Purpose Glutamate weighted Chemical Exchange Saturation Transfer (GluCEST) MRI is a noninvasive technique for mapping parenchymal glutamate in the brain. Because of the sensitivity to field (B0) inhomogeneity, the total acquisition time is prolonged due to the repeated image acquisitions at several saturation offset frequencies, which can cause practical issues such as increased sensitivity to patient motions. Because GluCEST signal is derived from the small z‐spectrum difference, it often has a low signal‐to‐noise‐ratio (SNR). We proposed a novel deep learning (DL)‐based algorithm armed with wide activation neural network blocks to address both issues. Methods B0 correction based on reduced saturation offset acquisitions was performed for the positive and negative sides of the z‐spectrum separately. For each side, a separate deep residual network was trained to learn the nonlinear mapping from few CEST‐weighted images acquired at different ppm values to the one at 3 ppm (where GluCEST peaks) in the same side of the z‐spectrum. Results All DL‐based methods outperformed the “traditional” method visually and quantitatively. The wide activation blocks‐based method showed the highest performance in terms of Structural Similarity Index (SSIM) and peak signal‐to‐noise ratio (PSNR), which were 0.84 and 25dB respectively. SNR increases in regions of interest were over 8dB. Conclusion We demonstrated that the new DL‐based method can reduce the entire GluCEST imaging time by ˜50% and yield higher SNR than current state‐of‐the‐art.

No abstract

Glutamate-weighted CEST (gluCEST) imaging is nearly unique in its ability to provide non-invasive, spatially resolved measurements of glutamate in vivo. In this article, we present an improved correction for B 1 inhomogeneity of gluCEST images of the human brain. Images were obtained on a Siemens 7.0 T Terra outfitted with a single-volume transmit/32-channel receive phased array head coil. Numerical Bloch-McConnell simulations, fitting and data processing were performed using inhouse code written in MATLAB and MEX (MATLAB executable). "Calibration" gluCEST data was acquired and fit with a phenomenological functional form first described here. The resulting surfaces were used to correct experimental data in accordance with a newly developed method. Healthy volunteers of varying ages were used for both fitted "calibration" data and corrected "experimental" data. Simulations allowed us to describe the dependence of CEST at 3.0 ppm (gluCEST) on saturation B 1 using a new functional form, whose validity was confirmed by successful fitting to real human data. This functional form was used to parameterize surfaces over the space (B 1 , T 1 ), which could then be used to correct the signal from each pixel. The resulting images show less signal loss in areas of low B 1 and greater contrast than those generated using the previously published method. We demonstrate that, using this method with appropriate nominal saturation B 1 , the major limitation of correcting for B 1 inhomogeneity becomes the effective flip angle of the acquisition module, rather than inability to correct for inhomogeneous saturation. The lower limit of our correction ability with respect to both saturation and acquisition B 1 is about 40% of the nominal value. In summary, we demonstrate a more rigorous and successful approach to correcting gluCEST images for B 1 inhomogeneity. Limitations of the method and further improvements to enable correction in regions with severe pathology are discussed.

A model for glucose sensing by pancreatic β‐cells is developed and compared with the available experimental data. The model brings together mathematical representations for the activities of the glucose sensor, glucokinase, and oxidative phosphorylation. Glucokinase produces glucose 6‐phosphate (G‐6‐P) in an irreversible reaction that determines glycolytic flux. The primary products of glycolysis are NADH and pyruvate. The NADH is reoxidized and the reducing equivalents transferred to oxidative phosphorylation by the glycerol phosphate shuttle, and some of the pyruvate is oxidized by pyruvate dehydrogenase and enters the citric acid cycle. These reactions are irreversible and result in a glucose concentration–dependent reduction of the intramitochondrial NAD pool. This increases the electrochemical energy coupled to ATP synthesis and thereby the cellular energy state ([ATP]/[ADP][Pi]). ATP and Pi are 10–100 times greater than ADP, so the increase in energy state is primarily through decrease in ADP. The decrease in ADP is considered responsible for altering ion channel conductance and releasing insulin. Applied to the reported glucose concentration–dependent release of insulin by perifused islet preparations (Doliba et al. 2012), the model predicts that the dependence of insulin release on ADP is strongly cooperative with a threshold of about 30 μmol/L and a negative Hill coefficient near −5.5. The predicted cellular energy state, ADP, creatine phosphate/creatine ratio, and cytochrome c reduction, including their dependence on glucose concentration, are consistent with experimental data. The ability of the model to predict behavior consistent with experiment is an invaluable resource for understanding glucose sensing and planning experiments.

Regulation of insulin release and glucose homeostasis by pancreatic β-cells is dependent on the metabolism of glucose by glucokinase (GK) and the influence of that activity on oxidative phosphorylation. Genetic alterations that result in hyperactivity of mitochondrial glutamate dehydrogenase (GDH-1) can cause hypoglycemia-hyperammonemia following high protein meals, but the role of GDH-1 remains poorly understood. GDH-1 activity is strongly inhibited by GTP, to near zero in the absence of ADP, and cooperatively activated ( n = 2.3) by ADP. The dissociation constant for ADP is near 200 µM in vivo, but leucine and its nonmetabolized analog 2-amino-2-norbornane-carboxylic acid (BCH) can activate GDH-1 by increasing the affinity for ADP. Under physiological conditions, as [ADP] increases GDH-1 activity remains very low until ~35 µM (threshold) and then increases rapidly. A model for GDH-1 and its regulation has been combined with a previously published model for glucose sensing that coupled GK activity and oxidative phosphorylation. The combined model (GK-GDH-core) shows that GK activity determines the energy state ([ATP]/[ADP][Pi]) in β-cells for glucose concentrations > 5 mM ([ADP] < 35 µM). As glucose falls < 5 mM the [ADP]-dependent increase in GDH-1 activity prevents [ADP] from rising above ~70 µM. Thus, GDH-1 dynamically buffers β-cell energy metabolism during hypoglycemia, maintaining the energy state and the basal rate of insulin release. GDH-1 hyperactivity suppresses the normal increase in [ADP] in hypoglycemia. This leads to hypoglycemia following a high protein meal by increasing the basal rate of insulin release (β-cells) and decreasing glucagon release (α-cells). NEW & NOTEWORTHY A model of β-cell metabolism and regulation of insulin release is presented. The model integrates regulation of oxidative phosphorylation, glucokinase (GK), and glutamate dehydrogenase (GDH-1). GDH-1 is near equilibrium under physiological conditions, but the activity is inhibited by GTP. In hypoglycemia, however, GK activity is low and [ADP], a potent activator of GDH-1, increases. Reducing equivalents from GDH dynamically buffers the intramitochondrial [NADH]/[NAD], and thereby the energy state, preventing hypoglycemia-induced substrate deprivation.