Reduction of respiratory motion artifacts in transverse relaxation rate (R2) images of the liver

A short-TR single-echo spin-echo breath-hold method for assessing liver T2

OBJECTIVE

Conventional single-echo spin-echo T2 mapping used for liver iron quantification is too long for breath-holding. This study investigated a short TR (~100 ms) single-echo spin-echo T2 mapping technique wherein each image (corresponding to a single TE) could be acquired in ~17 s-short enough for a breath-hold. TE images were combined for T2 fitting. To avoid T1 bias, each TE acquisition incremented TR to maintain a constant TR-TE.

MATERIALS AND METHODS

Experiments at 1.5T validated the technique's accuracy in phantoms, 9 healthy volunteers, and 5 iron overload patients. In phantoms and healthy volunteers, the technique was compared to the conventional approach of constant TR for all TEs. Iron overload results were compared to FerriScan.

RESULTS

In phantoms, the constant TR-TE technique provided unbiased estimates of T2, while the conventional constant TR approach underestimated it. In healthy volunteers, there was no significant discrepancy at the 95% confidence level between constant TR-TE and reference T2 values, whereas there was for constant TR scans. In iron overload patients, there was a high correlation between constant TR-TE and FerriScan T2 values (r2 = 0.95), with a discrepancy of 0.6+/- 1.4 ms.

DISCUSSION

The short-TR single-echo breath-hold spin-echo technique provided unbiased estimates of T2 in phantoms and livers.

Automatic MR image quality evaluation using a Deep CNN: A reference-free method to rate motion artifacts in neuroimaging

Motion artifacts on magnetic resonance (MR) images degrade image quality and thus negatively affect clinical and research scanning. Considering the difficulty in preventing patient motion during MR examinations, the identification of motion artifact has attracted significant attention from researchers. We propose an automatic method for the evaluation of motion corrupted images using a deep convolutional neural network (CNN). Deep CNNs has been used widely in image classification tasks. While such methods require a significant amount of annotated training data, a scarce resource in medical imaging, the transfer learning and fine-tuning approaches allow us to use a smaller amount of data. Here we selected four renowned architectures, initially trained on Imagenet contest dataset, to fine-tune. The models were fine-tuned using patches from an annotated dataset composed of 68 T1-weighted volumetric acquisitions from healthy volunteers. For training and validation 48 images were used, while the remaining 20 images were used for testing. Each architecture was fine-tuned for each MR axis, detecting the motion artifact per patches from the three orthogonal MR acquisition axes. The overall average accuracy for the twelve models (three axes for each of four architecture) was 86.3%. As our goal was to detect fine-grained corruption in the image, we performed an extensive search on lower layers from each of the four architectures, since they filter small regions in the original input. Experiments showed that architectures with fewer layers than the original ones reported the better results for image patches with an overall average accuracy of 90.4%. The accuracies per architecture were similar so we decided to explore all four architectures performing a result consensus. Also, to determine the probability of motion artifacts presence on the whole acquisition a combination of the three axes were performed. The final architecture consists of an artificial neural network (ANN) classifier combining all models from the four shallower architectures, which overall acquisition-based accuracy was 100.0%. The proposed method generalization was tested using three different MR data: (1) MR image acquired in epilepsy patients (93 acquisitions); (2) MR image presenting susceptibility artifact (22 acquisitions); and (3) MR image acquired from different scanner vendor (20 acquisitions). The achieved acquisition-based accuracy on generalization tests (1) 90.3%, (2) 63.6%, and (3) 75.0%) suggests that domain adaptation is necessary. Our proposed method can be rapidly applied to large amounts of image data, providing a motion probability p∈[0,1] per acquisition. This method output can be used as a scale to identify the motion corrupted images from the dataset, thus minimizing the time spent on visual quality control.

Inter-method reproducibility of biexponential R2 MR relaxometry for estimation of liver iron concentration

PURPOSE

To assess the reproducibility of biexponential R₂ -relaxometry MRI for estimation of liver iron concentration (LIC) between proprietary and nonproprietary analysis methods.

METHODS

This single-center retrospective study, approved by investigational review board and compliant with the Health Insurance Portability and Accountability Act, included 40 liver MRI exams in 38 subjects with suspected or known iron overload. From spin-echo images of the liver, acquired at 5 different echo times (TE = 6-18 ms), biexponential R₂ maps were calculated using 1 proprietary (FerriScan, Resonance Health Ltd., Claremont WA, Australia) and 3 nonproprietary (simulated annealing, nonlinear least squares, dictionary search) analysis methods. Each subject's average liver R₂ value was converted to LIC using a previously validated calibration curve. Inter-method reproducibility for liver R₂ and LIC were assessed for linearity using linear regression analysis and absolute agreement using intraclass correlation and Bland-Altman analysis. For point estimates, 95% confidence intervals were calculated; P values < 0.05 were considered statistically significant.

RESULTS

Linearity between the proprietary and nonproprietary methods was excellent across the observed range for R₂ (20-312 s^-1 ) and LIC (0.4-52.2 mg/g), with all coefficients of determination (R² ) ≥ 0.95. No statistically significant bias was found (slope estimates ∼ 1; intercept estimates ∼ 0; P values > 0.05). Agreement between the 4 methods was excellent for both liver R₂ and LIC (intraclass correlations ≥ 0.97). Bland-Altman 95% limits of agreement in % difference between the proprietary and nonproprietary methods were ≤ 9% and ≤ 16% for R₂ and LIC, respectively.

CONCLUSION

Biexponential R₂ -relaxometry MRI for LIC estimation is reproducible between proprietary and nonproprietary analysis methods.

Respiratory self-gated multiple gradient recalled echo sequence for free-breathing abdominal R2* mapping

Abdominal effective transverse relaxation rate (R(2)*) mapping is critical for a wide range of applications. However, respiratory motion can lead to significant image quality deterioration and R(2)* overestimation. For this work, we explored the feasibility of combining respiratory self-gating techniques with a multiple gradient-recalled echo sequence for free-breathing abdominal R(2)* measurements. In a series of eight normal volunteers, respiratory self-gated-multiple gradient-recalled echo methods effectively avoided motion artifacts to produce quantitative R(2)* measurements in liver, spleen, and kidneys that were comparable to R(2)* measurements produced while breath-holding. Respiratory self-gated-multiple gradient-recalled echo methods demonstrated the potential to avoid the need for breath-holding during abdominal R(2)* mapping. For clinical application, respiratory self-gated-multiple gradient-recalled echo approaches could be particularly useful for R(2)* measurements in those patients unable or unwilling to sustain sufficiently long breath-holds to avoid motion artifacts.

The effect of reducing repetition time TR on the measurement of liver R2 for the purpose of measuring liver iron concentration

The effects of reducing the pulse repetition time from 2500 ms to 1000 ms when using spin-density-projection-assisted R2-magnetic resonance imaging for the purpose of measuring liver iron concentration were evaluated. Repeated liver R2 measurements were made using both protocols on 60 subjects with liver iron concentrations ranging from 0.5 to 48.6 mg Fe (g dry tissue)(-1). The mean total scan time at repetition time 1000 ms was 42% of that at repetition time 2500 ms. The repeatability coefficients for the two protocols were not significantly different from each other. A systematic difference in the measured R2 using each protocol was found indicating that an adjustment factor is required when one protocol is used to replace the other. The 95% limits of agreement between the two protocols were not significantly different from their repeatability coefficients indicating that the protocols can be interchanged without any significant change in accuracy or precision of liver iron concentration measurement.

Improved transverse relaxation rate measurement techniques for the assessment of hepatic and myocardial iron content

PURPOSE

To develop and validate an optimized respiratory-gated, gradient-echo sampling of free induction decay and echo (GESFIDE) pulse sequence for the simultaneous measurement of R2, R2*, and R2' in the liver or heart.

MATERIALS AND METHODS

Fifteen subjects (12 thalassemia patients and three normal volunteers) were scanned using an optimized navigator-gated GESFIDE pulse sequence for the measurement of R2, R2*, and R2' in the liver and heart. For imaging the myocardium, dark-blood preparation was used to suppress the blood signal to improve accuracy. The results were compared with those obtained from breath-held GESFIDE and multi-gradient-echo (GRE) scans.

RESULTS

Good agreement between breath-held and navigator-gated scans was found for R2, R2*, and R2' values in the liver (slopes = 0.97-0.99, r = 0.997-0.998, P < 0.0001) and for R2* in the heart (slope = 1.02, r = 0.85, P < 0.0001). Both R2* and R2' were closely correlated to R2 in the liver, with correlation factors of 0.998 and 0.994, respectively, but weaker correlations were observed in the heart (r = 0.72 for R2* vs. R2 and r = 0.51 for R2' vs. R2).

CONCLUSION

The improved sequence enables free-breathing measurements of transverse relaxation rates of the myocardium and liver. The method precludes the need for multiple breath-held scans and possible misregistration issues, and may prove most beneficial for imaging young children and patients who may have difficulty with prolonged or repeated breath-holds.

Magnetic resonance imaging measurement of iron overload

PURPOSE OF REVIEW

To highlight recent advances in magnetic resonance imaging estimation of somatic iron overload. This review will discuss the need and principles of magnetic resonance imaging-based iron measurements, the validation of liver and cardiac iron measurements, and the key institutional requirements for implementation.

RECENT FINDINGS

Magnetic resonance imaging assessment of liver and cardiac iron has achieved critical levels of availability, utility, and validity to serve as the primary endpoint of clinical trials. Calibration curves for the magnetic resonance imaging parameters R2 and R2* (or their reciprocals, T2 and T2*) have been developed for the liver and the heart. Interscanner variability for these techniques has proven to be on the order of 5-7%.

SUMMARY

Magnetic resonance imaging assessment of tissue iron is becoming increasingly important in the management of transfusional iron load because it is noninvasive, relatively widely available and offers a window into presymptomatic organ dysfunction. The techniques are highly reproducible within and across machines and have been chemically validated in the liver and the heart. These techniques will become the standard of care as industry begins to support the acquisition and postprocessing software.

Measurement and Mapping of Liver Iron Concentrations Using Magnetic Resonance Imaging

Measurement of liver iron concentration (LIC) is an important clinical procedure in the management of transfusional iron overload with iron chelation. LIC gives an indication of over- or underchelation. Although chemical assay of needle biopsy samples from the liver has been considered the "gold standard" of LIC measurement, needle biopsy sampling errors can be surprisingly large owing to the natural spatial variation of LIC throughout the liver and the small size of biopsy specimens. A magnetic resonance imaging technique has now been developed that enables safe noninvasive measurement and imaging of LIC with a known accuracy and precision. Measurements of LIC can be made over the range of LIC encountered in clinical practice. The technique is based on the measurement and imaging of proton transverse relaxation rates (R2) within the liver. The R2 imaging technique can be implemented on most clinical 1.5-T MRI instruments, making it readily available to the clinical community.

Liver iron concentration evaluated by two magnetic methods: magnetic resonance imaging and magnetic susceptometry

Quantification of liver iron concentration (LIC) is crucial in the management of patients suffering from certain pathologies that can produce iron overload, such as Cooley's anemia and hemochromatosis. All of these patients must control the level of iron deposits in their organs to avoid the toxicity of high LIC, which is potentially lethal. This paper describes experimental protocols for LIC measurement using two magnetic techniques: magnetic resonance imaging (MRI) and biomagnetic liver susceptometry (BLS). MRI proton transverse relaxation rate (R2) and image intensity, evaluated pixel by pixel, were used as indicators of iron load in the tissue. LIC measurement by BLS was performed using an AC superconducting susceptometer system. A group of 23 patients with a large range of iron overload (0.9 to 34.5 mgFe/g(dry tissue)) was evaluated with both techniques (MRI x BLS). A significant linear correlation (r = 0.89-0.95) was found between the LIC by MRI and by BLS. These results show the feasibility of using two noninvasive methodologies to evaluate liver iron store in a large concentration range. Both methodologies represent an equivalent precision.

Single spin-echo proton transverse relaxometry of iron-loaded liver

A single-spin-echo methodology is described for the measurement and imaging of proton transverse relaxation rates (R2) in iron-loaded and normal human liver tissue in vivo. The methodology brings together previously reported techniques dealing with (i) the changes in gain between each spin-echo acquisition, (ii) signal level offset due to background noise, (iii) estimation of signal intensities in decay curves at time zero to enable reliable extraction of relaxation times from tissues with very short T2 values, (iv) bi-exponential modelling of decay curves with a small number of data points, and (v) reduction of respiratory motion artefacts. The accuracy of the technique is tested on aqueous manganese chloride solutions yielding a relaxivity of 74.1+/-0.3 s-1 (mM)-1, consistent with previous reports. The precision of the in vivo measurement of mean liver R2 values is tested through duplicate measurements on 10 human subjects with mean liver R2 values ranging from 26 to 220 s-1. The random uncertainty on the measurement of mean liver R2 was found to be 7.7%.

Noninvasive measurement and imaging of liver iron concentrations using proton magnetic resonance

Measurement of liver iron concentration (LIC) is necessary for a range of iron-loading disorders such as hereditary hemochromatosis, thalassemia, sickle cell disease, aplastic anemia, and myelodysplasia. Currently, chemical analysis of needle biopsy specimens is the most common accepted method of measurement. This study presents a readily available noninvasive method of measuring and imaging LICs in vivo using clinical 1.5-T magnetic resonance imaging units. Mean liver proton transverse relaxation rates (R2) were measured for 105 humans. A value for the LIC for each subject was obtained by chemical assay of a needle biopsy specimen. High degrees of sensitivity and specificity of R2 to biopsy LICs were found at the clinically significant LIC thresholds of 1.8, 3.2, 7.0, and 15.0 mg Fe/g dry tissue. A calibration curve relating liver R2 to LIC has been deduced from the data covering the range of LICs from 0.3 to 42.7 mg Fe/g dry tissue. Proton transverse relaxation rates in aqueous paramagnetic solutions were also measured on each magnetic resonance imaging unit to ensure instrument-independent results. Measurements of proton transverse relaxivity of aqueous MnCl2 phantoms on 13 different magnetic resonance imaging units using the method yielded a coefficient of variation of 2.1%.