Quo Vadis Hyperpolarized 13C MRI?

Abstract

Over the last two decades, hyperpolarized ¹³C MRI has gained significance in both preclinical and clinical studies, hereby relying on technologies like PHIP-SAH (ParaHydrogen-Induced Polarization-Side Arm Hydrogenation), SABRE (Signal Amplification by Reversible Exchange), and dDNP (dissolution Dynamic Nuclear Polarization), with dDNP being applied in humans. A clinical dDNP polarizer has enabled studies across 24 sites, despite challenges like high cost and slow polarization. Parahydrogen-based techniques like SABRE and PHIP offer faster, more cost-efficient alternatives but require molecule-specific optimization. The focus has been on imaging metabolism of hyperpolarized probes, which requires long T₁, high polarization and rapid contrast generation. Efforts to establish novel probes, improve acquisition techniques and enhance data analysis methods including artificial intelligence are ongoing. Potential clinical value of hyperpolarized ¹³C MRI was demonstrated primarily for treatment response assessment in oncology, but also in cardiology, nephrology, hepatology and CNS characterization. In this review on biomedical hyperpolarized ¹³C MRI, we summarize important and recent advances in polarization techniques, probe development, acquisition and analysis methods as well as clinical trials. Starting from those we try to sketch a trajectory where the field of biomedical hyperpolarized ¹³C MRI might go.

Keywords: Carbon-13; DNP; Hyperpolarization; Hyperpolarized (13)C MRI; Magnetic Resonance Imaging; Molecular Imaging; PHIP; SABRE.

Figure 1

¹³C Hyperpolarization techniques and examples of preclinical studies with hyperpolarized [1-¹³C]pyruvate in mice. (a) Simplified schematic overview of dDNP, PHIP and SABRE hyperpolarization highlighting the most important steps in the polarization process. Parameter ranges for the polarization transfer do not indicate the full range of suitable conditions but rather predominantly applied regimes as well as the regimes corresponding to the images shown in b-d. (b) dDNP pyruvate and lactate projection from 3D bSSFP dataset (resolution 1.75 x 1.75 x 1.75 mm³, indicated in the image by a white voxel), 10 repetitions are averaged into the shown images, data reproduced from Nagel et al. . (c) PHIP-SAH pyruvate and lactate projection from 3D bSSFP dataset (resolution 1.75 x 1.75 x 1.75 mm³, indicated in the image by a white voxel), 10 repetitions are averaged into the shown images, data adapted from . (d) SABRE pyruvate and lactate projection from 3D bSSFP dataset (resolution 2.5 x 2.5 x 2.5 mm³, indicated in the image by a white voxel), 14 repetitions are averaged into the shown images, data adapted from . Scale bars in (b,c,d) represent 10 mm.

Figure 2

Metabolic biomarkers and enzymatic pathways accessible by hyperpolarized ¹³C-labeled probes. (a) Schematic overview of selected biomarkers in the extra- and intracellular space and related pathways that can be probed in vivo by imaging using hyperpolarized ¹³C-labeled probes. Substrates highlighted in this review are marked in black and bold and observable products are colored in black and are displayed with smaller font size. Unobservable products or contextual metabolites are colored in gray. ¹³C-bicarbonate as a substrate rather resides in the extracellular and cytosolic compartment while ¹³C-bicarbonate produced from [1-¹³C]pyruvate originates from the intramitochondrial space. (b) pH-sensitive probes such as [1,5-¹³C₂]Z-OMPD enable detection of multiple pH compartments (bottom spectra) upon injection in healthy kidneys which correspond to the anatomical regions of cortex, medulla and pelvis (top image). Image adapted with permission from . (c) [1-¹³C]DHA shows increased reduction to [1-¹³C]AA in tumors (top images) under oxidative stress, which can be monitored following treatment (bottom cluster plots). Color bar indicates voxel intensity relative to respective maxima. Image adapted with permission from . (d) Conversion of [1,4-¹³C₂]fumarate to [1,4-¹³C₂]malate indicates cell death and related release of the intramitochondrial enzyme fumarase into the extracellular space in fast growing or treated tumors. Image adapted with permission from . The schematic graphic in (a) was created with BioRender.com. CAIX: Carbonic anhydrase 9; MCT: Monocarboxylate transporter; GLUT: Glucose transporter; PEP: Phosphoenolpyruvate; G3P: Glyceraldehyde-3-phosphate; TCA: tricaboxylic acid; NADP+/NADPH: Nicotinamide adenine dinucleotide phosphate (oxidized and reduced form respectively); GSSG: Glutathione disulfide; GSH: Glutathione; AA: Ascorbic Acid; Succinyl-CoA: Succinyl-coenzyme A.

Figure 3

Sequences, acquisition advances and performance trends in clinical hyperpolarized ¹³C MRI of [1-¹³C]pyruvate. (a) Three-dimensional, whole-abdomen-covering conversion rate mapping in healthy volunteers. Image adapted with permission from . (b) Fast stack of spirals readout trajectory and combination of sequences (gradient echo & bSSFP) that allows fine-tuning and optimization of magnetization usage and acquisition speed for hyperpolarized ¹³C MRI and e.g. enables sub-millimeter in-plane resolution, here demonstrated in ALS patients (c). Image in (b) and (c) adapted with permission from and respectively. Trend analysis for dimensionality reveals increased availability of 3D volume-covering scans (d) which rely more and more on metabolite-specific acquisition techniques (e). Improvements in more efficient signal usage and faster acquisition techniques are invested in favor of improving the spatial (f) and less so the temporal resolution (g). For trend analysis, hyperpolarized clinical research studies listed in the “Human HP C13 publication list (v2, 15 September 2023) were reviewed. Consecutive multi-slice sequences were classified as “3D” and spatial resolution represents the average spatial resolution in three dimensions. Sequences were considered spectroscopic when explicitly stated or spectra were displayed. Model-based sequences all rely on IDEAL reconstruction and were explicitly stated. Metabolite-specific sequences encompass spectral-spatial, narrow bandwidth excitation approaches or other techniques explicitly stating this term. One publication could contain more than one acquisition strategy in terms of sequence type, and dimensionality and spatial and temporal resolution.

Figure 4

Kinetic models, biomarker and novel trends in analysis of hyperpolarized imaging data using [1-¹³C]pyruvate. (a) One compartment model (left) and two compartment model (right), which includes the vascular blood volume fraction v_b, the extracellular volume fraction v_e, accounting for a vascular compartment and bidirectional conversion. Figure adapted with permission from , previously adapted from . (b) Typical shape of fitted pyruvate (red) and lactate (purple) signal time curves obtained in a single-voxel. Figure inspired by . (c) Patch-based HOSVD-denoising applied to human metabolic imaging brain data. Image adapted with permission from . (d) Human metabolic imaging brain data processed without (top row) and with applied (bottom row) structural guidance super-resolution approach. Image adapted with permission from .

Figure 5

Statistics and exemplary data for clinical studies using hyperpolarized [1-¹³C]pyruvate MRI. (a) Division of patients into examined tumor organs. (b) Classification of patient data into three different assessment categories. (c). Subdivision of patients into examined tumor subtypes. (d) k_PL in a prostate lesion is able to distinguish clinically relevant (black) from an indolent (red) lesion. Image adapted with permission from . (e) k_PL increase (red) in T_2w FLAIR enhancing lesions indicates radiological progression in brain tumor patients. Image adapted with permission from . Analyzed data exclusively stems from published studies. The UCSF Department of Radiology & Biomedical Engineering published a summary of human scans worldwide, including non-published studies . (f) Rest/stress test conducted in a healthy human heart. Image adapted with permission from .