Unsupervised motion artifact correction of turbo spin-echo MRI using deep image prior.

PURPOSE: In MRI, motion artifacts can significantly degrade image quality. Motion artifact correction methods using deep neural networks usually required extensive training on large datasets, making them time-consuming and resource-intensive. In this paper, an unsupervised deep learning-based motion artifact correction method for turbo-spin echo MRI is proposed using the deep image prior framework. THEORY AND METHODS: The proposed approach takes advantage of the high impedance to motion artifacts offered by the neural network parameterization to remove motion artifacts in MR images. The framework consists of parameterization of MR image, automatic spatial transformation, and motion simulation model. The proposed method synthesizes motion-corrupted images from the motion-corrected images generated by the convolutional neural network, where an optimization process minimizes the objective function between the synthesized images and the acquired images. RESULTS: In the simulation study of 280 slices from 14 subjects, the proposed method showed a significant increase in the averaged structural similarity index measure by 0.2737 in individual coil images and by 0.4550 in the root-sum-of-square images. In addition, the ablation study demonstrated the effectiveness of each proposed component in correcting motion artifacts compared to the corrected images produced by the baseline method. The experiments on real motion dataset has shown its clinical potential. CONCLUSION: The proposed method exhibited significant quantitative and qualitative improvements in correcting rigid and in-plane motion artifacts in MR images acquired using turbo spin-echo sequence.

Self-supervised learning for denoising of multidimensional MRI data.

PURPOSE: To develop a fast denoising framework for high-dimensional MRI data based on a self-supervised learning scheme, which does not require ground truth clean image. THEORY AND METHODS: Quantitative MRI faces limitations in SNR, because the variation of signal amplitude in a large set of images is the key mechanism for quantification. In addition, the complex non-linear signal models make the fitting process vulnerable to noise. To address these issues, we propose a fast deep-learning framework for denoising, which efficiently exploits the redundancy in multidimensional MRI data. A self-supervised model was designed to use only noisy images for training, bypassing the challenge of clean data paucity in clinical practice. For validation, we used two different datasets of simulated magnetization transfer contrast MR fingerprinting (MTC-MRF) dataset and in vivo DWI image dataset to show the generalizability. RESULTS: The proposed method drastically improved denoising performance in the presence of mild-to-severe noise regardless of noise distributions compared to previous methods of the BM3D, tMPPCA, and Patch2self. The improvements were even pronounced in the following quantification results from the denoised images. CONCLUSION: The proposed MD-S2S (Multidimensional-Self2Self) denoising technique could be further applied to various multi-dimensional MRI data and improve the quantification accuracy of tissue parameter maps.

Registration and quantification network (RQnet) for IVIM-DKI analysis in MRI.

PURPOSE: A deep learning method is proposed for aligning diffusion weighted images (DWIs) and estimating intravoxel incoherent motion-diffusion kurtosis imaging parameters simultaneously. METHODS: We propose an unsupervised deep learning method that performs 2 tasks: registration and quantification for intravoxel incoherent motion-diffusion kurtosis imaging analysis. A common registration method in diffusion MRI is based on minimizing dissimilarity between various DWIs, which may result in registration errors due to different contrasts in different DWIs. We designed a novel unsupervised deep learning method for both accurate registration and quantification of various diffusion parameters. In order to generate motion-simulated training data and test data, 17 volunteers were scanned without moving their heads, and 4 volunteers moved their heads during the scan in a 3 Tesla MRI. In order to investigate the applicability of the proposed method to other organs, kidney images were also obtained. We compared the registration accuracy of the proposed method, statistical parametric mapping, and a deep learning method with a normalized cross-correlation loss. In the quantification part of the proposed method, a deep learning method that considered the diffusion gradient direction was used. RESULTS: Simulations and experimental results showed that the proposed method accurately performed registration and quantification for intravoxel incoherent motion-diffusion kurtosis imaging analysis. The registration accuracy of the proposed method was high for all b values. Furthermore, quantification performance was analyzed through simulations and in vivo experiments, where the proposed method showed the best performance among the compared methods. CONCLUSION: The proposed method aligns the DWIs and accurately quantifies the intravoxel incoherent motion-diffusion kurtosis imaging parameters.

Only-train-once MR fingerprinting for B0 and B1 inhomogeneity correction in quantitative magnetization-transfer contrast.

PURPOSE: To develop a fast, deep-learning approach for quantitative magnetization-transfer contrast (MTC)-MR fingerprinting (MRF) that simultaneously estimates multiple tissue parameters and corrects the effects of B0 and B1 variations. METHODS: An only-train-once recurrent neural network was designed to perform the fast tissue-parameter quantification for a large range of different MRF acquisition schedules. It enabled a dynamic scan-wise linear calibration of the scan parameters using the measured B0 and B1 maps, which allowed accurate, multiple-tissue parameter mapping. MRF images were acquired from 8 healthy volunteers at 3 T. Estimated parameter maps from the MRF images were used to synthesize the MTC reference signal (Zref ) through Bloch equations at multiple saturation power levels. RESULTS: The B0 and B1 errors in MR fingerprints, if not corrected, would impair the tissue quantification and subsequently corrupt the synthesized MTC reference images. Bloch equation-based numerical phantom studies and synthetic MRI analysis demonstrated that the proposed approach could correctly estimate water and semisolid macromolecule parameters, even with severe B0 and B1 inhomogeneities. CONCLUSION: The only-train-once deep-learning framework can improve the reconstruction accuracy of brain-tissue parameter maps and be further combined with any conventional MRF or CEST-MRF method.

Automatic segmentation of thyroid with the assistance of the devised boundary improvement based on multicomponent small dataset.

Deep learning has been widely considered in medical image segmentation. However, the difficulty of acquiring medical images and labels can affect the accuracy of the segmentation results for deep learning methods. In this paper, an automatic segmentation method is proposed by devising a multicomponent neighborhood extreme learning machine to improve the boundary attention region of the preliminary segmentation results. The neighborhood features are acquired by training U-Nets with the multicomponent small dataset, which consists of original thyroid ultrasound images, Sobel edge images and superpixel images. Afterward, the neighborhood features are selected by min-redundancy and max-relevance filter in the designed extreme learning machine, and the selected features are used to train the extreme learning machine to obtain supplementary segmentation results. Finally, the accuracy of the segmentation results is improved by adjusting the boundary attention region of the preliminary segmentation results with the supplementary segmentation results. This method combines the advantages of deep learning and traditional machine learning, boosting the accuracy of thyroid segmentation accuracy with a small dataset in a multigroup test.

Learning-based optimization of acquisition schedule for magnetization transfer contrast MR fingerprinting.

Magnetization transfer contrast MR fingerprinting (MTC-MRF) is a novel quantitative imaging method that simultaneously quantifies free bulk water and semisolid macromolecule parameters using pseudo-randomized scan parameters. To improve acquisition efficiency and reconstruction accuracy, the optimization of MRF sequence design has been of recent interest in the MRF field, but has been challenging due to the large number of degrees of freedom to be optimized in the sequence. Herein, we propose a framework for learning-based optimization of the acquisition schedule (LOAS), which optimizes RF saturation-encoded MRF acquisitions with a minimal number of scan parameters for tissue parameter determination. In a supervised learning framework, scan parameters were subsequently updated to minimize a predefined loss function that can directly represent tissue quantification errors. We evaluated the performance of the proposed approach with a numerical phantom and in in vivo experiments. For validation, MRF images were synthesized using the tissue parameters estimated from a fully connected neural network framework and compared with references. Our results showed that LOAS outperformed existing indirect optimization methods with regard to quantification accuracy and acquisition efficiency. The proposed LOAS method could be a powerful optimization tool in the design of MRF pulse sequences.

Unveiling the Inhomogeneous Nature of Strong Field Ionization in Extended Systems.

Intense light-induced fragmentation of spherical clusters produces highly energetic ions with characteristic spatial distributions. By subjecting argon clusters to a wavelength tunable laser, we show that ion emission energy and anisotropy can be controlled through the wavelength-isotropic and energetic for shorter wavelengths and increasingly anisotropic at longer wavelengths. The anisotropic part of the energy spectrum, consisting of multiply charged high-energy ions, is considerably more prominent at longer wavelengths. Classical molecular dynamics simulations reveal that cluster ionization occurs inhomogeneously producing a columnlike charge distribution along the laser polarization direction. This previously unknown distribution results from the dipole response of the neutral cluster which creates an enhanced field at the surface, preferentially triggering ionization at the poles. The subsequently formed nanoplasma provides an additional wavelength-dependent ionization mechanism through collisional ionization, efficiently homogenizing the system only at short wavelengths close to resonance. Our results open the door to studying polarization induced effects in nanostructures and complex molecules and provide a missing piece in our understanding of anisotropic ion emission.

MC2 -Net: motion correction network for multi-contrast brain MRI.

PURPOSE: A motion-correction network for multi-contrast brain MRI is proposed to correct in-plane rigid motion artifacts in brain MR images using deep learning. METHOD: The proposed method consists of 2 parts: image alignment and motion correction. Alignment of multi-contrast MR images is performed in an unsupervised manner by a CNN work, yielding transformation parameters to align input images in order to minimize the normalized cross-correlation loss among multi-contrast images. Then, fine-tuning for image alignment is performed by maximizing the normalized mutual information. The motion correction network corrects motion artifacts in the aligned multi-contrast images. The correction network is trained to minimize the structural similarity loss and the VGG loss in a supervised manner. All datasets of motion-corrupted images are generated using motion simulation based on MR physics. RESULTS: A motion-correction network for multi-contrast brain MRI successfully corrected artifacts of simulated motion for 4 test subjects, showing 0.96%, 7.63%, and 5.03% increases in the average structural simularity and 5.19%, 10.2%, and 7.48% increases in the average normalized mutual information for T1 -weighted, T2 -weighted, and T2 -weighted fluid-attenuated inversion recovery images, respectively. The experimental setting with image alignment and artifact-free input images for other contrasts shows better performances in correction of simulated motion artifacts. Furthermore, the proposed method quantitatively outperforms recent deep learning motion correction and synthesis methods. Real motion experiments from 5 healthy subjects demonstrate the potential of the proposed method for use in a clinical environment. CONCLUSION: A deep learning-based motion correction method for multi-contrast MRI was successfully developed, and experimental results demonstrate the validity of the proposed method.

Quantification of intravoxel incoherent motion with optimized b-values using deep neural network.

PURPOSE: To develop a framework for quantifying intravoxel incoherent motion (IVIM) parameters, where a neural network for quantification and b-values for diffusion-weighted imaging are simultaneously optimized. METHOD: A deep neural network (DNN) method is proposed for accurate quantification of IVIM parameters from multiple diffusion-weighted images. In addition, optimal b-values are selected to acquire the multiple diffusion-weighted images. The proposed framework consists of an MRI signal generation part and an IVIM parameter quantification part. Monte-Carlo (MC) simulations were performed to evaluate the accuracy of the IVIM parameter quantification and the efficacy of b-value optimization. In order to analyze the effect of noise on the optimized b-values, simulations were performed with five different noise levels. For in vivo data, diffusion images were acquired with the b-values from four b-values selection methods for five healthy volunteers at 3T MRI system. RESULTS: Experiment results showed that both the optimization of b-values and the training of DNN were simultaneously performed to quantify IVIM parameters. We found that the accuracies of the perfusion coefficient (Dp ) and perfusion fraction (f) were more sensitive to b-values than the diffusion coefficient (D) was. Furthermore, when the noise level changed, the optimized b-values also changed. Therefore, noise level has to be considered when optimizing b-values for IVIM quantification. CONCLUSION: The proposed scheme can simultaneously optimize b-values and train DNN to minimize quantification errors of IVIM parameters. The trained DNN can quantify IVIM parameters from the diffusion-weighted images obtained with the optimized b-values.

Unsupervised learning for magnetization transfer contrast MR fingerprinting: Application to CEST and nuclear Overhauser enhancement imaging.

PURPOSE: To develop a fast, quantitative 3D magnetization transfer contrast (MTC) framework based on an unsupervised learning scheme, which will provide baseline reference signals for CEST and nuclear Overhauser enhancement imaging. METHODS: Pseudo-randomized RF saturation parameters and relaxation delay times were applied in an MR fingerprinting framework to generate transient-state signal evolutions for different MTC parameters. Prospectively compressed sensing-accelerated (four-fold) MR fingerprinting images were acquired from 6 healthy volunteers at 3 T. A convolutional neural network framework in an unsupervised fashion was designed to solve an inverse problem of a two-pool MTC Bloch equation, and was compared with a conventional Bloch equation-based fitting approach. The MTC images synthesized by the convolutional neural network architecture were used for amide proton transfer and nuclear Overhauser enhancement imaging as a reference baseline image. RESULTS: The fully unsupervised learning scheme incorporated with the two-pool exchange model learned a set of unique features that can describe the MTC-MR fingerprinting input, and allowed only small amounts of unlabeled data for training. The MTC parameter values estimated by the unsupervised learning method were in excellent agreement with values estimated by the conventional Bloch fitting approach, but dramatically reduced computation time by ~1000-fold. CONCLUSION: Given the considerable time efficiency compared to conventional Bloch fitting, unsupervised learning-based MTC-MR fingerprinting could be a powerful tool for quantitative MTC and CEST/nuclear Overhauser enhancement imaging.

A deep learning approach for magnetization transfer contrast MR fingerprinting and chemical exchange saturation transfer imaging.

Semisolid magnetization transfer contrast (MTC) and chemical exchange saturation transfer (CEST) MRI based on MT phenomenon have shown potential to evaluate brain development, neurological, psychiatric, and neurodegenerative diseases. However, a qualitative MT ratio (MTR) metric commonly used in conventional MTC imaging is limited in the assessment of quantitative semisolid macromolecular proton exchange rates and concentrations. In addition, CEST signals measured by MTR asymmetry analysis are unavoidably contaminated by upfield nuclear Overhauser enhancement (NOE) signals of mobile and semisolid macromolecules. To address these issues, we developed an MTC-MR fingerprinting (MTC-MRF) technique to quantify tissue parameters, which further allows an estimation of accurate MTC signals at a certain CEST frequency offset. A pseudorandomized RF saturation scheme was used to generate unique MTC signal evolutions for different tissues and a supervised deep neural network was designed to extract tissue properties from measured MTC-MRF signals. Through detailed Bloch equation-based digital phantom and in vivo studies, we demonstrated that the MTC-MRF can quantify MTC characteristics with high accuracy and computational efficiency, compared to a conventional Bloch equation fitting approach, and provide baseline reference signals for CEST and NOE imaging. For validation, MTC-MRF images were synthesized using the tissue parameters estimated from the deep-learning method and compared with experimentally acquired MTC-MRF images as the reference standard. The proposed MTC-MRF framework can provide quantitative 3D MTC, CEST, and NOE imaging of the human brain within a clinically acceptable scan time.

A locally segmented reconstruction method for parallel imaging.

PURPOSE: A locally segmented parallel imaging reconstruction method is proposed that efficiently utilizes sensitivity distribution of multichannel receiver coil. THEORY AND METHODS: A method of locally segmenting a MR signal is introduced to maximize the differences in sensitivity between receiver channels. A 1D Fourier transformation of the undersampled k-space data is performed along the readout direction, which generates a hybrid 2D space. The hybrid space is partitioned into localized segments along the readout direction. In every localized segment, kernels representing relation between adjacent signals are estimated from autocalibration signals, and data at unsampled points are estimated using the kernels. Then, the images are reconstructed from full k-space data that consists of the sampled data and the estimated data at unsampled points. RESULTS: In a computer simulation and in vivo experiments, the locally segmented reconstruction method produced fewer residual artifacts compared to the conventional parallel imaging reconstruction methods with the same kernel geometry. The performance gain of the proposed method comes from maximizing encoding capability of receiver channels, thus resulting in the accurately estimated kernel weights that reflect the relation between adjacent signals. CONCLUSION: The proposed spatial segmentation method maximally utilizes differences in the sensitivity of receiver channels to reconstruct images with reduced artifacts.

Unsupervised learning of a deep neural network for metal artifact correction using dual-polarity readout gradients.

PURPOSE: A new unsupervised learning method was developed to correct metal artifacts in MRI using 2 distorted images obtained with dual-polarity readout gradients. METHODS: An unsupervised learning method is proposed for a deep neural network architecture consisting of a deep neural network and an MR image generation module. The architecture is trained as an end-to-end process without the use of distortion-free images or off-resonance frequency maps. The deep neural network estimates frequency-shift maps between 2 distorted images that are obtained using dual-polarity readout gradients. From the estimated frequency-shift maps and 2 distorted input images, distortion-corrected images are obtained with the MR image generation module. Experiments using synthetic data and actual MR data were performed to compare images corrected by several metal-artifact-correction methods. RESULTS: The proposed method resolved the ripple and pile-up artifacts in the reconstructed images from synthetic data and actual MR data. The results from the proposed method were comparable to those from supervised-learning methods and superior to the compared model-based method. The proposed unsupervised learning method enabled the network to be trained without labels and to be more robust than supervised learning methods, for which overfitting problems can arise when using small training data sets. CONCLUSION: Metal artifacts in the MR image were drastically corrected by the proposed unsupervised learning method. Two distorted images obtained with dual-polarity readout gradients are used as the input of the deep neural network. The proposed method can train networks without labels and does not overfit the network, even with small training data sets.

Universal High-Energy Photoelectron Emission from Nanoclusters Beyond the Atomic Limit.

Rescattering by electrons on classical trajectories is central to understand photoelectron and high-harmonic emission from isolated atoms or molecules in intense laser pulses. By controlling the cluster size and the quiver amplitude of electrons, we demonstrate how rescattering influences the energy distribution of photoelectrons emitted from noble gas nanoclusters. Our experiments reveal a universal dependence of photoelectron energy distributions on the cluster size when scaled by the field driven electron excursion, establishing a unified rescattering picture for extended systems with the known atomic dynamics as the limit of zero extension. The result is supported by molecular dynamics calculations and rationalized with a one-dimensional classical model.

Robust water-fat separation for multi-echo gradient-recalled echo sequence using convolutional neural network.

PURPOSE: To accurately separate water and fat signals for bipolar multi-echo gradient-recalled echo sequence using a convolutional neural network (CNN). METHODS: A CNN architecture was designed and trained using the relationship between multi-echo images from the bipolar multi-echo gradient-recalled echo sequence and artifact-free water-fat-separated images. The artifact-free water-fat-separated images for training the CNN were obtained from multiple signals with different TEs by using iterative decomposition of water and fat with echo asymmetry and the least-squares estimation method, in which multiple signals at different TEs were acquired using a single-echo gradient-recalled echo sequence. We also proposed a data augmentation method using a synthetic field inhomogeneity to generate multi-echo signals, including various bipolar multi-echo gradient-recalled echo artifacts so that the CNN could prevent overfitting and increase the separation accuracy. We trained the CNN using in vivo knee images and tested it using in vivo knee, head, and ankle images. RESULTS: In vivo imaging results showed that the proposed CNN could separate water-fat images accurately. Although the proposed CNN was trained using only in vivo knee images, the proposed CNN could also separate water-fat images of different imaging regions. The proposed data augmentation method could prevent overfitting even with a limited number of training data sets and make the method robust to magnetic field inhomogeneities. CONCLUSION: The proposed CNN could obtain water-fat-separated images from the multi-echo images acquired from the bipolar multi-echo gradient-recalled echo sequence, which included artifacts from the bipolar gradients.

Optimization of steady-state pulsed CEST imaging for amide proton transfer at 3T MRI.

PURPOSE: To optimize a steady-state imaging sequence for maximizing the amide proton transfer effects in pulsed-CEST (pCEST) imaging. METHOD: The steady-state pCEST (SS-pCEST) sequence is a fast CEST imaging scheme that applies repetitive short RF pulses for generating CEST and acquiring MR imaging signal alternately. To maximize the obtainable amide proton transfer effects, the SS-pCEST scheme is analyzed and optimized with respect to not only the imaging parameters but also the imaging schemes of the signal acquisition part. Three imaging parameters such as the flip angle and RF power for saturation and the flip angle for imaging are selected as factors affecting the obtainable CEST effects; and 2 imaging schemes, namely, SSFP and spoiled gradient echo sequences, are analyzed and compared for numerical simulations and MRI experiments at 3 tesla. RESULTS: SS-pCEST combined with SSFP could provide higher amide proton transfer effects than that with spoiled gradient echo. Furthermore, in the proposed SS-pCEST imaging with SSFP, 3 imaging parameters can be independently optimized so that the optimization complexities can be reduced. CONCLUSION: We optimized the SS-pCEST imaging method with SSFP to maximize the amide proton transfer effects. In addition, our analysis showed the SSFP sequence was more efficient than the spoiled gradient echo sequence for SS-pCEST imaging.

High relative-phase precision beam duplicator for mid-infrared femtosecond pulses.

In this Letter, we use a 0-π square-wave phase grating to shape 1350 nm and 1450 nm femtosecond pulses and create two intense lobes at the focus of a lens. We show that the relative phase between these two lobes (the 1st and -1st orders of diffraction of the grating) is controlled very simply and precisely by shifting the position of the grating in its plane. We generate high harmonic orders from the two bright lobes and record the beating between the two emissions for each harmonic order up to the 53rd harmonic order.

A multicontrast imaging method using steady-state free precession with alternating RF flip angles.

PURPOSE: To obtain multicontrast images including fat-suppressed contrast image, a novel multicontrast imaging method using an SSFP sequence with alternating RF flip angles is proposed. METHODS: The proposed method uses the balanced SSFP sequence with 2 flip angles. In general, the conventional balanced SSFP sequence has its own unique contrast, which combines both FID signal and echo signal under a steady-state condition. By using alternating RF flip angles and RF phase cycling, various image contrasts weighted by proton density, T1 , and T2 can be obtained. The proposed method offers multicontrast images with fat suppression by using the combination of 2 images obtained just after alternating RF pulses, respectively. RESULTS: As demonstrated by simulations, phantom and in vivo experiments, the proposed method provides multicontrast knee images including fat-suppressed contrast images. CONCLUSION: The proposed method can be a useful tool for clinical diagnosis, such as the cartilage segmentation and the fast screening of lesions.

Non-contrast-enhanced peripheral MR angiography using velocity-selective excitation.

PURPOSE: To develop a new non-contrast-enhanced peripheral MR angiography that provides a high contrast angiogram without using electrocardiography triggering and saturation radiofrequency pulses. METHODS: A velocity-selective excitation technique is used in conjunction with the golden-angle radial sampling scheme. The signal amplitude varies according to the velocity of the flow by the velocity-selective excitation technique. Because the arterial blood velocity varies depending on the cardiac phase, the acquired data can be classified into systolic and diastolic phase based on the signal amplitude of the artery. Two images are then reconstructed from the systolic and diastolic phase data, respectively, and an image reflecting the differences between the two images is obtained to eliminate background and vein signals. The performance of the proposed method was compared with the quiescent-interval single shot (QISS) in eight healthy subjects and an elderly subject. RESULTS: The proposed method generated fewer residual venous and background signals than the QISS. Furthermore, the maximum intensity projection images, the relative contrast, and the apparent contrast-to-noise ratio results showed that the proposed method produced a better contrast than the QISS. CONCLUSIONS: The proposed non-contrast-enhanced peripheral MR angiography technique can provide a high contrast angiogram without the use of electrocardiography triggering and saturation radiofrequency pulses. Magn Reson Med 79:779-788, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

AutoVOI: real-time automatic prescription of volume-of-interest for single voxel spectroscopy.

PURPOSE: To develop a fast and automated volume-of-interest (VOI) prescription pipeline (AutoVOI) for single-voxel MRS that removes the need for manual VOI placement, allows flexible VOI planning in any brain region, and enables high inter- and intra-subject consistency of VOI prescription. METHODS: AutoVOI was designed to transfer pre-defined VOIs from an atlas to the 3D anatomical data of the subject during the scan. The AutoVOI pipeline was optimized for consistency in VOI placement (precision), enhanced coverage of the targeted tissue (accuracy), and fast computation speed. The tool was evaluated against manual VOI placement using existing T1 -weighted data sets and corresponding VOI prescriptions. Finally, it was implemented on 2 scanner platforms to acquire MRS data from clinically relevant VOIs that span the cerebrum, cerebellum, and the brainstem. RESULTS: The AutoVOI pipeline includes skull stripping, non-linear registration of the atlas to the subject's brain, and computation of the VOI coordinates and angulations using a minimum oriented bounding box algorithm. When compared against manual prescription, AutoVOI showed higher intra- and inter-subject spatial consistency, as quantified by generalized Dice coefficients (GDC), lower intra- and inter-subject variability in tissue composition (gray matter, white matter, and cerebrospinal fluid) and higher or equal accuracy, as quantified by GDC of prescribed VOI with targeted tissues. High quality spectra were obtained on Siemens and Philips 3T systems from 6 automatically prescribed VOIs by the tool. CONCLUSION: Robust automatic VOI prescription is feasible and can help facilitate clinical adoption of MRS by avoiding operator dependence of manual selection.