Synthesis of higher-B0 CEST Z-spectra from lower-B0 data via deep learning and singular value decomposition.

Chemical exchange saturation transfer (CEST) MRI at 3 T suffers from low specificity due to overlapping CEST effects from multiple metabolites, while higher field strengths (B0) allow for better separation of Z-spectral "peaks," aiding signal interpretation and quantification. However, data acquisition at higher B0 is restricted by equipment access, field inhomogeneity and safety issues. Herein, we aim to synthesize higher-B0 Z-spectra from readily available data acquired with 3 T clinical scanners using a deep learning framework. Trained with simulation data using models based on Bloch-McConnell equations, this framework comprised two deep neural networks (DNNs) and a singular value decomposition (SVD) module. The first DNN identified B0 shifts in Z-spectra and aligned them to correct frequencies. After B0 correction, the lower-B0 Z-spectra were streamlined to the second DNN, casting into the key feature representations of higher-B0 Z-spectra, obtained through SVD truncation. Finally, the complete higher-B0 Z-spectra were recovered from inverse SVD, given the low-rank property of Z-spectra. This study constructed and validated two models, a phosphocreatine (PCr) model and a pseudo-in-vivo one. Each experimental dataset, including PCr phantoms, egg white phantoms, and in vivo rat brains, was sequentially acquired on a 3 T human and a 9.4 T animal scanner. Results demonstrated that the synthetic 9.4 T Z-spectra were almost identical to the experimental ground truth, showing low RMSE (0.11% ± 0.0013% for seven PCr tubes, 1.8% ± 0.2% for three egg white tubes, and 0.79% ± 0.54% for three rat slices) and high R2 (>0.99). The synthesized amide and NOE contrast maps, calculated using the Lorentzian difference, were also well matched with the experiments. Additionally, the synthesis model exhibited robustness to B0 inhomogeneities, noise, and other acquisition imperfections. In conclusion, the proposed framework enables synthesis of higher-B0 Z-spectra from lower-B0 ones, which may facilitate CEST MRI quantification and applications.

In vivo characterization of glycogen storage disease type III in a mouse model using glycoNOE MRI.

PURPOSE: Glycogen storage disease type III (GSD III) is a rare inherited metabolic disease characterized by excessive accumulation of glycogen in liver, skeletal muscle, and heart. Currently, there are no widely available noninvasive methods to assess tissue glycogen levels and disease load. Here, we use glycogen nuclear Overhauser effect (glycoNOE) MRI to quantify hepatic glycogen levels in a mouse model of GSD III. METHODS: Agl knockout mice (n = 13) and wild-type controls (n = 10) were scanned for liver glycogen content using glycoNOE MRI. All mice were fasted for 12 to 16 h before MRI scans. GlycoNOE signal was quantified by fitting the Z-spectrum using a four-pool Voigt lineshape model. Next, the fitted direct water saturation pool was removed and glycoNOE signal was estimated from the integral of the residual Z spectrum within -0.6 to -1.4 ppm. Glycogen concentration was also measured ex vivo using a biochemical assay. RESULTS: GlycoNOE MRI clearly distinguished Agl knockout mice from wild-type controls, showing a statistically significant difference in glycoNOE signals in the livers across genotypes. There was a linear correlation between glycoNOE signal and glycogen concentration determined by the biochemical assay. The obtained glycoNOE maps of mouse livers also showed higher glycogen levels in Agl knockout mice compared to wild-type mice. CONCLUSION: GlycoNOE MRI was used successfully as a noninvasive method to detect liver glycogen levels in mice, suggesting the potential of this method to be applied to assess glycogen storage diseases.

Assessing the predictability of the H3K27M status in diffuse glioma patients using frequency importance analysis on chemical exchange saturation transfer MRI.

BACKGROUND AND OBJECTIVES: In diffuse glioma patients, Lys-27-Met mutations in histone 3 genes (H3K27M) are associated with an aggravated prognosis and further decreased overall survival. By using frequency importance analysis on chemical exchange saturation transfer (CEST) MRI, this study aimed to assess the predictability of the H3K27M status in diffuse glioma patients. METHODS: Twenty-two patients diagnosed with diffuse glioma, with a known H3K27M status, were included in the present study. All patients underwent CEST MRI scans. The previously proposed frequency importance analysis was performed to determine the relative contribution of the amide and aliphatic protons for the differentiation between normal tissues and tumors. For this comparison, the conventional MTRasym analysis of amide protons at 3.5 ppm, i.e., the amide proton transfer-weighted (APTw) signal, was employed. Statistical analysis was performed using the Mann-Whitney U test, and the receiver operating characteristic (ROC) and area under the curve (AUC) analyses. RESULTS: The mean and 90th percentile of the ΔAPTw intensities, amide and aliphatic frequency importance values revealed statistically significant differences between the wildtype and the H3K27M-altered patient groups (p < 0.05). For the prediction of the H3K27M status, amide frequency importance achieved highest AUCs of 0.97, with a specificity of 0.93. In contrast, the ΔAPTw intensities and aliphatic frequency importance showed relatively lower AUCs (<0.35) in predicting the H3K27M status. CONCLUSIONS: Amide frequency importance exhibited satisfactory performance in the prediction of the H3K27M status. As such, it may be considered as a non-invasive MRI biomarker for the diagnosis of diffuse gliomas.

Hierarchical K-means clustering method for accelerated Lorentzian estimation (KALE) in chemical exchange saturation transfer-magnetic resonance imaging quantification.

Background: Quantification of in vivo chemical exchange saturation transfer (CEST) magnetic resonance signals is challenging due to contamination from coexisting effects, including the direct water effect and asymmetric magnetization transfer. Fitting-based analysis allows the calculation of multiple types of signals from the line shape of Z-spectra. However, the conventional voxelwise method has several drawbacks, including its long computation time and its susceptibility to image noise and Z-spectra oscillations, and it is difficult to determine the initial fitting parameters. Methods: Herein, we propose a K-means clustering method for accelerated Lorentzian estimation (KALE) in CEST quantification. Briefly, voxels in CEST images are clustered into K groups according to their Z-spectra characteristics. A 'groupwise' fitting process is then performed with preset initial values, yielding a set of fitted spectra and fitted parameters for each group. With the updated initial values, each group is further clustered into subgroups, and groupwise fitting is performed again. This hierarchical K-means clustering and parameter updating process continues until the pixel number or intensity error meets the termination criteria. Voxelwise fitting could be further conducted to improve the quantification images (termed voxel-K) by utilizing the previous groupwise KALE results as the initial values (termed group-K). Results: Incorporated with Lorentzian difference (LD) quantification, KALE was first optimized and evaluated on 5 healthy human brain datasets at 3 Tesla. Compared with traditional voxel-by-voxel LD quantification, the computation times of group-K and voxel-K were significantly reduced by ~85% and ~70%, respectively (P<0.001). Furthermore, the group-K images exhibited better denoising performance than traditional LD and voxel-K. KALE was further validated on six ischemic rat brains acquired at 7 Tesla, with both LD_group-K and LD_voxel-K displaying almost identical contrast maps with traditional voxelwise maps. When incorporated with the five-pool Lorentzian fitting (LF), KALE exhibited an improved contrast-to-noise ratio (CNR) for amplitude maps of each pool [P=0.003, 0.015, 0.047, and 0.047 for amide, nuclear Overhauser effect (NOE), magnetic transfer (MT) and guanidine amine, respectively] and improved fitting goodness (P=0.033). Conclusions: KALE quantification provides comparable or even superior contrast maps to traditional voxelwise fitting, with significantly reduced computation time. The 'smart' and hierarchical voxel-clustering and parameter updating process of KALE may facilitate more preclinical and clinical CEST applications.

Radiofrequency labeling strategies in chemical exchange saturation transfer MRI.

Chemical exchange saturation transfer (CEST) MRI has generated great interest for molecular imaging applications because it can image low-concentration solute molecules in vivo with enhanced sensitivity. CEST effects are detected indirectly through a reduction in the bulk water signal after repeated perturbation of the solute proton magnetization using one or more radiofrequency (RF) irradiation pulses. The parameters used for these RF pulses-frequency offset, duration, shape, strength, phase, and interpulse spacing-determine molecular specificity and detection sensitivity, thus their judicious selection is critical for successful CEST MRI scans. This review article describes the effects of applying RF pulses on spin systems and compares conventional saturation-based RF labeling with more recent excitation-based approaches that provide spectral editing capabilities for selectively detecting molecules of interest and obtaining maximal contrast.

Ultrafast Z-spectroscopic imaging in vivo at 3T using through-slice spectral encoding (TS-UFZ).

PURPOSE: Acquisition of high-resolution Z-spectra for CEST or magnetization transfer contrast (MTC) MRI requires excessive scan times. Ultrafast Z-spectroscopy (UFZ) has been proposed to address this; however, the quality of in vivo UFZ spectra has been insufficient. Here, we present a simple approach to improve this. THEORY AND METHODS: UFZ imaging acquires full Z-spectra by encoding the spectral dimension spatially via a gradient applied concurrently with the RF saturation pulse. Different from previous implementations, both this saturation gradient and its readout were applied in the slice direction, resulting in a relatively uniform voxel composition. Phase-encoding was applied in both in-plane directions, allowing additional under-sampling and acceleration. RESULTS: In phantoms, UFZ imaging with through-slice Z-spectral encoding (TS-UFZ) provided Z-spectra of salicylic acid and egg white in excellent agreement with conventional acquisitions. In vivo brain Z-spectra were influenced by flow through the imaging slice which affected the Z-spectral baseline. Still, CEST signals could be quantified after baseline fitting and mapping the residual CEST signal. Amide proton transfer (APT) contrast intensities obtained by TS-UFZ were on the same order of magnitude as conventional CEST but with different contrast across slice which likely is a result of different tissue regions contributing. CONCLUSION: TS-UFZ approach improves signal stability and spectral uniformity over previous implementations and allows high spectral-resolution imaging of saturation transfer effects in the human brain at 3T. This implementation allows for further acceleration by reducing phase encoding steps and thus opens up the possibility of mapping dynamic CEST signals in vivo with a practical temporal resolution.

The relayed nuclear Overhauser effect in magnetization transfer and chemical exchange saturation transfer MRI.

Magnetic resonance (MR) is a powerful technique for noninvasively probing molecular species in vivo but suffers from low signal sensitivity. Saturation transfer (ST) MRI approaches, including chemical exchange saturation transfer (CEST) and conventional magnetization transfer contrast (MTC), allow imaging of low-concentration molecular components with enhanced sensitivity using indirect detection via the abundant water proton pool. Several recent studies have shown the utility of chemical exchange relayed nuclear Overhauser effect (rNOE) contrast originating from nonexchangeable carbon-bound protons in mobile macromolecules in solution. In this review, we describe the mechanisms leading to the occurrence of rNOE-based signals in the water saturation spectrum (Z-spectrum), including those from mobile and immobile molecular sources and from molecular binding. While it is becoming clear that MTC is mainly an rNOE-based signal, we continue to use the classical MTC nomenclature to separate it from the rNOE signals of mobile macromolecules, which we will refer to as rNOEs. Some emerging applications of the use of rNOEs for probing macromolecular solution components such as proteins and carbohydrates in vivo or studying the binding of small substrates are discussed.

Label-Free Assessment of Mannitol Accumulation Following Osmotic Blood-Brain Barrier Opening Using Chemical Exchange Saturation Transfer Magnetic Resonance Imaging.

PURPOSE: Mannitol is a hyperosmolar agent for reducing intracranial pressure and inducing osmotic blood-brain barrier opening (OBBBO). There is a great clinical need for a non-invasive method to optimize the safety of mannitol dosing. The aim of this study was to develop a label-free Chemical Exchange Saturation Transfer (CEST)-based MRI approach for detecting intracranial accumulation of mannitol following OBBBO. METHODS: In vitro MRI was conducted to measure the CEST properties of D-mannitol of different concentrations and pH. In vivo MRI and MRS measurements were conducted on Sprague-Dawley rats using a Biospec 11.7T horizontal MRI scanner. Rats were catheterized at the internal carotid artery (ICA) and randomly grouped to receive either 1 mL or 3 mL D-mannitol. CEST MR images were acquired before and at 20 min after the infusion. RESULTS: In vitro MRI showed that mannitol has a strong, broad CEST contrast at around 0.8 ppm with a mM CEST MRI detectability. In vivo studies showed that CEST MRI could effectively detect mannitol in the brain. The low dose mannitol treatment led to OBBBO but no significant mannitol accumulation, whereas the high dose regimen resulted in both OBBBO and mannitol accumulation. The CEST MRI findings were consistent with 1H-MRS and Gd-enhanced MRI assessments. CONCLUSION: We demonstrated that CEST MRI can be used for non-invasive, label-free detection of mannitol accumulation in the brain following BBBO treatment. This method may be useful as a rapid imaging tool to optimize the dosing of mannitol-based OBBBO and improve its safety and efficacy.

Detection of electrostatic molecular binding using the water proton signal.

PURPOSE: Saturation transfer MRI has previously been used to probe molecular binding interactions with signal enhancement via the water signal. Here, we detail the relayed nuclear overhauser effect (rNOE) based mechanisms of this signal enhancement, develop a strategy of quantifying molecular binding affinity, i.e., the dissociation constant ( KD$$ {K}_D $$ ), and apply the method to detect electrostatic binding of several charged small biomolecules. Another goal was to estimate the detection limit for transient receptor-substrate binding. THEORY AND METHODS: The signal enhancement mechanism was quantitatively described by a three-step magnetization transfer model, and numerical simulations were performed to verify this theory. The binding equilibria of arginine, choline, and acetyl-choline to anionic resin were studied as a function of ligand concentration, pH, and salt content. Equilibrium dissociation constants ( KD$$ {K}_D $$ ) were determined by fitting the multiple concentration data. RESULTS: The numerical simulations indicate that the signal enhancement is sufficient to detect the molecular binding of sub-millimolar (∼100 µM) concentration ligands to low micromolar levels of molecular targets. The measured rNOE signals from arginine, choline, and acetyl-choline binding experiments show that several magnetization transfer pathways (intra-ligand rNOEs and intermolecular rNOEs) can contribute. The rNOEs that arise from molecular ionic binding were influenced by pH and salt concentration. The molecular binding strengths in terms of KD$$ {K}_{\mathrm{D}} $$ ranged from 70-160 mM for the three cations studied. CONCLUSION: The capability to use MRI to detect the transient binding of small substrates paves a pathway towards the detection of micromolar level receptor-substrate binding in vivo.

Deep learning-based classification of preclinical breast cancer tumor models using chemical exchange saturation transfer magnetic resonance imaging.

Chemical exchange saturation transfer (CEST) magnetic resonance imaging has shown promise for classifying tumors based on their aggressiveness, but CEST contrast is complicated by multiple signal sources and thus prolonged acquisition times are often required to extract the signal of interest. We investigated whether deep learning could help identify pertinent Z-spectral features for distinguishing tumor aggressiveness as well as the possibility of acquiring only the pertinent spectral regions for more efficient CEST acquisition. Human breast cancer cells, MDA-MB-231 and MCF-7, were used to establish bi-lateral tumor xenografts in mice to represent higher and lower aggressive tumors, respectively. A convolutional neural network (CNN)-based classification model, trained on simulated data, utilized Z-spectral features as input to predict labels of different tissue types, including MDA-MB-231, MCF-7, and muscle tissue. Saliency maps reported the influence of Z-spectral regions on classifying tissue types. The model was robust to noise with an accuracy of more than 91.5% for low and moderate noise levels in simulated testing data (SD of noise less than 2.0%). For in vivo CEST data acquired with a saturation pulse amplitude of 2.0 µT, the model had a superior ability to delineate tissue types compared with Lorentzian difference (LD) and magnetization transfer ratio asymmetry (MTRasym ) analysis, classifying tissues to the correct types with a mean accuracy of 85.7%, sensitivity of 81.1%, and specificity of 94.0%. The model's performance did not improve substantially when using data acquired at multiple saturation pulse amplitudes or when adding LD or MTRasym spectral features, and did not change when using saliency map-based partial or downsampled Z-spectra. This study demonstrates the potential of CNN-based classification to distinguish between different tumor types and muscle tissue, and speed up CEST acquisition protocols.

Motion correction of chemical exchange saturation transfer MRI series using robust principal component analysis (RPCA) and PCA.

BACKGROUND: Chemical exchange saturation transfer (CEST) MRI requires the acquisition of multiple saturation-weighted images and can last several minutes. Misalignments among these images, which are often due to the inevitable motion of the subject, will corrupt CEST contrast maps and result in large quantification errors. Therefore, the registration of the CEST series is critical. However, registration is challenging since common intensity-based registration algorithms may fail to differentiate CEST signals from motion artifacts. Herein, we studied how different patterns of motion affect CEST quantification and proposed a cascaded two-step registration scheme by utilizing features extracted from the entire Z-spectral image series instead of direct registration to a single image. METHODS: The proposed approach is conducted in two stages: during the first coarse registration, the Z-spectral image series is decomposed by robust principal component analysis (RPCA) to separate CEST contrast from motion. The recomposed image series using only the low-rank component, which contains minimized motion, are averaged to generate a reference for the alignment of the image series. To further remove residual misalignments, the coarse registration is followed by a refinement stage, which uses PCA iteratively to generate motionless synthetic reference series with the first few principal components (PCs) that correspond to CEST contrast. In the end, the quality check is performed to exclude the images with unsuccessful registration. RESULTS: The proposed registration scheme (RPCA + PCA_R) was assessed by both phantom experiments and in vivo data of tumor-bearing mouse brain, with simulated random rigid motion in different patterns applied to the acquired static Z-spectral image series. For comparison, previous correction schemes using an explicit image [either S0 or Ssat(∆ω)] as registration reference were also performed, named as S0_R and Ssat_R respectively. To illustrate the advantage of combination of RPCA and PCA, registration was also exploited using either only the RPCA-based method (RPCA_R) or only the PCA-based one (PCA_R). Compared with the above four methods, RPCA + PCA_R allowed for more accurate correction of the corrupted Z-spectral images, exhibiting smaller MTRasym(∆ω) error maps and lower residual Z-spectra referring to the static data. Among all the five correction methods, the corrected Z-spectral image series by RPCA + PCA_R and the resulting MTRasym(∆ω) maps achieved the highest correlation coefficients (CC) with respect to the static ones. CONCLUSIONS: The registration scheme of RPCA + PCA_R provides robust motion correction between two specific Z-spectral images and among an entire image series, through extraction of the static component from the entire Z-spectra set and inclusion of a PCA-based refinement step. Therefore, this method can help improve CEST acquisition and quantification.

Voxel-wise Optimization of Pseudo Voigt Profile (VOPVP) for Z-spectra fitting in chemical exchange saturation transfer (CEST) MRI.

BACKGROUND: Chemical exchange saturation transfer (CEST) MRI is a promising approach for detecting biochemical alterations in cancers and neurological diseases, but the quantification can be challenging. Among numerous quantification methods, Lorentzian difference (LD) is relatively simple and widely used, which employs Lorentzian line-shape as a reference to describe the direct saturation (DS) of water and takes account of difference against experimental CEST spectra data. However, LD often overestimates CEST and nuclear overhauser enhancement (NOE) effects. Specifically, for fast-exchanging CEST species require higher saturation power (B1_sat) or in the presence of strong magnetization transfer (MT) contrast, Z-spectrum appears more like a Gaussian line-shape rather than a Lorentzian line-shape. METHODS: To improve the conventional LD analysis, the present study developed and validated a novel fitting algorithm through a linear combination of Gaussian and Lorentzian function as the reference spectra, namely, Voxel-wise Optimization of Pseudo Voigt Profile (VOPVP). The experimental Z-spectra were pre-fitted with Gaussian and Lorentzian method independently, in order to determine Lorentzian proportionality coefficient (a). To further compensate for the line-shape changes under different B1_sat's, a B1-dependent adjustment was applied to the experimental Z-spectra (Z_exp) according to the prior knowledge learned from 5-pool Bloch equation-based simulations at a range of B1_sat's. Then, the obtained Z-spectra (Z_B1adj) was fitted by the previously defined VOPVP function. Considering the asymmetric component of MT, the positive- and negative-side of Z-spectra were fitted separately, while the middle part (-0.6 to 0.6 ppm, consisted primarily of DS) was fitted using Lorentzian function. Finally, the difference between Z_VOPVP and Z_exp was defined as the CEST and NOE contrast. To validate our VOPVP method, an extensive simulation of CEST Z-spectra was performed using 5-pool model and 6-pool model with greater MT component. RESULTS: In comparison with LD approach, VOPVP exhibited lower sum of squares due to error (SSE) and higher goodness of fit (R-square) for the experimental Z-spectra at all B1_sat. Moreover, the results indicated that VOPVP fitting improved the overestimated contributions from amide proton transfer (APT) and NOE through LD at all B1_sat. Despite that the relationship for B1-dependent adjustment was pre-determined using a single 5-pool model, the VOPVP fittings obtained accurate quantification for multiple 6-pool models with a range of T1w's and T2w's. The robustness of VOPVP fitting was also proved by simulations using 3T parameters. Furthermore, we assessed VOPVP in vivo in a glioblastoma-bearing mouse. Compared to LD maps, VOPVP quantification maps displayed higher contrast-to-noise ratio between tumor and normal contralateral tissue for APT, glutamate and nuclear overhauser effect (NOE), when B1_sat >1 µT. CONCLUSIONS: As an improvement of LD method, VOPVP fitting can serve as a simple, robust and more accurate approach for quantifying CEST and NOE contrast.