Single Field Strength Research Articles

Large-scale, multi-vendor, multi-protocol, quality-controlled analysis of clinical cine CMR using artificial intelligence

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): Advancing Impact Award scheme of the EPSRC Impact Acceleration Account at King’s College London Background Artificial intelligence (AI) has the potential to facilitate the automation of CMR analysis for biomarker extraction. However, most AI algorithms are trained on a specific input domain (e.g., scanner vendor or hospital-tailored imaging protocol) and lack the robustness to perform optimally when applied to CMR data from other input domains. Purpose To develop and validate a robust CMR analysis tool for automatic segmentation and cardiac function analysis which achieves state-of-the-art performance for multi-vendor short-axis cine CMR images. Methods The current work is an extension of our previously published quality-controlled AI-based tool for cine CMR analysis [1]. We deployed an AI algorithm that is equipped to handle different image sizes and domains automatically - the ‘nnU-Net’ framework [2] - and retrained our tool using the UK Biobank (UKBB) cohort population (n = 4,872) and a large database of clinical CMR studies obtained from two NHS hospitals (n = 3,406). The NHS hospital data came from three different scanner types: Siemens Aera 1.5T (n = 1,419), Philips Achieva 1.5T and 3T (n = 1,160), and Philips Ingenia 1.5T (n = 827). The ‘nnU-net’ was used to segment both ventricles and the myocardium. The proposed method was evaluated on randomly selected test sets from UKBB (n = 488) and NHS (n = 331) and on two external publicly available databases of clinical CMRs acquired on Philips, Siemens, General Electric (GE), and Canon CMR scanners – ACDC (n = 100) [3] and M&Ms (n = 321) [4]. We calculated the Dice scores - which measure the overlap between manual and automatic segmentations - and compared manual vs AI-based measures of biventricular volumes and function. Results Table 1 shows that the Dice scores for the NHS, ACDC, and M&Ms scans are similar to those obtained in the highly controlled, single vendor and single field strength UKBB scans. Although our AI-based tool was only trained on CMR scans from two vendors (Philips and Siemens), it performs similarly in unseen vendors (GE and Canon). Furthermore, it achieves state-of-the-art performance in online segmentation challenges, without being specifically trained on these databases. Table 1 also shows good agreement between manual and automated clinical measures of ejection fraction and ventricular volume and mass. Conclusions We show that our proposed AI-based tool, which combines training on a large-scale multi-domain CMR database with a state-of-the-art AI algorithm, allows us to robustly deal with routine clinical data from multiple centres, vendors, and field strengths. This is a fundamental step for the clinical translation of AI algorithms. Moreover, our method yields a range of additional metrics of cardiac function (filling and ejection rates, regional wall motion, and strain) at no extra computational cost.

Standardization of T1-mapping in cardiovascular magnetic resonance using clustered structuring for benchmarking normal ranges

BackgroundCardiovascular magnetic resonance T1-mapping is increasingly used for tissue characterization, commonly based on Modified Look-Locker Inversion recovery (MOLLI). However, there are numerous MOLLI variants with differing normal ranges. This lack of standardization presents confusion and difficulty in inter-center comparisons, hindering widespread adoption of T1-mapping. MethodsTo address this, we performed a structured literature search for native left ventricular myocardial T1-mapping in healthy humans measured using MOLLI variants at 1.5 and 3 Tesla, across scanner vendors. We then used k-means clustering to structure normal MOLLI-T1 values according to magnetic field strength, and investigated correlations between common imaging parameters: repetition time (TR), echo time (TE), flip angle (FA). ResultsWe analyzed data from 2207 healthy controls in 76 independent reports. Normal MOLLI-T1 standard deviations varied by 11-fold, and dependencies on TE, TR, and FA differed between 1.5 T and 3 T, thwarting meaningful T1 standardization even within a single field strength, including the use of Z-score. However, divergent MOLLI-T1 norms may be structured using data clustering. For 1.5 T, two clusters emerged: Cluster11.5T: T1 = 958 ± 16 ms (n = 1280); Cluster21.5T: T1 = 1027 ± 19 ms (n = 386). For 3 T, three clusters emerged: Cluster13T: T1 = 1160 ± 21 ms (n = 330); Cluster23T: T1 = 1067 ± 18 ms (n = 178); Cluster33T: T1 = 1227 ± 19 ms (n = 41). We then propose the concept of an online calculator for assigning local norms to a known MOLLI-T1 cluster, allowing benchmarking against published norms. ConclusionClustered structuring allows T1 standardization of widely-divergent MOLLI variants, benchmarking local norms (usually based on smaller samples) against published norms (larger samples). This may increase confidence and quality control in method implementation, facilitating wider clinical adoption of T1-mapping.

Uncertainty Estimations for Collision Cross Section Determination via Uniform Field Drift Tube-Ion Mobility-Mass Spectrometry.

Uniform field drift tube ion mobility-mass spectrometry (DTIM-MS) has emerged as a valuable tool for a range of analytical applications. In focus here are standardized collisional cross section values from DTIM-MS (DTCCS) as a candidate identification point for various analytical workflows. Of critical importance in establishing this parameter as a valid identification point is a rugged estimation of uncertainties according to the procedures used for their derivation. Relying on the assumption of the zero-field limit, the primary method of measurement for DTCCS values involves experimental determination of arrival times of an ion measured at several different field strengths transiting a drift tube filled with high purity drift gas, while a method using measurements of external calibrants at a single field strength is employed to allow for online measurements of transient signals (e.g., chromatographic peaks). Both approaches are here considered with respect to the uncertainty of input experimental variables (temperature, pressure, voltages, physical constants) and the steps of the calibration function employed. Estimations of uncertainty were performed according to EURACHEM with Monte Carlo simulations and reveal that existing consensus calibration standards from experimental stepped-field IM-MS determinations have estimated expanded uncertainties in the range of 2.7 to 4.6% (k = 2). Application of these standards for calibration considering these input uncertainties reveals uncertainty estimates of 4.7-9.1% (k = 2) for measured values using an established single-field calibration approach. Finally, directions for improving this situation via new experimental efforts toward standard reference and calibration materials are presented.

Investigating conditional GAN performance with different generator architectures, an ensemble model, and different MR scanners for MR-sCT conversion

Recent developments in magnetic resonance (MR) to synthetic computed tomography (sCT) conversion have shown that treatment planning is possible without an initial planning CT. Promising conversion results have been demonstrated recently using conditional generative adversarial networks (cGANs). However, the performance is generally only tested on images from one MR scanner, which neglects the potential of neural networks to find general high-level abstract features. In this study, we explored the generalizability of the generator models, trained on a single field strength scanner, to data acquired with higher field strengths. T2-weighted 0.35T MRIs and CTs from 51 patients treated for prostate (40) and cervical cancer (11) were included. 25 of them were used to train four different generators (SE-ResNet, DenseNet, U-Net, and Embedded Net). Further, an ensemble model was created from the four network outputs. The models were validated on 16 patients from a 0.35T MR scanner. Further, the trained models were tested on the Gold Atlas dataset, containing T2-weighted MR scans of different field strengths; 1.5T(7) and 3T(12), and 10 patients from the 0.35T scanner. The sCTs were dosimetrically compared using clinical VMAT plans for all test patients. For the same scanner (0.35T), the results from the different models were comparable on the test set, with only minor differences in the mean absolute error (MAE) (35-51HU body). Similar results were obtained for conversions of 3T GE Signa and the 3T GE Discovery images (40-62HU MAE) for three of the models. However, larger differences were observed for the 1.5T images (48-65HU MAE). The overall best model was found to be the ensemble model. All dose differences were below 1%. This study shows that it is possible to generalize models trained on images of one scanner to other scanners and different field strengths. The best metric results were achieved by the combination of all networks.

Multi-Tuned Cable Traps for Multinuclear MRI and MRS.

The method of pole-insertion for multi-tuning cable traps was studied for multinuclear MRI and MRS applications. Relative efficiency of the different cable trap modes was studied as component values were varied and at four different magnetic field strengths. In all cases, efficiencies were compared to equivalent single-tuned designs. The multi-tuned traps were able to block shield currents at multiple frequencies with only slightly degraded efficiencies as compared to their single-tuned counterparts. As in double-tuned coil design, the cable trap effectiveness at each frequency was found to be highly dependent on the trap inductor value with larger trap inductances leading to worse efficiency at the lower frequency but better efficiency at the higher frequency. This relationship held at all field strengths examined. This work presents design guidelines for the double-tuning method that are useful when designing RF coils for multinuclear studies. The design takes up less space than using two single-tuned cable traps mounted in series as is commonly done. Triple-tuned and "floating" designs were also demonstrated as proofs-of-concept for a single field strength and showed great promise to prove similarly useful in future studies. For many applications such as when using high-density array coils, finding a space-efficient solution to eliminate common-mode currents could be of significant benefit. This multi-tuned approach provides space efficiency at a small cost in trapping efficiency.

Determination of current-induced spin-orbit effective magnetic field in GaMnAs ferromagnetic semiconductor

We report a method for accurate determination of the strength of the current-induced spin-orbit (SO) field in ferromagnetic GaMnAs films. The SO-field manifests itself in the form of a hysteresis between planar Hall resistances (PHR) measured with positive and negative currents as an applied magnetic field is rotated in the sample plane at constant field strength. The width of the hysteresis, which is related to the strength of the SO-field, is observed to change significantly for different values of the rotating external field strength. Since the SO field occurring at a given current is an intrinsic property of the crystal, such a field dependence of the hysteresis indicates that the width of the hysteresis measured with a single field strength is insufficient for determining the SO field. However, using a model based on magnetic free energy that includes the effects of magnetic anisotropy and the SO-field as developed in the present paper, we show that the SO field for a given current density can be accurately established by fitting to the experimentally observed dependence of transition angles of PHR measured with different applied field strengths. Using the known dependence of magnetic anisotropy of GaMnAs on temperature, we also show that this method applies reliably as the temperature varies.

Variation in longitudinal trajectories of regional brain volumes of healthy men and women (ages 10 to 85years) measured with atlas-based parcellation of MRI

Numerous cross-sectional MRI studies have characterized age-related differences in regional brain volumes that differ with structure and tissue type. The extent to which cross-sectional assumptions about change are accurate depictions of actual longitudinal measurement remains controversial. Even longitudinal studies can be limited by the age range of participants, sex distribution of the samples, and scan intervals. To address these issues, we calculated trajectories of regional brain volume changes from T1-weighted (SPGR) MRI data, quantified with our automated, unsupervised SRI24 atlas-based registration and parcellation method. Longitudinal MRIs were acquired at 3T in 17 boys and 12 girls, age 10 to 14 years, and 41 men and 41 women, age 20 to 85 years at first scan. Application of a regression-based correction function permitted merging of data acquired at 3T field strength with data acquired at 1.5T from additional subjects, thereby expanding the sample to a total of 55 men and 67 women, age 20 to 85 years at first scan. Adjustment for individual supratentorial volume removed regional volume differences between men and women due to sex-related differences in head size. Individual trajectories were computed from data collected on 2 to 6 MRIs at a single field strength over a ~1 to 8 year interval. Using linear mixed-effects models, the pattern of trajectories over age indicated: rises in ventricular and Sylvian fissure volumes, with older individuals showing faster increases than younger ones; declines in selective cortical volumes with faster tissue shrinkage in older than younger individuals; little effect of aging on volume of the corpus callosum; more rapid expansion of CSF-filled spaces in men than women after age 60 years; and evidence for continued growth in central white matter through early adulthood with accelerated decline in senescence greater in men than women.

Combining atlas-based parcellation of regional brain data acquired across scanners at 1.5T and 3.0T field strengths

Longitudinal brain morphometric studies designed for data acquisition at a single MRI field strength can be seriously limited by system replacements from lower to higher field strength. Merging data across field strengths has not been endorsed for a variety of reasons, yet the ability to combine such data would broaden longitudinal investigations. To determine whether structural T1-weighted MRI data acquired across MR field strengths could be merged, parcellations of archival SPGR data acquired in 114 individuals at 1.5 T and at 3.0 T within 3 weeks of each other were compared. The first set of analyses examined 1) the correspondence between regional tissue volumes derived from data collected at 1.5 T and 3.0 T and 2) whether there were systematic differences for which a correction factor could be determined and applied to improve measurement agreement. Comparability of regional volume determination at 1.5 T and 3.0 T was assessed with intraclass correlation (ICC) computed on volumes derived from the automated and unsupervised SRI24 atlas registration and parcellation method. A second set of analyses measured the reliability of the registration and quantification using the same approach on longitudinal data acquired in 69 healthy adults at a single field strength, 1.5 T, at an interval < 2 years. The mainstay of the analyses was based on the SRI24 method; to examine the potential of merging data across field strengths and across image analysis packages, a secondary set of analyses used FreeSurfer instead of the SRI24 method. For both methods, a regression-based linear correction function significantly improved correspondence. The results indicated high correspondence between most selected cortical, subcortical, and CSF-filled spaces; correspondence was lowest in the globus pallidus, a region rich in iron, which in turn has a considerable field-dependent effect on signal intensity. Thus, the application of a regression-based correction function that improved the correspondence in regional volume estimations argues well for the proposition that selected T1-weighted regional anatomical brain data can be reliably combined across 1.5 T and 3.0 T field strengths with the application of an appropriate correction procedure.

MRI estimates of brain iron concentration in normal aging using quantitative susceptibility mapping

Quantifying tissue iron concentration in vivo is instrumental for understanding the role of iron in physiology and in neurological diseases associated with abnormal iron distribution. Herein, we use recently-developed Quantitative Susceptibility Mapping (QSM) methodology to estimate the tissue magnetic susceptibility based on MRI signal phase. To investigate the effect of different regularization choices, we implement and compare ℓ1 and ℓ2 norm regularized QSM algorithms. These regularized approaches solve for the underlying magnetic susceptibility distribution, a sensitive measure of the tissue iron concentration, that gives rise to the observed signal phase. Regularized QSM methodology also involves a pre-processing step that removes, by dipole fitting, unwanted background phase effects due to bulk susceptibility variations between air and tissue and requires data acquisition only at a single field strength. For validation, performances of the two QSM methods were measured against published estimates of regional brain iron from postmortem and in vivo data. The in vivo comparison was based on data previously acquired using Field-Dependent Relaxation Rate Increase (FDRI), an estimate of MRI relaxivity enhancement due to increased main magnetic field strength, requiring data acquired at two different field strengths. The QSM analysis was based on susceptibility-weighted images acquired at 1.5 T, whereas FDRI analysis used Multi-Shot Echo-Planar Spin Echo images collected at 1.5 T and 3.0 T. Both datasets were collected in the same healthy young and elderly adults. The in vivo estimates of regional iron concentration comported well with published postmortem measurements; both QSM approaches yielded the same rank ordering of iron concentration by brain structure, with the lowest in white matter and the highest in globus pallidus. Further validation was provided by comparison of the in vivo measurements, ℓ1-regularized QSM versus FDRI and ℓ2-regularized QSM versus FDRI, which again yielded perfect rank ordering of iron by brain structure. The final means of validation was to assess how well each in vivo method detected known age-related differences in regional iron concentrations measured in the same young and elderly healthy adults. Both QSM methods and FDRI were consistent in identifying higher iron concentrations in striatal and brain stem ROIs (i.e., caudate nucleus, putamen, globus pallidus, red nucleus, and substantia nigra) in the older than in the young group. The two QSM methods appeared more sensitive in detecting age differences in brain stem structures as they revealed differences of much higher statistical significance between the young and elderly groups than did FDRI. However, QSM values are influenced by factors such as the myelin content, whereas FDRI is a more specific indicator of iron content. Hence, FDRI demonstrated higher specificity to iron yet yielded noisier data despite longer scan times and lower spatial resolution than QSM. The robustness, practicality, and demonstrated ability of predicting the change in iron deposition in adult aging suggest that regularized QSM algorithms using single-field-strength data are possible alternatives to tissue iron estimation requiring two field strengths.

Impact of non-uniform surface magnetic fields on stellar winds

Observations of active stars reveal highly non-uniform surface distributions of magnetic flux. Theoretical models considering magnetised stellar winds however often presume uniform surface magnetic fields, characterised by a single magnetic field strength. The present work investigates the impact of non-uniform surface magnetic field distributions on the stellar mass and angular momentum loss rates. The approach of Weber & Davis (1967) is extended to non-equatorial latitudes to quantify the impact of latitude-dependent magnetic field distributions over a large range of stellar rotation rates and thermal wind properties. The analytically prescribed field patterns are dominated by magnetic flux concentrations at intermediate and high latitudes. The global stellar mass loss rates are found to be rather insensitive to non-uniformities of the surface magnetic field. Depending on the non-uniformity of the field distribution, the angular momentum loss rates deviate in contrast at all rotation rates between -60% and 10% from the Weber & Davis-values, and the effective Alfvenic radii up to about +/-25%. These large variations albeit equal amounts of total magnetic flux indicate that a classification of stellar surface magnetic fields through a single field strength is insufficient, and that their non-uniformity has to be taken into account. The consequences for applications involving magnetised stellar winds are discussed in view of the rotational evolution of solar-like stars and of the observational determination of their mass loss rates using the terminal velocity and ram pressure of the wind. For rapidly rotating stars the latitudinal variation of the wind ram pressure is found to exceed, depending on the actual field distribution on the stellar surface, over two orders of magnitude.