- Discussion
9
- 10.1002/hep.27808
Is it time to advance pediatric NAFLD diagnosis to the magnetic resonance imaging era?
- Apr 25, 2015
- Hepatology
- Miriam B Vos
Is it time to advance pediatric NAFLD diagnosis to the magnetic resonance imaging era?
To propose a motion-robust chemical shift-encoded (CSE) method with high signal-to-noise (SNR) for accurate quantification of liver proton density fat fraction (PDFF) and . A free-breathing multi-repetition 2D CSE acquisition with motion-corrected averaging using nonlocal means (NLM) was proposed. PDFF and quantified with 2D CSE-NLM were compared to two alternative 2D techniques: direct averaging and single acquisition (2D 1ave) in a digital phantom. Further, 2D NLM was compared in patients to 3D techniques (standard breath-hold, free-breathing and navigated), and the alternative 2D techniques. A reader study and quantitative analysis (Bland-Altman, correlation analysis, paired Student's t-test) were performed to evaluate the image quality and assess PDFF and measurements in regions of interest. In simulations, 2D NLM resulted in lower standard deviations (STDs) of PDFF (2.7%) and (8.2 ) compared to direct averaging (PDFF: 3.1%, : 13.6 ) and 2D 1ave (PDFF: 8.7%, : 33.2 ). In patients, 2D NLM resulted in fewer motion artifacts than 3D free-breathing and 3D navigated, less signal loss than 2D direct averaging, and higher SNR than 2D 1ave. Quantitatively, the STDs of PDFF and of 2D NLM were comparable to those of 2D direct averaging (p>0.05). 2D NLM reduced bias, particularly in (-5.73 to -0.36 ) that arises in direct averaging (-3.96 to 11.22 ) in the presence of motion. 2D CSE-NLM enables accurate mapping of PDFF and in the liver during free-breathing.
Is it time to advance pediatric NAFLD diagnosis to the magnetic resonance imaging era?
Is it time to advance pediatric NAFLD diagnosis to the magnetic resonance imaging era?
Correcting gradient chain induced fat quantification errors in radial multi-echo Dixon imaging using a gradient modulation transfer function
Purpose: Multi-echo Stack-of-stars (SoS) radial k-space trajectories with golden angle ordering are becoming popular for free-breathing abdominal Dixon imaging and proton density fat fraction (PDFF) mapping. Gradient chain imperfections including eddy currents and system delays are known to affect the image quality of radial imaging and to confound the estimation of PDFF mapping. This work proposes a retrospective trajectory correction method based on a simple gradient modulation transfer function (GMTF) measurement to predict and correct gradient chain induced k-space trajectory errors.Methods: The GMTF was measured using the standard hardware of a 3 Tesla scanner on a phantom using the thin slice method and was applied to a 3D radial SoS Dixon imaging sequence. The impact of the GMTF-based correction on image reconstruction and PDFF quantification was investigated using numerical simulations and validated on experimental phantom data as well as on in vivo leg and liver data of healthy volunteers.Results: Correcting the k-space trajectories with the measured GMTF during image reconstruction reduced PDFF quantification errors for phantom and in vivo acquisitions. A Bland-Altman comparison of the measured PDFF phantom and reference data confirmed that the GMTF correction narrowed down the limits of agreement (LoA) from 1.3% ± 8.1% (uncorrected) to 1.9% ± 5.4% (GMTF-corrected) over the full PDFF range (0%–100%) and from −0.26% ± 2.8% (uncorrected) to 0.12% ± 1.5% (GMTF-corrected) within the 0%–50% PDFF range. Liver PDFF estimation was improved by reducing the standard deviation of the mean liver PDFF and the bias of the mean liver PDFF for all subjects.Conclusion: The proposed GMTF-based k-space trajectory correction is a fast alternative method for avoiding PDFF quantitation errors caused by gradient-system induced k-space trajectory errors in 3D radial multi-echo gradient-echo acquisitions.
Read moreStability of liver proton density fat fraction and changes in R 2* measurements induced by administering gadoxetic acid at 3T MRI.
To assess changes in liver proton density fat fraction (PDFF) and R 2* measurements in the presence of changes in tissue relaxation rates induced by administrating gadoxetic acid, using two different image reconstruction methods at 3T MRI. Forty-five patients were imaged at 3T with chemical-shift-based MRI sequences before and 20min after administration of gadoxetic acid. Image reconstructions were performed using hybrid and complex methods to obtain PDFF and R 2* images. A single radiologist measured PDFF and R 2* values on precontrast and postcontrast images. Precontrast and postcontrast PDFF values were compared using intraclass correlation coefficient (ICC), linear regression, and Bland-Altman analysis. Changes in R 2* values from precontrast to postcontrast were correlated with relative liver enhancement (RLE) based on signal intensities on T 1-weighted images using Spearman's rank correlation. PDFF values were similar between precontrast and postcontrast images (ICC=0.99, linear regression slopes=0.98, mean difference=-0.21 to -0.31%). PDFF measurements were stable between precontrast and postcontrast images. Changes in R 2* values were correlated with RLE (p<0.001, r=0.49-0.71). PDFF measurements from both image reconstruction methods are stable in the presence of changes in tissue relaxation rates after administering gadoxetic acid at 3T MRI. Changes in R 2* values correlate with established measures of gadoxetic acid uptake based on T 1-weighted images.
Read moreLumacaftor-ivacaftor effects on cystic fibrosis-related liver involvement in adolescents with homozygous F508 del-CFTR
Lumacaftor-ivacaftor effects on cystic fibrosis-related liver involvement in adolescents with homozygous F508 del-CFTR
High SNR Acquisitions Improve the Repeatability of Liver Fat Quantification Using Confounder-corrected Chemical Shift-encoded MR Imaging
PurposeTo determine whether high signal-to-noise ratio (SNR) acquisitions improve the repeatability of liver proton density fat fraction (PDFF) measurements using confounder-corrected chemical shift-encoded magnetic resonance (MR) imaging (CSE-MRI).Materials and MethodsEleven fat-water phantoms were scanned with 8 different protocols with varying SNR. After repositioning the phantoms, the same scans were repeated to evaluate the test-retest repeatability. Next, an in vivo study was performed with 20 volunteers and 28 patients scheduled for liver magnetic resonance imaging (MRI). Two CSE-MRI protocols with standard- and high-SNR were repeated to assess test-retest repeatability. MR spectroscopy (MRS)-based PDFF was acquired as a standard of reference. The standard deviation (SD) of the difference (Δ) of PDFF measured in the two repeated scans was defined to ascertain repeatability. The correlation between PDFF of CSE-MRI and MRS was calculated to assess accuracy. The SD of Δ and correlation coefficients of the two protocols (standard- and high-SNR) were compared using F-test and t-test, respectively. Two reconstruction algorithms (complex-based and magnitude-based) were used for both the phantom and in vivo experiments.ResultsThe phantom study demonstrated that higher SNR improved the repeatability for both complex- and magnitude-based reconstruction. Similarly, the in vivo study demonstrated that the repeatability of the high-SNR protocol (SD of Δ = 0.53 for complex- and = 0.85 for magnitude-based fit) was significantly higher than using the standard-SNR protocol (0.77 for complex, P < 0.001; and 0.94 for magnitude-based fit, P = 0.003). No significant difference was observed in the accuracy between standard- and high-SNR protocols.ConclusionHigher SNR improves the repeatability of fat quantification using confounder-corrected CSE-MRI.
Read moreThe relationship between liver triglyceride composition and proton density fat fraction as assessed by 1 H MRS.
The aim of this study was to estimate parameters determining liver triglyceride composition (TC) using 1 H MRS and to assess how TC estimability is affected by proton density fat fraction (PDFF) in adults with nonalcoholic fatty liver disease (NAFLD). In this prospective single-site study, 199 adults with known or suspected NAFLD in whom other causes of liver disease were excluded underwent two 1 H MRS STimulated Echo Acquisition Method (STEAM) sequences at 3 T. A respiratory-gated water-suppressed free breathing sequence (TE 10 ms, 16 signal averages) was used to assess TC in terms of the number of double bonds (ndb) and methylene-interrupted double bonds (nmidb), and a single breath-hold-long TR, multi-TE sequence (TR 3500 ms), which acquired five single average spectra over TE 10-30 ms, was used to estimate liver PDFF. Ndb and nmidb estimability was qualitatively assessed for each case and summarized descriptively. The consistency of ndb and nmidb estimation was examined using ROC analysis. The relationship between ndb and nmidb values and PDFF was presented graphically. Quality-of-fit of ndb and nmidb versus PDFF was evaluated by Pearson-r correlation. A significance level of 0.05 was used. In 263 1 H MRS examinations performed on 199 adult participants, ndb and nmidb were successfully estimated in 7/53 (13.2%) examinations with PDFF < 4%, 13/30 (43.3%) examinations with PDFF between 4% and 7%, 33/41 (80.5%) examinations with PDFF between 7% and 10%, and 124/139 (89.2%) examinations with PDFF > 10% (maximum PDFF 38.1%). Liver TC could be estimated consistently for PDFF > 6.7%. Both ndb and nmidb decreased with increasing PDFF (ndb = 2.83-0.0160·PDFF, r = -0.449, P < 0.0001); nmidb = 0.75-0.0088·PDFF, r = -0.350, P < 0.0001). In a cohort of adults with known or suspected NAFLD, liver TC becomes more saturated as PDFF increases.
Read moreLinearity and Bias of Proton Density Fat Fraction as a Quantitative Imaging Biomarker: A Multicenter, Multiplatform, Multivendor Phantom Study.
Background Proton density fat fraction (PDFF) estimated by using chemical shift-encoded (CSE) MRI is an accepted imaging biomarker of hepatic steatosis. This work aims to promote standardized use of CSE MRI to estimate PDFF. Purpose To assess the accuracy of CSE MRI methods for estimating PDFF by determining the linearity and range of bias observed in a phantom. Materials and Methods In this prospective study, a commercial phantom with 12 vials of known PDFF values were shipped across nine U.S. centers. The phantom underwent 160 independent MRI examinations on 27 1.5-T and 3.0-T systems from three vendors. Two three-dimensional CSE MRI protocols with minimal T1 bias were included: vendor and standardized. Each vendor's confounder-corrected complex or hybrid magnitude-complex based reconstruction algorithm was used to generate PDFF maps in both protocols. The Siemens reconstruction required a configuration change to correct for water-fat swaps in the phantom. The MRI PDFF values were compared with the known PDFF values by using linear regression with mixed-effects modeling. The 95% CIs were calculated for the regression slope (ie, proportional bias) and intercept (ie, constant bias) and compared with the null hypothesis (slope = 1, intercept = 0). Results Pooled regression slope for estimated PDFF values versus phantom-derived reference PDFF values was 0.97 (95% CI: 0.96, 0.98) in the biologically relevant 0%-47.5% PDFF range. The corresponding pooled intercept was -0.27% (95% CI: -0.50%, -0.05%). Across vendors, slope ranges were 0.86-1.02 (vendor protocols) and 0.97-1.0 (standardized protocol) at 1.5 T and 0.91-1.01 (vendor protocols) and 0.87-1.01 (standardized protocol) at 3.0 T. The intercept ranges (absolute PDFF percentage) were -0.65% to 0.18% (vendor protocols) and -0.69% to -0.17% (standardized protocol) at 1.5 T and -0.48% to 0.10% (vendor protocols) and -0.78% to -0.21% (standardized protocol) at 3.0 T. Conclusion Proton density fat fraction estimation derived from three-dimensional chemical shift-encoded MRI in a commercial phantom was accurate across vendors, imaging centers, and field strengths, with use of the vendors' product acquisition and reconstruction software. © RSNA, 2021 See also the editorial by Dyke in this issue.
Read moreHighly Reproducible, Vendor-Agnostic, Motion-Insensitive Liver PDFF Mapping at 0.55T, 1.5T, and 3T.
To develop and validate a vendor-agnostic, motion-insensitive proton-density fat-fraction (PDFF) quantification method. Flip-angle-modulated (FAM) 2D chemical-shift-encoded (CSE) MRI for PDFF quantification was implemented in both the vendor-agnostic platform Pulseq ("Pulseq-FAM") and one vendor-specific platform ("GE-specific FAM"). These implementations were distributed to four sites with twelve MR systems of three vendors (Siemens/GE/Philips) and field strengths (0.55T/1.5T/3T). A sequentially-shipped 16-vial phantom (PDFF = 0%-30%/T1water = 200-1400 ms) underwent confounder-corrected PDFF mapping with commercial 3D-CSE methods and GE-specific FAM as available on each system, and Pulseq-FAM on every system. To assess bias, phantom PDFF measurements were compared to reference. Between-system variance was evaluated with linear mixed-effects modeling. Different volunteers were also imaged at each site to assess free-breathing PDFF mapping feasibility. A prospective single-site volunteer study was also conducted. Adult patients and children were imaged with breath-held 3D-CSE and free-breathing GE-specific and Pulseq-FAM. Radiologists evaluated images for overall quality and motion artifacts. To assess bias, Pulseq-FAM PDFF measurements were compared to 3D-CSE and GE-specific FAM. Test-retest repeatability was assessed by re-imaging after repositioning. Between-field-strength reproducibility was assessed at 1.5T and 3.0T. In the multi-center study, Pulseq-FAM showed reduced T1-bias and between-system variability versus 3D-CSE in phantom PDFF measurements, and free-breathing feasibility in volunteers. In the single-site volunteer study (N = 57), Pulseq-FAM improved image quality and motion artifacts versus 3D-CSE (p < 0.01). Pulseq-FAM showed excellent agreement with 3D-CSE (95% limits-of-agreement (LoA) = 3.4% PDFF) and GE-specific FAM (LoA = 2.0%). Pulseq-FAM showed excellent repeatability (repeatability coefficient (RC) = 1.6% PDFF) and between-field-strength reproducibility (reproducibility coefficient (RDC) = 2.4%) versus 3D-CSE (RC = 2.7%/RDC = 3.4%; differences p < 0.05). Pulseq-FAM enables accurate, reproducible, vendor-agnostic, and motion-insensitive PDFF quantification in adults and children.
Read moreValue of two-point Dixon water-only Look-Locker T1 mapping on the assessment of liver fibrosis in chronic liver disease with hepatic steatosis.
To evaluate the influence of fat on the assessment of liver fibrosis in chronic liver disease using two-point Dixon water-only Look-Locker T1 mapping by comparing water-only derived sequence (W-Dixon) with in-phase (IP) and opposed-phase (OP)-based sequences. 2.89-T MRI included 2D two-point Dixon Look-Locker T1 mapping and proton density fat fraction (PDFF) mapping. The correlations between liver T1 values and PDFF were assessed using Spearman correlation coefficient. T1 values on each T1 map were compared among three FIB-4 index range groups (FIB-4 < 1.3, 1.3-2.67, > 2.67) in patients with and without hepatic steatosis using one-way analysis of variance and Kruskal-Wallis test. 204 patients with chronic liver disease were retrospectively evaluated. T1 values on IP or OP images were significantly correlated with PDFF (r = -0.373, 0.220), while no significant correlation was found between T1 values on W-Dixon images and PDFF (r = -0.071). In patients without hepatic steatosis, T1 values on each T1 map in FIB-4 > 2.67 group were significantly higher than in FIB-4 1.3-2.67 group (p < 0.01). Conversely, in patients with hepatic steatosis, only W-Dixon sequence statistically differentiated FIB-4 > 2.67 group from FIB-4 1.3-2.67 group based on T1 values (p < 0.05). The assessment of liver fibrosis based on T1 values obtained by Dixon water-only T1 mapping was less influenced by the presence of fat. Two-point Dixon water-only Look-Locker T1 mapping minimizes the confounding effect of fat, enabling proper assessment of liver fibrosis in steatotic chronic liver disease.
Read moreValidation of a standardized MRI method for liver fat and T2* quantification
PurposeSeveral studies have demonstrated the accuracy, precision, and reproducibility of proton density fat fraction (PDFF) quantification using vendor-specific image acquisition protocols and PDFF estimation methods. The purpose of this work is to validate a confounder-corrected, cross-vendor, cross field-strength, in-house variant LMS IDEAL of the IDEAL method licensed from the University of Wisconsin, which has been developed for routine clinical use.MethodsLMS IDEAL is implemented using a combination of patented and/or published acquisition and some novel model fitting methods required to correct confounds which result from the imaging and estimation processes, including: water-fat ambiguity; T2* relaxation; multi-peak fat modelling; main field inhomogeneity; T1 and noise bias; bipolar readout gradients; and eddy currents. LMS IDEAL has been designed to use image acquisition protocols that can be installed on most MRI scanners and cloud-based image processing to provide fast, standardized clinical results. Publicly available phantom data were used to validate LMS IDEAL PDFF calculations against results from originally published IDEAL methodology. LMS PDFF and T2* measurements were also compared with an independent technique in human volunteer data (n = 179) acquired as part of the UK Biobank study.ResultsWe demonstrate excellent agreement of LMS IDEAL across vendors, field strengths, and over a wide range of PDFF and T2* values in the phantom study. The performance of LMS IDEAL was then assessed in vivo against widely accepted PDFF and T2* estimation methods (LMS Dixon and LMS T2*, respectively), demonstrating the robustness of LMS IDEAL to potential sources of error.ConclusionThe development and clinical validation of the LMS IDEAL algorithm as a chemical shift-encoded MRI method for PDFF and T2* estimation contributes towards robust, unbiased applications for quantification of hepatic steatosis and iron overload, which are key features of chronic liver disease.
Read moreMagnitude-intrinsic water-fat ambiguity can be resolved with multipeak fat modeling and a multipoint search method.
PurposeTo develop a postprocessing algorithm for multiecho chemical‐shift encoded water–fat separation that estimates proton density fat fraction (PDFF) maps over the full dynamic range (0‐100%) using multipeak fat modeling and multipoint search optimization. To assess its accuracy, reproducibility, and agreement with state‐of‐the‐art complex‐based methods, and to evaluate its robustness to artefacts in abdominal PDFF maps.MethodsWe introduce MAGO (MAGnitude‐Only), a magnitude‐based reconstruction that embodies multipeak liver fat spectral modeling and multipoint optimization, and which is compatible with asymmetric echo acquisitions. MAGO is assessed first for accuracy and reproducibility on publicly available phantom data. Then, MAGO is applied to N = 178 UK Biobank cases, in which its liver PDFF measures are compared using Bland‐Altman analysis with those from a version of the hybrid iterative decomposition of water and fat with echo asymmetry and least squares estimation (IDEAL) algorithm, LiverMultiScan IDEAL (LMS IDEAL, Perspectum Diagnostics Ltd, Oxford, UK). Finally, MAGO is tested on a succession of high field challenging cases for which LMS IDEAL generated artefacts in the PDFF maps.ResultsPhantom data showed accurate, reproducible MAGO PDFF values across manufacturers, field strengths, and acquisition protocols. Moreover, we report excellent agreement between MAGO and LMS IDEAL for 6‐echo, 1.5 tesla human acquisitions (bias = −0.02% PDFF, 95% confidence interval = ±0.13% PDFF). When tested on 12‐echo, 3 tesla cases from different manufacturers, MAGO was shown to be more robust to artefacts compared to LMS IDEAL.ConclusionMAGO resolves the water–fat ambiguity over the entire fat fraction dynamic range without compromising accuracy, therefore enabling robust PDFF estimation where phase data is inaccessible or unreliable and complex‐based and hybrid methods fail.
Read moreHbA1c and Liver Fat After 16 Weeks of Fasted versus Fed Exercise Training in Adults With Type 2 Diabetes.
Exercise-nutrient timing is of interest for people with type 2 diabetes (T2D) as a potential method to optimize glycemic control. However, the optimal nutritional environment for exercise is not well understood over the long term. The Fasted Exercise for Type 2 Diabetes (FED) Trial compared 16 wk of fasted versus postprandial morning exercise on glycated hemoglobin (HbA1c) and liver proton density fat fraction (PDFF). Twenty adults with T2D were recruited and randomized to complete exercise after an overnight fast versus after their morning meal. Participants walked three mornings per week, progressing to 180 min·wk -1 over 16 wk. Groups were balanced with five males and five females each. Sixteen participants completed the trial (8 in each group, 50% female). Age, HbA1c, and PDFF were 59.8 ± 9.0 yr, 7.2 ± 0.7%, and 9.3 ± 4.1%, respectively. On average, both groups completed 98% of their walking sessions but there was no change in HbA1c (-0.2%, P = 0.588). However, one participant from each group had changes in their glucose-lowering medication during the trial, and when excluded, the fasted training group had greater improvements in HbA1c compared with the postprandial group (-0.3% vs 0.0%, P = 0.033). There was no difference in changes in liver PDFF between groups (-1.6% vs 0.3%, P = 0.221) but visceral fat and intramuscular fat decreased to a greater extent after fasted exercise. Although our study had a small sample size, it suggests that exercise after an overnight fast can have high adherence and represents an option for people with T2D to improve longer-term indicators of glycemia and ectopic fat depots.
Read moreComparison of Vendor-Independent Software Tools for Liver Proton Density Fat Fraction Estimation at 1.5 T.
(1) Background: Open-source software tools are available to estimate proton density fat fraction (PDFF). (2) Methods: We compared four algorithms: complex-based with graph cut (GC), magnitude-based (MAG), magnitude-only estimation with Rician noise modeling (MAG-R), and multi-scale quadratic pseudo-Boolean optimization with graph cut (QPBO). The accuracy and reliability of the methods were evaluated in phantoms with known fat/water ratios and a patient cohort with various grades (S0-S3) of steatosis. Image acquisitions were performed at 1.5 Tesla (T). (3) Results: The PDFF estimates showed a nearly perfect correlation (Pearson r = 0.999, p < 0.001) and inter-rater agreement (ICC = from 0.995 to 0.999, p < 0.001) with true fat fractions. The absolute bias was low with all methods (0.001-1%), and an ANCOVA detected no significant difference between the algorithms in vitro. The agreement across the methods was very good in the patient cohort (ICC = 0.891, p < 0.001). However, MAG estimates (-2.30% ± 6.11%, p = 0.005) were lower than MAG-R. The field inhomogeneity artifacts were most frequent in MAG-R (70%) and GC (39%) and absent in QPBO images. (4) Conclusions: The tested algorithms all accurately estimate PDFF in vitro. Meanwhile, QPBO is the least affected by field inhomogeneity artifacts in vivo.
Read moreAgreement between region-of-interest- and parametric map-based hepatic proton density fat fraction estimation in adults with chronic liver disease.
To compare agreement between region-of-interest (ROI)- and parametric map-based methods of hepatic proton density fat fraction (PDFF) estimation in adults with known or suspected hepatic steatosis secondary to chronic liver disease over a range of imaging and analysis conditions. In this IRB approved HIPAA compliant prospective single-site study, 31 adults with chronic liver disease undergoing clinical gadoxetic acid-enhanced liver magnetic resonance imaging at 3T were recruited. Multi-echo gradient-echo imaging at flip angles of 10° and 50° was performed before and after administration of gadoxetic acid. Six echoes were acquired at successive nominally out-of-phase and in-phase echo times. PDFF was estimated with a nonlinear fitting algorithm using the first two, three, four, five, and (all) six echoes. Hence, 20 different imaging and analysis conditions were used (pre/post contrast x low/high flip angle x 2/3/4/5/6 echoes). For each condition, PDFF estimation was done in corresponding liver locations using two methods: a region-of-interest (ROI)-based method in which mean signal intensity values within ROIs were run through the fitting algorithm, and a parametric map-based method in which individual signal intensities were run through the fitting algorithm pixel by pixel. Agreement between ROI- and map-based PDFF estimation was assessed by Bland-Altman and intraclass correlation (ICC) analysis. Depending on the condition and method, PDFF ranged from -2.52% to 45.57%. Over all conditions, mean differences between ROI- and map-based PDFF estimates ranged from 0.04% to 0.24%, with all ICCs ≥0.999. Agreement between ROI- and parametric map-based PDFF estimation is excellent over a wide range of imaging and analysis conditions.
Read moreDeep Learning Reveals Liver MRI Features Associated With PNPLA3 I148M in Steatotic Liver Disease
ABSTRACTBackgroundSteatotic liver disease (SLD) is the most common liver disease worldwide, affecting 30% of the global population. It is strongly associated with the interplay of genetic and lifestyle‐related risk factors. The genetic variant accounting for the largest fraction of SLD heritability is PNPLA3 I148M, which is carried by 23% of the western population and increases the risk of SLD two to three‐fold. However, identification of variant carriers is not part of routine clinical care and prevents patients from receiving personalised care.MethodsWe analysed MRI images and common genetic variants in PNPLA3, TM6SF2, MTARC1, HSD17B13 and GCKR from a cohort of 45 603 individuals from the UK Biobank. Proton density fat fraction (PDFF) maps were generated using a water‐fat separation toolbox, applied to the magnitude and phase MRI data. The liver region was segmented using a U‐Net model trained on 600 manually segmented ground truth images. The resulting liver masks and PDFF maps were subsequently used to calculate liver PDFF values. Individuals with (PDFF ≥ 5%) and without SLD (PDFF < 5%) were selected as the study cohort and used to train and test a Vision Transformer classification model with five‐fold cross validation. We aimed to differentiate individuals who are homozygous for the PNPLA3 I148M variant from non‐carriers, as evaluated by the area under the receiver operating characteristic curve (AUROC). To ensure a clear genetic contrast, all heterozygous individuals were excluded. To interpret our model, we generated attention maps that highlight the regions that are most predictive of the outcomes.ResultsHomozygosity for the PNPLA3 I148M variant demonstrated the best predictive performance among five variants with AUROC of 0.68 (95% CI: 0.64–0.73) in SLD patients and 0.57 (95% CI: 0.52–0.61) in non‐SLD patients. The AUROCs for the other SNPs ranged from 0.54 to 0.57 in SLD patients and from 0.52 to 0.54 in non‐SLD patients. The predictive performance was generally higher in SLD patients compared to non‐SLD patients. Attention maps for PNPLA3 I148M carriers showed that fat deposition in regions adjacent to the hepatic vessels, near the liver hilum, plays an important role in predicting the presence of the I148M variant.ConclusionOur study marks novel progress in the non‐invasive detection of homozygosity for PNPLA3 I148M through the application of deep learning models on MRI images. Our findings suggest that PNPLA3 I148M might affect the liver fat distribution and could be used to predict the presence of PNPLA3 variants in patients with fatty liver. The findings of this research have the potential to be integrated into standard clinical practice, particularly when combined with clinical and biochemical data from other modalities to increase accuracy, enabling easier identification of at‐risk individuals and facilitating the development of tailored interventions for PNPLA3 I148M‐associated liver disease.
Read more