AI as a New Frontier in Contrast Media Research: Bridging the Gap Between Contrast Media Reduction, the Contrast-Free Question and New Application Discoveries.

ABSTRACT: Artificial intelligence (AI) techniques are currently harnessed to revolutionize the domain of medical imaging. This review investigates 3 major AI-driven approaches for contrast agent management: new frontiers in contrast agent dose reduction, the contrast-free question, and new applications. By examining recent studies that use AI as a new frontier in contrast media research, we synthesize the current state of the field and provide a comprehensive understanding of the potential and limitations of AI in this context. In doing so, we show the dose limits of reducing the amount of contrast agents and demonstrate why it might not be possible to completely eliminate contrast agents in the future. In addition, we highlight potential new applications to further increase the radiologist's sensitivity at normal doses. At the same time, this review shows which network architectures provide promising approaches and reveals possible artifacts of a paired image-to-image conversion. Furthermore, current US Food and Drug Administration regulatory guidelines regarding AI/machine learning-enabled medical devices are highlighted.

Assessment of the sensitivity of 2 H MR spectroscopy measurements of [2,3-2 H2 ]fumarate metabolism for detecting tumor cell death.

Imaging the metabolism of [2,3-2 H2 ]fumarate to produce malate can be used to detect tumor cell death post-treatment. Here, we assess the sensitivity of the technique for detecting cell death by lowering the concentration of injected [2,3-2 H2 ]fumarate and by varying the extent of tumor cell death through changes in drug concentration. Mice were implanted subcutaneously with human triple negative breast cancer cells (MDA-MB-231) and injected with 0.1, 0.3, and 0.5 g/kg [2,3-2 H2 ]fumarate before and after treatment with a multivalent TRAlL-R2 agonist (MEDI3039) at 0.1, 0.4, and 0.8 mg/kg. Tumor conversion of [2,3-2 H2 ]fumarate to [2,3-2 H2 ]malate was assessed from a series of 13 spatially localized 2 H MR spectra acquired over 65 min using a pulse-acquire sequence with a 2-ms BIR4 adiabatic excitation pulse. Tumors were then excised and stained for histopathological markers of cell death: cleaved caspase 3 (CC3) and DNA damage (terminal deoxynucleotidyl transferase dUTP nick end labeling [TUNEL]). The rate of malate production and the malate/fumarate ratio plateaued at tumor fumarate concentrations of 2 mM, which were obtained with injected [2,3-2 H2 ]fumarate concentrations of 0.3 g/kg and above. Tumor malate concentration and the malate/fumarate ratio increased linearly with the extent of cell death determined histologically. At an injected [2,3-2 H2 ]fumarate concentration of 0.3 g/kg, 20% CC3 staining corresponded to a malate concentration of 0.62 mM and a malate/fumarate ratio of 0.21. Extrapolation indicated that there would be no detectable malate at 0% CC3 staining. The use of low and nontoxic fumarate concentrations and the production of [2,3-2 H2 ]malate at concentrations that are within the range that can be detected clinically suggest this technique could translate to the clinic.

Imaging Glioblastoma Response to Radiotherapy Using 2H Magnetic Resonance Spectroscopy Measurements of Fumarate Metabolism.

Early detection of tumor cell death in glioblastoma following treatment with chemoradiation has the potential to distinguish between true disease progression and pseudoprogression. Tumor cell death can be detected noninvasively in vivo by imaging the production of [2,3-2H2]malate from [2,3-2H2]fumarate using 2H magnetic resonance (MR) spectroscopic imaging. We show here that 2H MR spectroscopy and spectroscopic imaging measurements of [2,3-2H2]fumarate metabolism can detect tumor cell death in orthotopically implanted glioblastoma models within 48 hours following the completion of chemoradiation. Following the injection of [2,3-2H2]fumarate into tumor-bearing mice, production of [2,3-2H2]malate was measured in a human cell line-derived model and in radiosensitive and radioresistant patient-derived models of glioblastoma that were treated with temozolomide followed by targeted fractionated irradiation. The increase in the [2,3-2H2]malate/[2,3-2H2]fumarate signal ratio posttreatment, which correlated with histologic assessment of cell death, was a more sensitive indicator of treatment response than diffusion-weighted and contrast agent-enhanced 1H MRI measurements, which have been used clinically to detect responses of glioblastoma to chemoradiation. Overall, early detection of glioblastoma cell death using 2H MRI of malate production from fumarate could help improve the clinical evaluation of response to chemoradiation. SIGNIFICANCE: 2H magnetic resonance imaging of labeled fumarate metabolism can detect early evidence of tumor cell death following chemoradiation, meeting a clinical need to reliably detect treatment response in glioblastoma.

Deuterium MRSI of tumor cell death in vivo following oral delivery of 2 H-labeled fumarate.

PURPOSE: There is an unmet clinical need for direct and sensitive methods to detect cell death in vivo, especially with regard to monitoring tumor treatment response. We have shown previously that tumor cell death can be detected in vivo from 2 H MRS and MRSI measurements of increased [2,3-2 H2 ]malate production following intravenous injection of [2,3-2 H2 ]fumarate. We show here that cell death can be detected with similar sensitivity following oral administration of the 2 H-labeled fumarate. METHODS: Mice with subcutaneously implanted EL4 tumors were fasted for 1 h before administration (200 µl) of [2,3-2 H2 ]fumarate (2 g/kg bodyweight) via oral gavage without anesthesia. The animals were then anesthetized, and after 30 min, tumor conversion of [2,3-2 H2 ]fumarate to [2,3-2 H2 ]malate was assessed from a series of 13 2 H spectra acquired over a period of 65 min. The 2 H spectra and 2 H spectroscopic images were acquired using a surface coil before and at 48 h after treatment with a chemotherapeutic drug (etoposide, 67 mg/kg). RESULTS: The malate/fumarate signal ratio increased from 0.022 ± 0.03 before drug treatment to 0.12 ± 0.04 following treatment (p = 0.023, n = 4). Labeled malate was undetectable in spectroscopic images acquired before treatment and increased in the tumor area following treatment. The increase in the malate/fumarate signal ratio was similar to that observed previously following intravenous administration of labeled fumarate. CONCLUSION: Orally administered [2,3-2 H2 ]fumarate can be used to detect tumor cell death noninvasively following treatment with a sensitivity that is similar to that obtained with intravenous administration.

Genetic algorithm-based optimization of pulse sequences.

PURPOSE: The performance of pulse sequences in vivo can be limited by fast relaxation rates, magnetic field inhomogeneity, and nonuniform spin excitation. We describe here a method for pulse sequence optimization that uses a stochastic numerical solver that in principle is capable of finding a global optimum. The method provides a simple framework for incorporating any constraint and implementing arbitrarily complex cost functions. Efficient methods for simulating spin dynamics and incorporating frequency selectivity are also described. METHODS: Optimized pulse sequences for polarization transfer between protons and X-nuclei and excitation pulses that eliminate J-coupling modulation were evaluated experimentally using a surface coil on phantoms, and also the detection of hyperpolarized [2-13 C]lactate in vivo in the case of J-coupling modulation-free excitation. RESULTS: The optimized polarization transfer pulses improved the SNR by ~50% with a more than twofold reduction in the B1 field, and J-coupling modulation-free excitation was achieved with a more than threefold reduction in pulse length. CONCLUSION: This process could be used to optimize any pulse when there is a need to improve the uniformity and frequency selectivity of excitation as well as to design new pulses to steer the spin system to any desired achievable state.

Monitoring tumor cell death in murine tumor models using deuterium magnetic resonance spectroscopy and spectroscopic imaging.

2H magnetic resonance spectroscopic imaging has been shown recently to be a viable technique for metabolic imaging in the clinic. We show here that 2H MR spectroscopy and spectroscopic imaging measurements of [2,3-2H2]malate production from [2,3-2H2]fumarate can be used to detect tumor cell death in vivo via the production of labeled malate. Production of [2,3-2H2]malate, following injection of [2,3-2H2]fumarate (1 g/kg) into tumor-bearing mice, was measured in a murine lymphoma (EL4) treated with etoposide, and in human breast (MDA-MB-231) and colorectal (Colo205) xenografts treated with a TRAILR2 agonist, using surface-coil localized 2H MR spectroscopy at 7 T. Malate production was also imaged in EL4 tumors using a fast 2H chemical shift imaging sequence. The malate/fumarate ratio increased from 0.016 ± 0.02 to 0.16 ± 0.14 in EL4 tumors 48 h after drug treatment (P = 0.0024, n = 3), and from 0.019 ± 0.03 to 0.25 ± 0.23 in MDA-MB-231 tumors (P = 0.0001, n = 5) and from 0.016 ± 0.04 to 0.28 ± 0.26 in Colo205 tumors (P = 0.0002, n = 5) 24 h after drug treatment. These increases were correlated with increased levels of cell death measured in excised tumor sections obtained immediately after imaging. 2H MR measurements of [2,3-2H2]malate production from [2,3-2H2]fumarate provide a potentially less expensive and more sensitive method for detecting cell death in vivo than 13C MR measurements of hyperpolarized [1,4-13C2]fumarate metabolism, which have been used previously for this purpose.

Increasing the sensitivity of hyperpolarized [15 N2 ]urea detection by serial transfer of polarization to spin-coupled protons.

PURPOSE: Hyperpolarized 15 N-labeled molecules have been proposed as imaging agents for investigating tissue perfusion and pH. However, the sensitivity of direct 15 N detection is limited by the isotope's low gyromagnetic ratio. Sensitivity can be increased by transferring 15 N hyperpolarization to spin-coupled protons provided that there is not significant polarization loss during transfer. However, complete polarization transfer would limit the temporal window for imaging to the order of the proton T1 (2-3 s). To exploit the long T1 offered by storing polarization in 15 N and the higher sensitivity of 1 H detection, we have developed a pulse sequence for partial polarization transfer. METHODS: A polarization transfer pulse sequence was modified to allow partial polarization transfer, as is required for dynamic measurements, and that can be implemented with inhomogeneous B1 fields, as is often the case in vivo. The sequence was demonstrated with dynamic spectroscopy and imaging measurements with [15 N2 ]urea. RESULTS: When compared to direct 15 N detection, the sequence increased the signal-to-noise ratio (SNR) by a factor of 1.72 ± 0.25, where both experiments depleted ~20% of the hyperpolarization (>10-fold when 100% of the hyperpolarization is used). Simulations with measured cross relaxation rates showed that this sequence gave up to a 50-fold increase in urea proton polarization when compared to spontaneous polarization transfer via cross relaxation. CONCLUSION: The sequence gave an SNR increase that was close to the theoretical limit and can give a significant SNR benefit when compared to direct 13 C detection of hyperpolarized [13 C]urea.

Measuring Tumor Glycolytic Flux in Vivo by Using Fast Deuterium MRI.

Background Tumor cells frequently show high rates of aerobic glycolysis, which provides the glycolytic intermediates needed for the increased biosynthetic demands of rapid cell growth and proliferation. Existing clinical methods (fluorodeoxyglucose PET and carbon 13 MRI and spectroscopy) do not allow quantitative images of glycolytic flux. Purpose To evaluate the use of deuterium (hydrogen 2 [2H]) MR spectroscopic imaging for quantitative mapping of tumor glycolytic flux and to assess response to chemotherapy. Materials and Methods A fast three-dimensional 2H MR spectroscopic imaging pulse sequence, with a time resolution of 10 minutes, was used to image glycolytic flux in a murine tumor model after bolus injection of D-[6,6'-2H2]glucose before and 48 hours after treatment with a chemotherapeutic agent. Tumor lactate labeling, expressed as the lactate-to-water and lactate-to-glucose signal ratios, was also assessed in localized 2H MR spectra. Statistical significance was tested with a one-sided paired t test. Results 2H MR spectroscopic imaging showed heterogeneity in glycolytic flux across the tumor and an early decrease in flux following treatment with a chemotherapeutic drug. Spectroscopy measurements on five animals showed a decrease in the lactate-to-water signal ratio, from 0.33 ± 0.10 to 0.089 ± 0.039 (P = .005), and in the lactate-to-glucose ratio, from 0.27 ± 0.12 to 0.12 ± 0.06 (P = .04), following drug treatment. Conclusion Rapidly acquired deuterium (hydrogen 2) MR spectroscopic images can provide quantitative and spatially resolved measurements of glycolytic flux in tumors that can be used to assess treatment response. Published under a CC BY 4.0 license. Online supplemental material is available for this article. See also the editorial by Ouwerkerk in this issue.

Glyoxalase activity in human erythrocytes and mouse lymphoma, liver and brain probed with hyperpolarized 13C-methylglyoxal.

Methylglyoxal is a faulty metabolite. It is a ubiquitous by-product of glucose and amino acid metabolism that spontaneously reacts with proximal amino groups in proteins and nucleic acids, leading to impairment of their function. The glyoxalase pathway evolved early in phylogeny to bring about rapid catabolism of methylglyoxal, and an understanding of the role of methylglyoxal and the glyoxalases in many diseases is beginning to emerge. Metabolic processing of methylglyoxal is very rapid in vivo and thus notoriously difficult to detect and quantify. Here we show that 13C nuclei in labeled methylglyoxal can be hyperpolarized using dynamic nuclear polarization, providing 13C nuclear magnetic resonance signal enhancements in the solution state close to 5,000-fold. We demonstrate the applications of this probe of metabolism for kinetic characterization of the glyoxalase system in isolated cells as well as mouse brain, liver and lymphoma in vivo.

Photogenerated Radical in Phenylglyoxylic Acid for in Vivo Hyperpolarized 13C MR with Photosensitive Metabolic Substrates.

Whether for 13C magnetic resonance studies in chemistry, biochemistry, or biomedicine, hyperpolarization methods based on dynamic nuclear polarization (DNP) have become ubiquitous. DNP requires a source of unpaired electrons, which are commonly added to the sample to be hyperpolarized in the form of stable free radicals. Once polarized, the presence of these radicals is unwanted. These radicals can be replaced by nonpersistent radicals created by the photoirradiation of pyruvic acid (PA), which are annihilated upon dissolution or thermalization in the solid state. However, since PA is readily metabolized by most cells, its presence may be undesirable for some metabolic studies. In addition, some 13C substrates are photosensitive and therefore may degrade during the photogeneration of a PA radical, which requires ultraviolet (UV) light. We show here that the photoirradiation of phenylglyoxylic acid (PhGA) using visible light produces a nonpersistent radical that, in principle, can be used to hyperpolarize any molecule. We compare radical yields in samples containing PA and PhGA upon photoirradiation with broadband and narrowband UV-visible light sources. To demonstrate the suitability of PhGA as a radical precursor for DNP, we polarized the gluconeogenic probe 13C-dihydroxyacetone, which is UV-sensitive, using a commercial 3.35 T DNP polarizer and then injected this into a mouse and followed its metabolism in vivo.

Metabolic Imaging Detects Low Levels of Glycolytic Activity That Vary with Levels of c-Myc Expression in Patient-Derived Xenograft Models of Glioblastoma.

13C MRI of hyperpolarized [1-13C]pyruvate metabolism has been used in oncology to detect disease, investigate disease progression, and monitor response to treatment with a view to guiding treatment in individual patients. This technique has translated to the clinic with initial studies in prostate cancer. Here, we use the technique to investigate its potential uses in patients with glioblastoma (GB). We assessed the metabolism of hyperpolarized [1-13C]pyruvate in an orthotopically implanted cell line model (U87) of GB and in patient-derived tumors, where these were produced by orthotopic implantation of cells derived from different patients. Lactate labeling was higher in the U87 tumor when compared with patient-derived tumors, which displayed intertumoral heterogeneity, reflecting the intra- and intertumoral heterogeneity in the patients' tumors from which they were derived. Labeling in some patient-derived tumors could be observed before their appearance in morphologic images, whereas in other tumors it was not significantly greater than the surrounding brain. Increased lactate labeling in tumors correlated with c-Myc-driven expression of hexokinase 2, lactate dehydrogenase A, and the monocarboxylate transporters and was accompanied by increased radioresistance. Because c-Myc expression correlates with glioma grade, this study demonstrates that imaging with hyperpolarized [1-13C]pyruvate could be used clinically with patients with GB to determine disease prognosis, to detect early responses to drugs that modulate c-Myc expression, and to select tumors, and regions of tumors for increased radiotherapy dose.Significance: Metabolic imaging with hyperpolarized [1-13C]pyruvate detects low levels of c-Myc-driven glycolysis in patient-derived glioblastoma models, which, when translated to the clinic, could be used to detect occult disease, determine disease prognosis, and target radiotherapy. Cancer Res; 78(18); 5408-18. ©2018 AACR.

Dynamic 1 H imaging of hyperpolarized [1-13 C]lactate in vivo using a reverse INEPT experiment.

PURPOSE: Dynamic magnetic resonance spectroscopic imaging of hyperpolarized 13 C-labeled cell substrates has enabled the investigation of tissue metabolism in vivo. Currently observation of these hyperpolarized substrates is limited mainly to 13 C detection. We describe here an imaging pulse sequence that enables proton observation by using polarization transfer from the hyperpolarized 13 C nucleus to spin-coupled protons. METHODS: The pulse sequence transfers 13 C hyperpolarization to 1 H using a modified reverse insensitive nuclei enhanced by polarization transfer (INEPT) sequence that acquires a fully refocused echo. The resulting hyperpolarized 1 H signal is acquired using a 2D echo-planar trajectory. The efficiency of polarization transfer was investigated using simulations with and without T1 and T2 relaxation of both the 1 H and 13 C nuclei. RESULTS: Simulations showed that 1 H detection of the hyperpolarized 13 C nucleus in lactate should increase significantly the signal-to-noise ratio when compared with direct 13 C detection at 3T. However the advantage of 1 H detection is expected to disappear at higher fields. Dynamic 1 H images of hyperpolarized [1-13 C]lactate, with a spatial resolution of 1.25 × 1.25 mm2 , were acquired from a phantom injected with hyperpolarized [1-13 C]lactate and from tumors in vivo following injection of hyperpolarized [1-13 C]pyruvate. CONCLUSIONS: The sequence allows 1 H imaging of hyperpolarized 13 C-labeled substrates in vivo. Magn Reson Med 79:741-747, 2018. © 2017 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.