Liver PDFF estimation using a multi-decoder water-fat separation neural network with a reduced number of echoes.

OBJECTIVE: To accurately estimate liver PDFF from chemical shift-encoded (CSE) MRI using a deep learning (DL)-based Multi-Decoder Water-Fat separation Network (MDWF-Net), that operates over complex-valued CSE-MR images with only 3 echoes. METHODS: The proposed MDWF-Net and a U-Net model were independently trained using the first 3 echoes of MRI data from 134 subjects, acquired with conventional 6-echoes abdomen protocol at 1.5 T. Resulting models were then evaluated using unseen CSE-MR images obtained from 14 subjects that were acquired with a 3-echoes CSE-MR pulse sequence with a shorter duration compared to the standard protocol. Resulting PDFF maps were qualitatively assessed by two radiologists, and quantitatively assessed at two corresponding liver ROIs, using Bland Altman and regression analysis for mean values, and ANOVA testing for standard deviation (STD) (significance level: .05). A 6-echo graph cut was considered ground truth. RESULTS: Assessment of radiologists demonstrated that, unlike U-Net, MDWF-Net had a similar quality to the ground truth, despite it considered half of the information. Regarding PDFF mean values at ROIs, MDWF-Net showed a better agreement with ground truth (regression slope = 0.94, R2 = 0.97) than U-Net (regression slope = 0.86, R2 = 0.93). Moreover, ANOVA post hoc analysis of STDs showed a statistical difference between graph cuts and U-Net (p < .05), unlike MDWF-Net (p = .53). CONCLUSION: MDWF-Net showed a liver PDFF accuracy comparable to the reference graph cut method, using only 3 echoes and thus allowing a reduction in the acquisition times. CLINICAL RELEVANCE STATEMENT: We have prospectively validated that the use of a multi-decoder convolutional neural network to estimate liver proton density fat fraction allows a significant reduction in MR scan time by reducing the number of echoes required by 50%. KEY POINTS: • Novel water-fat separation neural network allows for liver PDFF estimation by using multi-echo MR images with a reduced number of echoes. • Prospective single-center validation demonstrated that echo reduction leads to a significant shortening of the scan time, compared to standard 6-echo acquisition. • Qualitative and quantitative performance of the proposed method showed no significant differences in PDFF estimation with respect to the reference technique.

Toward a realistic in silico abdominal phantom for QSM.

PURPOSE: QSM outside the brain has recently gained interest, particularly in the abdominal region. However, the absence of reliable ground truths makes difficult to assess reconstruction algorithms, whose quality is already compromised by additional signal contributions from fat, gases, and different kinds of motion. This work presents a realistic in silico phantom for the development, evaluation and comparison of abdominal QSM reconstruction algorithms. METHODS: Synthetic susceptibility and R 2 * $$ {R}_2^{\ast } $$ maps were generated by segmenting and postprocessing the abdominal 3T MRI data from a healthy volunteer. Susceptibility and R 2 * $$ {R}_2^{\ast } $$ values in different tissues/organs were assigned according to literature and experimental values and were also provided with realistic textures. The signal was simulated using as input the synthetic QSM and R 2 * $$ {R}_2^{\ast } $$ maps and fat contributions. Three susceptibility scenarios and two acquisition protocols were simulated to compare different reconstruction algorithms. RESULTS: QSM reconstructions show that the phantom allows to identify the main strengths and limitations of the acquisition approaches and reconstruction algorithms, such as in-phase acquisitions, water-fat separation methods, and QSM dipole inversion algorithms. CONCLUSION: The phantom showed its potential as a ground truth to evaluate and compare reconstruction pipelines and algorithms. The publicly available source code, designed in a modular framework, allows users to easily modify the susceptibility, R 2 * $$ {R}_2^{\ast } $$ and TEs, and thus creates different abdominal scenarios.

Multi-parametric liver tissue characterization using MR fingerprinting: Simultaneous T1 , T2 , T2 *, and fat fraction mapping.

PURPOSE: Quantitative T1 , T2 , T2 *, and fat fraction (FF) maps are promising imaging biomarkers for the assessment of liver disease, however these are usually acquired in sequential scans. Here we propose an extended MR fingerprinting (MRF) framework enabling simultaneous liver T1 , T2 , T2 *, and FF mapping from a single ~14 s breath-hold scan. METHODS: A gradient echo (GRE) liver MRF sequence with nine readouts per TR, low flip angles (5-15°), varying magnetisation preparation and golden angle radial trajectory is acquired at 1.5T to encode T1 , T2 , T2 *, and FF simultaneously. The nine-echo time-series are reconstructed using a low-rank tensor constrained reconstruction and used to fit T2 *, B0 and to separate the water and fat signals. Water- and fat-specific T1 , T2, and M0 are obtained through dictionary matching, whereas FF estimation is extracted from the M0 maps. The framework was evaluated in a standardized T1 /T2 phantom, a water-fat phantom, and 12 subjects in comparison to reference methods. Preliminary clinical feasibility is shown in four patients. RESULTS: The proposed water T1 , water T2 , T2 *, and FF maps in phantoms showed high coefficients of determination (r2 > 0.97) relative to reference methods. Measured liver MRF values in vivo (mean ± SD) for T1 , T2 , T2 *, and FF were 671 ± 60 ms, 43.2 ± 6.8 ms, 29 ± 6.6 ms, and 3.2 ± 2.6% with biases of 92 ms, -7.1 ms, -1.4 ms, and 0.63% when compared to conventional methods. CONCLUSION: A nine-echo liver MRF sequence allows for quantitative multi-parametric liver tissue characterization in a single breath-hold scan of ~14 s. Future work will aim to validate the proposed approach in patients with liver disease.

[Quantification of liver fat infiltration by magnetic resonance]. / Cuantificación de esteatosis hepática no alcohólica por resonancia magnética.

BACKGROUND: A simple and inexpensive method is required to assess fatty infiltration of the liver non-invasively. AIM: To develop and compare different methods to quantify liver fat by magnetic resonance and compare it against ultrasound. MATERIAL AND METHODS: Three algorithms were implemented: region growing (RG), graph cuts (GC) and hierarchical (HR), all based on the IDEAL method to obtain water and fat images. Using these images, the proton density fat fraction (PDFF) was calculated. The three methods were tested in phantoms with known fat percentages and later on we acquired images from 20 volunteers with an ultrasound diagnosis of fatty liver disease in different stages. For everyone, the PDFF of the nine liver segments was determined. RESULTS: In phantoms, the mean error between the real fat percentage and the value obtained through the three methods was -1,26, -1 and -0,8 for RG, GC and HR, respectively. The hierarchical method was more precise and efficient to obtain PDFF. The results in volunteers revealed that ultrasound showed errors categorizing the severity of hepatic steatosis in more than 50% of volunteers. CONCLUSIONS: We developed a tool for magnetic resonance, which allows to quantify fat in the liver. This method is less operator dependent than ultrasound and describes the heterogeneity in the fat distribution along the nine hepatic segments.

Cuantificación de esteatosis hepática no alcohólica por resonancia magnética / Quantification of liver fat infiltration by magnetic resonance

Background: A simple and inexpensive method is required to assess fatty infiltration of the liver non-invasively. Aim: To develop and compare different methods to quantify liver fat by magnetic resonance and compare it against ultrasound. Material and Methods: Three algorithms were implemented: region growing (RG), graph cuts (GC) and hierarchical (HR), all based on the IDEAL method to obtain water and fat images. Using these images, the proton density fat fraction (PDFF) was calculated. The three methods were tested in phantoms with known fat percentages and later on we acquired images from 20 volunteers with an ultrasound diagnosis of fatty liver disease in different stages. For everyone, the PDFF of the nine liver segments was determined. Results: In phantoms, the mean error between the real fat percentage and the value obtained through the three methods was −1,26, −1 and −0,8 for RG, GC and HR, respectively. The hierarchical method was more precise and efficient to obtain PDFF. The results in volunteers revealed that ultrasound showed errors categorizing the severity of hepatic steatosis in more than 50% of volunteers. Conclusions: We developed a tool for magnetic resonance, which allows to quantify fat in the liver. This method is less operator dependent than ultrasound and describes the heterogeneity in the fat distribution along the nine hepatic segments.

Is a single-level measurement of paraspinal muscle fat infiltration and cross-sectional area representative of the entire lumbar spine?

PURPOSE: Lumbar paraspinal muscle morphology has recently been evaluated in several studies with conflicting results. Several studies have performed single-slice evaluations of paraspinal muscle morphology, whereas other studies have done a multi-level assessment; this methodological difference might explain the observed different results. Our study evaluated if a single-slice axial measurement is representative of the entire lumbar musculature. METHODS: We included 80 adult patients who were consecutively evaluated with magnetic resonance imaging (MRI) for spinal symptoms. Using T2-weighted axial images, we measured the fat signal fractions (FSF) and cross-sectional area (CSA) of the erector spinae and multifidus at the five levels of the lumbar spine (from L1-L2 to L5-S1). We used the ANOVA test for repeated measurements (with Bonferroni correction) to compare the FSF and CSA among the levels. RESULTS: Erector spinae showed an increasing FSF from L1-L2 to L5-S1; all erector spinae FSF comparisons among the different levels were significantly different. Multifidus FSF also increased caudally below L2-L3, although significant differences were observed only with two or more levels of distance. The CSA of the erector spinae showed a caudal decrease (L4-L5 and L5-S1 being significantly smaller than all the levels above). The CSA of the multifidus showed that all levels exhibited a significantly different area compared to their adjacent level (except L5-S1 compared to L4-L5). CONCLUSIONS: No single-level FSF or CSA is representative of the whole lumbar spine. A standardized multi-level evaluation of the paraspinal musculature should be used in future research.

Lumbar paraspinal muscle fat infiltration is independently associated with sex, age, and inter-vertebral disc degeneration in symptomatic patients.

PURPOSE: To determine the association of paraspinal muscles and psoas relative cross-sectional area (RCSA) and fat signal fraction (FSF) with sex, age, and intervertebral disc degeneration (IDD) in symptomatic patients. METHODS: We retrospectively evaluated 80 adult patients with spinal symptoms using T2-weighted magnetic resonance images. We determined RCSA and FSF of the paraspinal muscles (erector spinae and multifidus) and psoas from L1-L2 to L5-S1; we determined IDD using the Pfirrmann classification. We compared differences in muscle RCSA and FSF based on sex and IDD, and we correlated age and IDD with RCSA and FSF. Using multivariate linear regression analyses, we determined the impact of sex, age, and IDD on RCSA and FSF. RESULTS: Men exhibited larger psoas RCSA but not larger paraspinal muscles RCSA than women. Women had larger FSF in the paraspinal muscles and psoas. Increasing IDD was associated with larger FSF if ≥2 Pfirrmann grades were observed. IDD correlated with FSF of the paraspinal muscles, and age correlated with FSF of the paraspinal muscles and psoas. IDD was less consistently correlated with RCSA, but age correlated negatively with RCSA of all three muscles. Linear regression analyses demonstrated that sex, age, and IDD were each independently associated with FSF of the paraspinal muscles; additionally, sex and age, but not IDD, were associated with psoas FSF. RCSA was less consistently influenced by these three variables. CONCLUSIONS: Sex, age, and IDD are independently associated with paraspinal muscles FSF; only sex and age influence psoas FSF.

Sensitivity analysis of geometric errors in additive manufacturing medical models.

Additive manufacturing (AM) models are used in medical applications for surgical planning, prosthesis design and teaching. For these applications, the accuracy of the AM models is essential. Unfortunately, this accuracy is compromised due to errors introduced by each of the building steps: image acquisition, segmentation, triangulation, printing and infiltration. However, the contribution of each step to the final error remains unclear. We performed a sensitivity analysis comparing errors obtained from a reference with those obtained modifying parameters of each building step. Our analysis considered global indexes to evaluate the overall error, and local indexes to show how this error is distributed along the surface of the AM models. Our results show that the standard building process tends to overestimate the AM models, i.e. models are larger than the original structures. They also show that the triangulation resolution and the segmentation threshold are critical factors, and that the errors are concentrated at regions with high curvatures. Errors could be reduced choosing better triangulation and printing resolutions, but there is an important need for modifying some of the standard building processes, particularly the segmentation algorithms.