Search

The adoption of machine learning (ML) and, more specifically, deep learning (DL) applications into all major areas of our lives is underway. The development of trustworthy AI is especially important in medicine due to the large implications for patients' lives. While trustworthiness concerns various aspects including ethical, transparency and safety requirements, we focus on the importance of data quality (training/test) in DL. Since data quality dictates the behaviour of ML products, evaluating data quality will play a key part in the regulatory approval of medical ML products. We perform a systematic review following PRISMA guidelines using the databases Web of Science, PubMed and ACM Digital Library. We identify 5408 studies, out of which 120 records fulfil our eligibility criteria. From this literature, we synthesise the existing knowledge on data quality frameworks and combine it with the perspective of ML applications in medicine. As a result, we propose the METRIC-framework, a specialised data quality framework for medical training data comprising 15 awareness dimensions, along which developers of medical ML applications should investigate the content of a dataset. This knowledge helps to reduce biases as a major source of unfairness, increase robustness, facilitate interpretability and thus lays the foundation for trustworthy AI in medicine. The METRIC-framework may serve as a base for systematically assessing training datasets, establishing reference datasets, and designing test datasets which has the potential to accelerate the approval of medical ML products. [ABSTRACT FROM AUTHOR]

Purpose: Traditional phase‐contrast MRI is affected by displacement artifacts caused by non‐synchronized spatial‐ and velocity‐encoding time points. The resulting inaccurate velocity maps can affect the accuracy of derived hemodynamic parameters. This study proposes and characterizes a 3D radial phase‐contrast UTE (PC‐UTE) sequence to reduce displacement artifacts. Furthermore, it investigates the displacement of a standard Cartesian flow sequence by utilizing a displacement‐free synchronized‐single‐point‐imaging MR sequence (SYNC‐SPI) that requires clinically prohibitively long acquisition times. Methods: 3D flow data was acquired at 3T at three different constant flow rates and varying spatial resolutions in a stenotic aorta phantom using the proposed PC‐UTE, a Cartesian flow sequence, and a SYNC‐SPI sequence as reference. Expected displacement artifacts were calculated from gradient timing waveforms and compared to displacement values measured in the in vitro flow experiments. Results: The PC‐UTE sequence reduces displacement and intravoxel dephasing, leading to decreased geometric distortions and signal cancellations in magnitude images, and more spatially accurate velocity quantification compared to the Cartesian flow acquisitions; errors increase with velocity and higher spatial resolution. Conclusion: PC‐UTE MRI can measure velocity vector fields with greater accuracy than Cartesian acquisitions (although pulsatile fields were not studied) and shorter scan times than SYNC‐SPI. As such, this approach is superior to traditional Cartesian 3D and 4D flow MRI when spatial misrepresentations cannot be tolerated, for example, when computational fluid dynamics simulations are compared to or combined with in vitro or in vivo measurements, or regional parameters such as wall shear stress are of interest. [ABSTRACT FROM AUTHOR]

Purpose: Depicting the stiffness of biological soft tissues, MR elastography (MRE) has a wide range of diagnostic applications. The purpose of this study was to improve the temporal resolution of 2D hepatic MRE in order to provide more rapid feedback on the quality of the wavefield and ensure better temporal sampling of respiration‐induced stiffness changes. Methods: We developed a rapid MRE sequence that uses 2D segmented gradient‐echo spiral readout to encode 40 Hz harmonic vibrations and generate stiffness maps within 625 ms. We demonstrate the use of this technique as a rapid test for shear wave amplitudes and overall MRE image quality and as a method for monitoring respiration‐induced stiffness changes in the liver in comparison to 3D MRE and ultrasound‐based time‐harmonic elastography. Results: Subsecond MRE allowed monitoring of increasing shear wave amplitudes in the liver with increasing levels of external stimulation within a single breath‐hold. Furthermore, the technique detected respiration‐induced changes in liver stiffness with peak values (1.83 ± 0.22 m/s) at end‐inspiration, followed by softer values during forced abdominal pressure (1.60 ± 0.22 m/s) and end‐expiration (1.49 ± 0.22 m/s). The effects of inspiration and expiration were confirmed by time‐harmonic elastography. Conclusion: Our results suggest that subsecond MRE of the liver is useful for checking MRE driver settings and monitoring breathing‐induced changes in liver stiffness in near real time. [ABSTRACT FROM AUTHOR]

Background: Cardiac MRI has become the gold‐standard imaging technique for assessing cardiovascular morphology and function. In spite of this, its slow data acquisition process presents imaging challenges due to the motion from heartbeats, respiration, and blood flow. In recent studies, deep learning (DL) algorithms have shown promising results for the task of image reconstruction. However, there have been instances where they have introduced artifacts that may be misinterpreted as pathologies or may obscure the detection of pathologies. Therefore, it is important to obtain a metric, such as the uncertainty of the network output, that identifies such artifacts. However, this can be quite challenging for large‐scale image reconstruction problems such as dynamic multi‐coil non‐Cartesian MRI. Purpose: To efficiently quantify uncertainties of a physics‐informed DL‐based image reconstruction method for a large‐scale accelerated 2D multi‐coil dynamic radial MRI reconstruction problem, and demonstrate the benefits of physics‐informed DL over model‐agnostic DL in reducing uncertainties while at the same time improving image quality. Methods: We extended a recently proposed physics‐informed 2D U‐Net that learns spatio‐temporal slices (named XT‐YT U‐Net), and employed it for the task of uncertainty quantification (UQ) by using Monte Carlo dropout and a Gaussian negative log‐likelihood loss function. Our data comprised 2D dynamic MR images acquired with a radial balanced steady‐state free precession sequence. The XT‐YT U‐Net, which allows for training with a limited amount of data, was trained and validated on a dataset of 15 healthy volunteers, and further tested on data from four patients. An extensive comparison between physics‐informed and model‐agnostic neural networks (NNs) concerning the obtained image quality and uncertainty estimates was performed. Further, we employed calibration plots to assess the quality of the UQ. Results: The inclusion of the MR‐physics model of data acquisition as a building block in the NN architecture led to higher image quality (NRMSE: −33±8.2%$-33 \pm 8.2 \%$, PSNR: 6.3±1.3%$6.3 \pm 1.3 \%$, and SSIM: 1.9±0.96%$1.9 \pm 0.96 \%$), lower uncertainties (−46±8.7%$-46 \pm 8.7 \%$), and, based on the calibration plots, an improved UQ compared to its model‐agnostic counterpart. Furthermore, the UQ information can be used to differentiate between anatomical structures (e.g., coronary arteries, ventricle boundaries) and artifacts. Conclusions: Using an XT‐YT U‐Net, we were able to quantify uncertainties of a physics‐informed NN for a high‐dimensional and computationally demanding 2D multi‐coil dynamic MR imaging problem. In addition to improving the image quality, embedding the acquisition model in the network architecture decreased the reconstruction uncertainties as well as quantitatively improved the UQ. The UQ provides additional information to assess the performance of different network approaches. [ABSTRACT FROM AUTHOR]

Purpose: To allow for T1 mapping of the myocardium within 2.3 s for a 2D slice utilizing cardiac motion‐corrected, model‐based image reconstruction. Methods: Golden radial data acquisition is continuously carried out for 2.3 s after an inversion pulse. In a first step, dynamic images are reconstructed which show both contrast changes due to T1 recovery and anatomical changes due to the heartbeat. An image registration algorithm with a signal model for T1 recovery is applied to estimate non‐rigid cardiac motion. In a second step, estimated motion fields are applied during an iterative model‐based T1 reconstruction. The approach was evaluated in numerical simulations, phantom experiments and in in‐vivo scans in healthy volunteers. Results: The accuracy of cardiac motion estimation was shown in numerical simulations with an average motion field error of 0.7 ± 0.6 mm for a motion amplitude of 5.1 mm. The accuracy of T1 estimation was demonstrated in phantom experiments, with no significant difference (p = 0.13) in T1 estimated by the proposed approach compared to an inversion‐recovery reference method. In vivo, the proposed approach yielded 1.3 × 1.3 mm T1 maps with no significant difference (p = 0.77) in T1 and SDs in comparison to a cardiac‐gated approach requiring 16 s scan time (i.e., seven times longer than the proposed approach). Cardiac motion correction improved the precision of T1 maps, shown by a 40% reduced SD. Conclusion: We have presented an approach that provides T1 maps of the myocardium in 2.3 s by utilizing both cardiac motion correction and model‐based T1 reconstruction. [ABSTRACT FROM AUTHOR]

Background: Unrolled neural networks (NNs) have been extensively applied to different image reconstruction problems across all imaging modalities. A key component of the latter is that they allow for physics‐informed learning of the regularization method, which is parametrized by the NN. However, due to the lack of understanding of deep NNs from a theoretical point of view, unrolled NNs are still black‐boxes when the regularizers are given by deep NNs, for example, U‐Nets. Purpose: Dictionarylearning (DL) is a well‐established regularization method, which is based on learning a transform to sparsely approximate the signals of interest. Typically, DL‐based image reconstruction either employs a dictionary, which was pretrained on a set of patches which were extracted from ground‐truth images or a dictionary which is jointly trained during the reconstruction. However, in both cases, the used DL‐algorithms are not designed to take into account the reconstruction problem or the underlying physical model, which describes the imaging process. In this work, we propose a DL‐algorithm based on unrolled NNs to overcome these limitations. Methods: We construct an unrolled NN, which corresponds to an unrolled DL‐based reconstruction algorithm and train the unrolled NN to optimize its weights, that is, the atoms of the dictionary, by back‐propagation in a supervised manner. Further, we propose a new way to employ a 2D dictionary in the spatio‐temporal domain. We tested and evaluated the method on an accelerated cardiac cine MR image reconstruction problem using 216/36/36 dynamic images for training, validation, and testing and compared it to two well‐known state‐of‐the‐art approaches for cardiac cine MRI based on deep iterative CNNs. Further, we analyze the obtained dictionaries in terms of dictionary‐coherence and structure of the atoms. Last, we compare the reported methods in terms of stability by applying them to an entirely different dataset consisting of 49 different test images. Results: The investigated physics‐informed DL‐approach yields significantly more accurate reconstructions compared to the DL‐method, which uses dictionaries obtained by decoupled pretraining, thereby providing an improvement of up to 4.90 dB in terms of PSNR and 5% in terms of SSIM. Further, the proposed spatio‐temporal 2D dictionary outperforms the 1D and 3D dictionaries by preventing smoothing of image details while still accurately removing undersampling artifacts and noise resulting in an increase of up to 1.10 dB in terms of PSNR and 4% in terms of SSIM. Although being surpassed by the CNNs on the first dataset, the proposed NNs‐based DL method is more stable compared to the latter approach and yields comparable results on the second dataset. Last, it has the advantage of being entirely interpretable in each component. Conclusions: The presented physics‐informed NN can be used as training algorithm for a classical and interpretable data‐driven regularization method based on a learned dictionary, which can then not only be linked to the considered data but also to the reconstruction method that the NN defines. [ABSTRACT FROM AUTHOR]

Multiple sites within Germany operate human MRI systems with magnetic fields either at 7 Tesla or 9.4 Tesla. In 2013, these sites formed a network to facilitate and harmonize the research being conducted at the different sites and make this technology available to a larger community of researchers and clinicians not only within Germany, but also worldwide. The German Ultrahigh Field Imaging (GUFI) network has defined a strategic goal to establish a 14 Tesla whole-body human MRI system as a national research resource in Germany as the next progression in magnetic field strength. This paper summarizes the history of this initiative, the current status, the motivation for pursuing MR imaging and spectroscopy at such a high magnetic field strength, and the technical and funding challenges involved. It focuses on the scientific and science policy process from the perspective in Germany, and is not intended to be a comprehensive systematic review of the benefits and technical challenges of higher field strengths. [ABSTRACT FROM AUTHOR]

Purpose: Subject‐tailored parallel transmission pulses for ultra‐high fields body applications are typically calculated based on subject‐specific B1+$$ {\mathrm{B}}_1^{+} $$‐maps of all transmit channels, which require lengthy adjustment times. This study investigates the feasibility of using deep learning to estimate complex, channel‐wise, relative 2D B1+$$ {\mathrm{B}}_1^{+} $$‐maps from a single gradient echo localizer to overcome long calibration times. Methods: 126 channel‐wise, complex, relative 2D B1+$$ {\mathrm{B}}_1^{+} $$‐maps of the human heart from 44 subjects were acquired at 7T using a Cartesian, cardiac gradient‐echo sequence obtained under breath‐hold to create a library for network training and cross‐validation. The deep learning predicted maps were qualitatively compared to the ground truth. Phase‐only B1+$$ {\mathrm{B}}_1^{+} $$‐shimming was subsequently performed on the estimated B1+$$ {\mathrm{B}}_1^{+} $$‐maps for a region of interest covering the heart. The proposed network was applied at 7T to 3 unseen test subjects. Results: The deep learning‐based B1+$$ {\mathrm{B}}_1^{+} $$‐maps, derived in approximately 0.2 seconds, match the ground truth for the magnitude and phase. The static, phase‐only pulse design performs best when maximizing the mean transmission efficiency. In‐vivo application of the proposed network to unseen subjects demonstrates the feasibility of this approach: the network yields predicted B1+$$ {\mathrm{B}}_1^{+} $$‐maps comparable to the acquired ground truth and anatomical scans reflect the resulting B1+$$ {\mathrm{B}}_1^{+} $$‐pattern using the deep learning‐based maps. Conclusion: The feasibility of estimating 2D relative B1+$$ {\mathrm{B}}_1^{+} $$‐maps from initial localizer scans of the human heart at 7T using deep learning is successfully demonstrated. Because the technique requires only sub‐seconds to derive channel‐wise B1+$$ {\mathrm{B}}_1^{+} $$‐maps, it offers high potential for advancing clinical body imaging at ultra‐high fields. [ABSTRACT FROM AUTHOR]

Purpose: Flow quantification by phase‐contrast MRI is hampered by spatially varying background phase offsets. Correction performance by polynomial regression on stationary tissue may be affected by outliers such as wrap‐around or constant flow. Therefore, we propose an alternative, M‐estimate SAmple Consensus (MSAC) to reject outliers, and improve and fully automate background phase correction. Methods: The MSAC technique fits polynomials to randomly drawn small samples from the image. Over several trials, it aims to find the best consensus set of valid pixels by rejecting outliers to the fit and minimizing the residuals of the remaining pixels. The robustness of MSAC to its few parameters was investigated and verified using third‐order polynomial correction fits on a total of 118 2D flow (97 with wrap‐around) and 18 4D flow data sets (14 with wrap‐around), acquired at 1.5 T and 3 T. Background phase was compared with standard stationary correction and phantom correction. Pulmonary/systemic flow ratios in 2D flow were derived, and exemplary 4D flow analysis was performed. Results: The MSAC technique is robust over a range of parameter choices, and a unique set of parameters is suitable for both 2D and 4D flow. In 2D flow, phase errors were significantly reduced by MSAC compared with stationary correction (p = 0.005), and stationary correction shows larger errors in pulmonary/systemic flow ratios compared with MSAC. In 4D flow, MSAC shows similar performance as stationary correction. Conclusions: The MSAC method provides fully automated background phase correction to 2D and 4D flow data and shows improved robustness over stationary correction, especially with outliers present. [ABSTRACT FROM AUTHOR]

Purpose: Myocardial fat infiltrations are associated with a range of cardiomyopathies. The purpose of this study was to perform cardio‐respiratory motion‐correction for model‐based water‐fat separation to image fatty infiltrations of the heart in a free‐breathing, non‐cardiac‐triggered high‐resolution 3D MRI acquisition. Methods: Data were acquired in nine patients using a free‐breathing, non‐cardiac‐triggered high‐resolution 3D Dixon gradient‐echo sequence and radial phase encoding trajectory. Motion correction was combined with a model‐based water‐fat reconstruction approach. Respiratory and cardiac motion models were estimated using a dual‐mode registration algorithm incorporating both motion‐resolved water and fat information. Qualitative comparisons of fat structures were made between 2D clinical routine reference scans and reformatted 3D motion‐corrected images. To evaluate the effect of motion correction the local sharpness of epicardial fat structures was analyzed for motion‐averaged and motion‐corrected fat images. Results: The reformatted 3D motion‐corrected reconstructions yielded qualitatively comparable fat structures and fat structure sharpness in the heart as the standard 2D breath‐hold. Respiratory motion correction improved the local sharpness on average by 32% ± 24% with maximum improvements of 81% and cardiac motion correction increased the sharpness further by another 15% ± 11% with maximum increases of 31%. One patient showed a fat infiltration in the myocardium and cardio‐respiratory motion correction was able to improve its visualization in 3D. Conclusion: The 3D water‐fat separated cardiac images were acquired during free‐breathing and in a clinically feasible and predictable scan time. Compared to a motion‐averaged reconstruction an increase in sharpness of fat structures by 51% ± 27% using the presented motion correction approach was observed for nine patients. [ABSTRACT FROM AUTHOR]

Purpose: Respiratory motion‐compensated (MC) 3D cardiac fat‐water imaging at 7T. Methods: Free‐breathing bipolar 3D triple‐echo gradient‐recalled‐echo (GRE) data with radial phase‐encoding (RPE) trajectory were acquired in 11 healthy volunteers (7M\4F, 21–35 years, mean: 30 years) with a wide range of body mass index (BMI; 19.9–34.0 kg/m2) and volunteer tailored B1+ shimming. The bipolar‐corrected triple‐echo GRE‐RPE data were binned into different respiratory phases (self‐navigation) and were used for the estimation of non‐rigid motion vector fields (MF) and respiratory resolved (RR) maps of the main magnetic field deviations (ΔB0). RR ΔB0 maps and MC ΔB0 maps were compared to a reference respiratory phase to assess respiration‐induced changes. Subsequently, cardiac binned fat‐water images were obtained using a model‐based, respiratory motion‐corrected image reconstruction. Results: The 3D cardiac fat‐water imaging at 7T was successfully demonstrated. Local respiration‐induced frequency shifts in MC ΔB0 maps are small compared to the chemical shifts used in the multi‐peak model. Compared to the reference exhale ΔB0 map these changes are in the order of 10 Hz on average. Cardiac binned MC fat‐water reconstruction reduced respiration induced blurring in the fat‐water images, and flow artifacts are reduced in the end‐diastolic fat‐water separated images. Conclusion: This work demonstrates the feasibility of 3D fat‐water imaging at UHF for the entire human heart despite spatial and temporal B1+ and B0 variations, as well as respiratory and cardiac motion. [ABSTRACT FROM AUTHOR]

Purpose: MRI at ultra‐high fields in the human body is highly challenging and requires lengthy calibration times to compensate for spatially heterogeneous B1+ profiles. This study investigates the feasibility of using pre‐computed universal pulses for calibration‐free homogeneous 3D flip angle distribution in the human heart at 7T. Methods: Twenty‐two channel‐wise 3D B1+ data sets were acquired under free‐breathing in 19 subjects to generate a library for an offline universal pulse (UP) design (group 1: 12 males [M] and 7 females [F], 21‐66 years, 19.8‐28.3 kg/m2). Three of these subjects (2M/1F, 21‐33 years, 20.8‐23.6 kg/m2) were re‐scanned on different days. A 4kT‐points UP optimized for the 22 channel‐wise 3D B1+ data sets in group 1 (UP22‐4kT) is proposed and applied at 7T in 9 new and unseen subjects (group 2: 4M/5F, 25‐56 years, 19.5‐35.3 kg/m2). Multiple tailored and universal static and dynamic parallel‐transmit (pTx) pulses were designed and evaluated for different permutations of the B1+ data sets in group 1 and 2. Results: The proposed UP22‐4kT provides low B1+ variation in all subjects, seen and unseen, without severe signal drops. Experimental data at 7T acquired with UP22‐4kT shows comparable image quality as data acquired with tailored‐4kT pulses and demonstrates successful calibration‐free pTx of the human heart. Conclusion: UP22‐4kT allows for calibration‐free homogeneous flip angle distributions across the human heart at 7T. Large inter‐subject variations because of sex, age, and body mass index are well tolerated. The proposed universal pulse removes the need for lengthy (10‐15 min) calibration scans and therefore has the potential to bring body imaging at 7T closer to the clinical application. [ABSTRACT FROM AUTHOR]