing use of MRI for body composition analysis. Possible reasons for this are that the availability of highly automated methods has been limited for non-imaging experts, and also that there is a limited number of studies investigating the reproducibility of automated methods for MRI-based body composition analysis.

Advanced body composition assessment: from body mass index to body composition profiling

This paper gives a brief overview of common non-invasive techniques for body composition analysis and a more in-depth review of a body composition assessment method based on fat-referenced quantitative MRI. Earlier published studies of this method are summarized, and a previously unpublished validation study, based on 4753 subjects from the UK Biobank imaging cohort, comparing the quantitative MRI method with dual-energy X-ray absorptiometry (DXA) is presented. For whole-body measurements of adipose tissue (AT) or fat and lean tissue (LT), DXA and quantitative MRIs show excellent agreement with linear correlation of 0.99 and 0.97, and coefficient of variation (CV) of 4.5 and 4.6 per cent for fat (computed from AT) and LT, respectively, but the agreement was found significantly lower for visceral adipose tissue, with a CV of >20 per cent. The additional ability of MRI to also measure muscle volumes, muscle AT infiltration and ectopic fat, in combination with rapid scanning protocols and efficient image analysis tools, makes quantitative MRI a powerful tool for advanced body composition assessment.

Precision of MRI-based body composition measurements of postmenopausal women

OBJECTIVES

To determine precision of magnetic resonance imaging (MRI) based fat and muscle quantification in a group of postmenopausal women. Furthermore, to extend the method to individual muscles relevant to upper-body exercise.

MATERIALS AND METHODS

This was a sub-study to a randomized control trial investigating effects of resistance training to decrease hot flushes in postmenopausal women. Thirty-six women were included, mean age 56 ± 6 years. Each subject was scanned twice with a 3.0T MR-scanner using a whole-body Dixon protocol. Water and fat images were calculated using a 6-peak lipid model including R2*-correction. Body composition analyses were performed to measure visceral and subcutaneous fat volumes, lean volumes and muscle fat infiltration (MFI) of the muscle groups' thigh muscles, lower leg muscles, and abdominal muscles, as well as the three individual muscles pectoralis, latissimus, and rhomboideus. Analysis was performed using a multi-atlas, calibrated water-fat separated quantification method. Liver-fat was measured as average proton density fat-fraction (PDFF) of three regions-of-interest. Precision was determined with Bland-Altman analysis, repeatability, and coefficient of variation.

RESULTS

All of the 36 included women were successfully scanned and analysed. The coefficient of variation was 1.1% to 1.5% for abdominal fat compartments (visceral and subcutaneous), 0.8% to 1.9% for volumes of muscle groups (thigh, lower leg, and abdomen), and 2.3% to 7.0% for individual muscle volumes (pectoralis, latissimus, and rhomboideus). Limits of agreement for MFI was within ± 2.06% for muscle groups and within ± 5.13% for individual muscles. The limits of agreement for liver PDFF was within ± 1.9%.

CONCLUSION

Whole-body Dixon MRI could characterize a range of different fat and muscle compartments with high precision, including individual muscles, in the study-group of postmenopausal women. The inclusion of individual muscles, calculated from the same scan, enables analysis for specific intervention programs and studies.

Automatic Measurement of the Total Visceral Adipose Tissue From Computed Tomography Images by Using a Multi-Atlas Segmentation Method

BACKGROUND

The visceral adipose tissue (VAT) volume is a predictive and/or prognostic factor for many cancers. The objective of our study was to develop an automatic measurement of the whole VAT volume using a multi-atlas segmentation (MAS) method from a computed tomography.

METHODS

A total of 31 sets of whole-body computed tomography volume data were used. The reference VAT volume was defined on the basis of manual segmentation (VATMANUAL). We developed an algorithm, which measured automatically the VAT volumes using a MAS based on a nonrigid volume registration algorithm coupled with a selective and iterative method for performance level estimation (SIMPLE), called VATMAS_SIMPLE. The results were evaluated using intraclass correlation coefficient and dice similarity coefficients.

RESULTS

The intraclass correlation coefficient of VATMAS_SIMPLE was excellent, at 0.976 (confidence interval, 0.943-0.989) (P < 0.001). The dice similarity coefficient of VATMAS_SIMPLE was also good, at 0.905 (SD, 0.076).

CONCLUSIONS

This newly developed algorithm based on a MAS can measure accurately the whole abdominopelvic VAT.

Sedentary Time and MRI-Derived Measures of Adiposity in Active Versus Inactive Individuals

OBJECTIVE

The aim of this study was to examine cross-sectional associations between objectively measured sedentary time and magnetic resonance imaging (MRI)-assessed adiposity in a population at high risk for type 2 diabetes (T2DM) and to determine whether associations are modified by the recommended levels of moderate-to-vigorous physical activity (MVPA).

METHODS

Sedentary time and MVPA were measured objectively by using accelerometers. Linear regression models examined the association of sedentary time with liver, visceral, subcutaneous, and total abdominal fat (quantified by using MRI). Interaction terms determined whether results were consistent across activity categories (active [> 150 min/wk of MVPA] vs. inactive [< 150 min/wk of MVPA]).

RESULTS

One hundred and twenty-four participants (age = 64.0 ± 7.1 years; male = 65.3%; BMI = 31.8 ± 5.6 kg/m² ) were included. Following adjustment, each 60 minutes of sedentary time was associated with 1.74 L higher total abdominal fat, 0.62 L higher visceral fat, 1.14 L higher subcutaneous fat, and 1.86% higher liver fat. When results were stratified by MVPA (active vs. inactive), sedentary time was associated with greater liver, visceral, and total abdominal fat in the inactive group only.

CONCLUSIONS

These findings suggest that sedentary time is associated with higher levels of inter- and intraorgan fat, but associations with liver, visceral, and total abdominal fat were stronger in those who do not reach the current exercise recommendations for health.

Whole-body adipose tissue and lean muscle volumes and their distribution across gender and age: MR-derived normative values in a normal-weight Swiss population

PURPOSE

To determine age- and gender-dependent whole-body adipose tissue and muscle volumes in healthy Swiss volunteers in Dixon MRI in comparison with anthropometric and bioelectrical impedance (BIA) measurements.

METHODS

Fat-water-separated whole-body 3 Tesla MRI of 80 healthy volunteers (ages 20 to 62 years) with a body mass index (BMI) of 17.5 to 26.2 kg/m² (10 men, 10 women per decade). Age and gender-dependent volumes of total adipose tissue (TAT), visceral adipose tissue (VAT), total abdominal subcutaneous adipose tissue (ASAT) and total abdominal adipose tissue (TAAT), and the total lean muscle tissue (TLMT) normalized for body height were determined by semi-automatic segmentation, and correlated with anthropometric and BIA measurements as well as lifestyle parameters.

RESULTS

The TAT, ASAT, VAT, and TLMT indexes (TATi, ASATi, VATi, and TLMTi, respectively) (L/m²  ± standard deviation) for women/men were 6.4 ± 1.8/5.3 ± 1.7, 1.6 ± 0.7/1.2 ± 0.5, 0.4 ± 0.2/0.8 ± 0.5, and 5.6 ± 0.6/7.1 ± 0.7, respectively. The TATi correlated strongly with ASATi (r > 0.93), VATi, BMI and BIA (r > 0.70), and TAATi (r > 0.96), and weak with TLMTi for both genders (r > -0.34). The VAT was the only parameter showing an age dependency (r > 0.32). The BMI and BIA showed strong correlation with all MR-derived adipose tissue volumes. The TAT mass was estimated significantly lower from BIA than from MRI (both genders P < .001; mean bias -5 kg).

CONCLUSIONS

The reported gender-specific MRI-based adipose tissue and muscle volumes might serve as normative values. The estimation of adipose tissue volumes was significantly lower from anthropometric and BIA measurements than from MRI. Magn Reson Med 79:449-458, 2018. © 2017 International Society for Magnetic Resonance in Medicine.