The state of imaging glycolytic metabolism in cancer with magnetic resonance

Abstract

Imaging is essential for probing cancer biology and tumor surveillance in humans. Combining isotopically labeled substrates with advanced imaging approaches yields a new platform, metabolic imaging. Although cancer metabolism research began ~100 years ago, breakthroughs in magnetic resonance imaging (MRI) like hyperpolarization, deuterium metabolic imaging, and novel probes have revolutionized our ability to appreciate deregulated glycolysis in cancer. Here, we discuss the state and future of glycolytic imaging with MRI.

Imaging glycolytic activity in cancer

The Warburg effect

Nearly 100 years ago, Otto Warburg pioneered the study of cancer metabolism by discovering that unlike healthy tissues, tumors exhibit excess lactate production with increased glucose consumption^1,2. This distinct phenotype of cancer sparked numerous advancements in molecular and biological tools geared towards elucidating targetable metabolic vulnerabilities. With the development of multiple metabolically sensitive imaging techniques, this phenotype emerged as a primary aspiration for positron tomography and magnetic resonance imaging (MRI). Another major advancement in this field was the utilization of isotope tracers to image cancer metabolism. Isotope tracers are molecules with indistinguishable chemical structures from natural molecules but are composed of one or more atomic nuclei that contain the same number of protons as the natural nuclei, but a different number of neutrons³. This feature allows one to track the biodistribution of isotope tracers and their enzymatic products in a system, thus providing a tool to spatially assess metabolite import and retention, enzymatic flux, and product export when using an isotopically labeled metabolic substrate. Here, we discuss the history, current state, and future of imaging techniques designed to probe glycolytic activity in cancer.

Positron emission tomography

The first imaging applications of isotope tracers began with the use of radioisotopes. These tracers could be used in very small quantities given the extremely high sensitivity of the detection of radioactive decay. This feature was exceptionally attractive in the fact that picomolar quantities of tracer could be used in biological systems without triggering physiological effects³. Since the first biological application of radioisotopes in the 1920s⁴, isotope tracer technology was rapidly transformed with developments that mediated the generation of images. In 1949, Ansell and Rotblat generated the first radioisotope image of a patient with a retrosternal goiter by administering radioactive iodine (¹³¹I) and utilizing a manual Geiger counter to scan and detect accumulated ¹³¹I signal⁵. While this study was limited to the use of a single tracer, the development of cyclotrons allowed the production of alternative radioactive tracers. This led to the seminal development and widespread clinical application of 2-[¹⁸F]fluorodeoxyglucose-positron emission tomography (FDG-PET)^6–8, the most widely used metabolic imaging strategy at pre-clinical and clinical levels. This imaging technique directly targets the Warburg effect through the administration of a radioactive glucose derivative. More specifically, FDG probes glucose uptake, hexokinase activity, and glucose-6-phosphatase in cancer cells through the production and retention of FDG-6-phosphate^9,10. Today, FDG-PET is the gold standard of tumoral metabolic imaging and is performed ~2.2 million times a year in the US alone for clinical cancer detection and staging^11,12. Although FDG-PET has excellent sensitivity for the detection of cancer and potential for assessing treatment effects, the use of a single step radioactive probe limits this approach through constrained serial monitoring capacity and lack of sensitivity to downstream metabolism.

Magnetic resonance spectroscopy and imaging

Given the inherent limitations imposed by radioisotopes, nonradioactive imaging paradigms are emerging as potentially safer alternatives for serial examinations. The development of MRI for in vivo applications reshaped clinical care standards by establishing the capability of generating high resolution images of internal anatomy without the use of ionizing radiation. This imaging technology has become a vital tool for cancer diagnostics and monitoring at pre-clinical and clinical levels. ¹H MRI alone has achieved valuable breakthroughs, ranging from the detection of tumors through standard weighted imaging of water signals as well as the development of noninvasive metabolic imaging of cancer^10,13. In parallel with high resolution anatomical imaging, in vivo ¹H magnetic resonance spectroscopy and imaging (MRS/I) has been utilized to generate tumoral contrast by imaging metabolite signals of cancer cells^10,13–15. Images with notable metabolic contrast have been acquired in clinical cases of breast cancer where total choline is used as a biomarker to clearly distinguish cancer from neighboring healthy tissue¹⁶. Similar strategies have been utilized for glycolytic imaging of lactate signals, but these approaches are often precluded by overlapping fat signals¹⁷. Other approaches generate contrast with chemical exchange saturation transfer (CEST) MRI, which is sensitive to protons in rapid exchange with water. By saturating the frequency of the exchangeable proton, a quantifiable decrease in the water pool resonance can be measured due to saturation transfer between the two pools and the high SNR of the water signal^18,19. In the context of cancer metabolism, pre-clinical and clinical CEST has mainly been used to image glucose uptake in tumors after the administration of a bolus dose of glucose.

While ¹H MRS/I can take advantage of the high abundance of unlabeled hydrogen to generate high SNR and resolution images, it is limited in assessing kinetics. Without the use of isotope tracers, there is no straightforward way of tracking the rates and fates of the specific metabolic reactions that generate detectable signals, making ¹H MRS/I of glycolytic activity difficult to assess and clinically translate.

Hyperpolarization

To enhance the capability of MRS/I, alternative stable isotopes and magnetic resonance (MR) based approaches were developed to image cancer metabolism at pre-clinical and clinical levels. These paradigms took advantage of the development of stable isotopes composed of MR sensitive nuclei like ¹³C, ²H, and ¹⁵N. Combining novel chemistry and physics, dissolution dynamic nuclear polarization (dDNP), a prepolarization process that utilizes microwave irradiation to transfer polarization from stable radical electrons to ¹³C nuclei, achieved unprecedented success by reaching significant polarization enhancements of several orders of magnitude beyond the Boltzman state of compatible ¹³C labeled substrates^9,10,20,21. The enhanced ¹³C sensitivity drastically transformed glycolytic imaging using ¹³C labeled pyruvate. The enhanced SNR gained from dDNP allows for the acquisition of real-time metabolic flux data within the span of a few minutes, achieving comparable scan times as conventional computer tomography (CT) scans with contrast. The short duration of these experiments essentially circumvents confounding effects from a bolus administration of substrate since the timescale is not enough to sense disturbed physiology²², thus allowing ¹³C labeled pyruvate to be a “first-pass” approach used like a true tracer in human cancer patients^23,24. In addition, computational metabolic flux modeling of hyperpolarized imaging data accounting for the temporal evolution of substrate, spatial compartmentalization, and compartment specific metabolic conversion has led to the generation of flux imaging which offers greater insight through dynamic data as opposed to static images of a labeled pool size^25,26. Alongside the development of dDNP, parahydrogen induced polarization (PHIP) techniques emerged as alternative methods for hyperpolarizing chemical compounds. As first demonstrated by Bowers and Weitekamp, PHIP approaches depend on the addition of parahydrogen to asymmetric unsaturated bonds which upon breaking the symmetry of the nascent hydrogens, leads to hyperpolarization of the compound²⁷. This methodology paved the way for Aime to develop parahydrogen induced polarization-side arm hydrogenation (PHIP-SAH), which achieved hyperpolarization of a pyruvate derivative equipped with a cleavable side arm, ultimately generating hyperpolarized pyruvate after hydrogenation and cleavage²⁸. This is a chemistry-based approach that is easier to handle and less expensive than dDNP and can achieve comparable performance while generating hyperpolarized samples within minutes for pre-clinical applications. Although hyperpolarization techniques provide an avenue to image cancer metabolism beyond uptake and retention without ionizing radiation, it is limited by lack of compatible probes, expensive instrumentation, need for extensive interdisciplinary expertise, and therefore clinical availability as of now.

Deuterium metabolic imaging

Although the potential for ²H labeled substrates as imaging probes was theorized since the 1930s^29–31, it was not until the seminal work of De Feyter et al. in 2018 that the application of deuterated glucose as a pre-clinical and clinical metabolic contrast agent in cancer was demonstrated³². In this study, researchers were able to orally administer a bolus of [6,6’-²H₂]glucose, without any prior polarization, and image the production of deuterated lactate and glutamate/glutamine (glx) in a human glioblastoma. Unlike FDG-PET images, these spatial metabolite maps leveraged downstream metabolism to generate increased contrast compared to glucose maps that could not distinguish tumor from healthy brain tissue. Since then, numerous deuterated probes have been developed to image cancer metabolism to track disease and therapeutic efficacy.

Imaging glycolysis with CEST MRI

Beyond imaging glucose uptake with glucose CEST, CEST MRI has transitioned into exploring new strategies like refined processing pipelines for clinically useful images of glucose uptake in metastases, endogenous imaging of phosphorylated glycolytic intermediates, and imaging glycolytic energy reserves through the detection of glycogen. As of today, the primary application of glucose CEST is based on dynamic glucose enhanced MRI which involves pre-injection, dynamic scanning, and post-injection images after the administration of intravenous glucose¹⁸. One challenge of any CEST MRI approach is overcoming B₀ inhomogeneities which can cause considerable artifacts in CEST image acquisitions. To overcome this limitation, new strategies have employed dynamic B₀ corrections along with principal component analysis based denoising which have allowed for the detection of brain metastases in human patients (Fig. 1A)³³. Additionally, other new applications of CEST have been designed to probe alternative molecular targets like endogenous phosphorylated metabolites and glycogen stores at pre-clinical levels. Vassallo et al. (2025), demonstrated that ³¹P CEST could be used to monitor glycolysis by tracking the endogenous pool sizes of phosphorylated glycolytic metabolites like glucose-6-phosphate, phosphoenol pyruvate, glyceraldehyde-3-phosphate, and dihydroxyacetone phosphate, for example. This method involved selective saturation transfer between phosphorylated metabolites and native ATP, inorganic phosphate, and phosphocreatine pools which exchange by enzymatic activity. When tested with in vivo tumor models, in vivo ³¹P CEST detected greater glycolytic metabolite pool sizes in tumors compared to adjacent muscle tissue (Fig. 1B)³⁴. Lastly, in a similar endogenous mapping analysis as ³¹P CEST, glycogen nuclear Overhauser effect (glycoNOE) imaging was developed to image glycogen storage in vivo. This approach makes advantageous gains in sensitivity by leveraging the through-space magnetic coupling between glycogen and water to achieve higher resolution and intensity images compared to glycogen CEST (glycoCEST)^35–37. While the application of CEST strategies for glycogen detection has only been published in the context of hepatology, future work will have to explore the role of glycogen storage in tumors and how metabolic modulation through hormone or substrate stimulation may reveal valuable metabolic contrast for tumoral imaging.

Fig. 1. In vivo CEST MRI for glycolytic imaging of cancer.

A Motion and B₀ corrections with principal components analysis based denoising can generate images of glucose uptake in human brain metastases. B ³¹P CEST demonstrates the ability to distinguish tumors from surrounding healthy muscle tissue through the detection of greater pool sizes of phosphorylated glycolytic intermediates. Adapted from Wu et al. 2023, Springer © 2023 and Vassallo et al. 2025, Wiley © 2025.

Imaging glycolytic metabolism with stable isotopes

New developments in ¹³C hyperpolarization

A major direction that the field of hyperpolarization is taking is the simultaneous use of multiple probes to gain additional readouts beyond glycolytic information derived from ¹³C pyruvate^26,38–42. This is highly advantageous for glycolytic imaging since it is intimately tied to processes like redox balance and tumoral perfusion. At the forefront of quantitative redox and glycolytic pre-clinical imaging with dual hyperpolarized probes is the work of Patel et al. 2024, in which researchers utilized hyperpolarized dehydroascorbic acid (DHA) and pyruvate to image brain metabolism in a murine model³⁹. This combined probe approach revealed increased production of lactate in white matter and increased reductive and oxidative capacity in gray matter (Fig. 2A, B), thus offering valuable biological insights and foundational methodology for clinical translation. Further, the investigators were able to use combination of pyruvate labeling (C1 and C2) as well as DHA to assess flux through multiple pathways simultaneously, taking advantage of the ability of chemical shift to multiplex MRI. While this work was limited to the study of brain metabolism, it clearly has great potential to serve as a probe for dysregulated energy metabolism in cancer which is heavily implicated by glycolytic and redox arms⁴³.

Fig. 2. Co-hyperpolarized imaging probes enhance metabolic contrast.

A Metabolite images generate post injection of simultaneously hyperpolarized [1-¹³C]dehydroascorbic acid (DHA) and [1-¹³C]pyruvate in the mouse brain. B Representative in vivo spectra demonstrating the presence of lactate, vitamin C, DHA, and pyruvate. C Left to right: proton anatomical reference image of cancerous prostate gland in a human patient, k_PL rate constant image overlay, AUC_urea image overlay, and k_PL/AUC_urea image overlay demonstrating distinct contrast generated from dual hyperpolarized [1-¹³C]pyruvate and [¹³C]urea administration. Adapted from Patel et al. 2024, AAAS © 2024, and de Kouchkovsky et al. 2024, Elsevier © 2024. Abbreviations: pyruvate to lactate rate constant (k_PL), area under the curve (AUC).

Another important approach is the use of co-hyperpolarized ¹³C pyruvate and urea to simultaneously assess glycolytic fluxes and tumoral perfusion^38,44–46. de Kouchkovsky et al. 2024 demonstrated that hyperpolarized [1-¹³C]pyruvate and [¹³C]urea clinical MRI could be utilized for the detection of clinically occult prostate cancer and for distinguishing intra-tumoral metabolism and perfusion in advanced cancer⁴⁰. This methodology generated images of k_PL rate constants (s^-1), AUC_urea (AU) measurements, and the ratio of the two parameters (Fig. 2C). k_PL/AUC_urea revealed images in-line with histopathological analysis of tumor biopsies performed after imaging. These images suggest that necrotic or hypoxic areas of tumors exhibit elevated lactate production and/or retention, but reduced perfusion. This marks the unique ability of dual hyperpolarized pyruvate and urea probes to interrogate tumor areas that are typically missed by conventional multiparametric MRI.

Advancements in hyperpolarization technology have rapidly accelerated the application of this molecular imaging tool for noninvasively probing dysregulated cancer metabolism in vivo. Beyond the application of hyperpolarized pyruvate, alternative glycolytic probes like hyperpolarized glucose and fructose are paving a path to imaging glycolytic rates in vivo. While hyperpolarized glucose has been used to probe glycolysis in a variety of pre-clinical cancer models, this approach has limited clinical potential due to the considerably short T₁ of glucose⁴⁷. This limitation could be potentially overcome with the application of hyperpolarized fructose which can probe glycolysis through hexokinase activity and ultimately LDH activity. Perdeuterating ¹³C fructose and dissolving hyperpolarized sample in D₂O can achieve a T₁ greater than 90 seconds making clinical translation a possibility⁴⁸. With this in mind, a major challenge that needs to be addressed for widespread clinical translation is the technical feasibility of generating hyperpolarized samples^49–56. Today, dDNP is the most common method for generating hyperpolarized samples, but emerging techniques like PHIP may be the answer to bridging this gap. Preclinical development of PHIP polarizers has made substantial progress towards rapid and cost-effective sample preparation while generating comparable polarization enhancements as dDNP^57–59. While dDNP can require 1–3 h to generate a hyperpolarized sample, PHIP polarizers have been able to generate ready to use samples within minutes. This speed will revolutionize the field of hyperpolarized imaging by allowing rapid serial monitoring, reduced cost, and increase scanner time efficiency. To make this possible, several challenges need to be overcome. First, given that PHIP probes must be modified with a parahydrogen reactive functional group prior to polarization and cleaved before administration of the polarized metabolic substrate, further development is needed to generate a wide range of functional glycolytic probes beyond pyruvate. Second, the significant removal of components of the chemical reaction, including catalyst and solvent, will be essential to guarantee the absence of any potential confounding effects caused by the administration of residual solvent. Lastly, effective clinical translation needs to be achieved and demonstrated across various institutions and MRI scanners to validate the robustness of this approach. Nevertheless, active research in this area will certainly advance the field of glycolytic imaging in cancer.

Advancements in deuterium metabolic imaging

The breakthrough application of [6,6’-²H₂]glucose as a contrast agent for glycolytic cancer metabolism invigorated the development of alternative deuterium metabolic imaging (DMI) approaches. While [6,6’-²H₂]glucose paradigms can generate substantial metabolic contrast, detectable signal for lactate and glx are not achievable at early time points. The majority of clinical studies administer [6,6’-²H₂]glucose orally, wait for ~1–2 h, and then perform imaging on the patient. This delay in imaging is essential to allow deuterated lactate and glx to accumulate to sufficient levels for signal detection and image generation^32,60–68, but comes at the cost of lacking a straightforward manner of determining tumoral kinetics. A promising fast imaging alternative was demonstrated by Chang et al. 2025, in which intravenous administration of a perdeuterated glucose tracer, [²H₇]glucose, was utilized to generate metabolic contrast with rapid detection of deuterated water (HDO) production pre-clinically⁶⁹. Unlike [6,6’-²H₂]glucose, [²H₇]glucose can generate significantly greater HDO at early time points with tumors maximizing production within the first 5 min post tracer administration (Fig. 3A). Imaging HDO is advantageous over lactate and glx imaging, since a readily detectable natural abundance signal is already present before tracer administration⁷⁰. This allows one to immediately start tracking glucose utilization and kinetics with high SNR as opposed to waiting for signal to evolve from undetectable levels. One concern with this paradigm is the contribution of peripheral HDO production. To address this concern, the researchers demonstrated that [²H₇]glucose and D₂O injections generated completely different spatial maps of total ²H signal, indicating that glucose is handled metabolically by specific tissues whereas exogenous D₂O diffused homogenously across all tissues in the image plane (Fig. 3B). If substantial peripheral HDO was produced from [²H₇]glucose at early timepoints, then images generated with [²H₇]glucose should look much more like the images generated with D₂O. Overall, [²H₇]glucose was demonstrated to work as a metabolic contrast agent at early timepoints that could report on anti-glycolytic therapeutic efficacy and outperform [6,6’-²H₂]glucose as a fast imaging tracer in a murine flank tumor model of melanoma.

Fig. 3. Advances in deuterium metabolic imaging of glycolytic activity in cancer.

A Kinetic analysis of HDO production from [²H₇]glucose demonstrates rapid utilization and a distinct kinetic profile of tumors in comparison to healthy tissues. B Left to right: representative total ²H images (axial) after the injection of [²H₇]glucose (1.95 g/kg), [6,6’-²H₂]glucose (1.95 g/kg), and D₂O (12.5%) in flank tumor bearing mice. C Chemical shift imaging-steady state free precession (CSI-SSFP) sequence diagram. D Left to right: representative in vivo sagittal images of a proton anatomical reference, HDO, [6,6’-²H₂]glucose, and ²H-lactate acquired with CSI-SSFP at 79 minutes post injection. E [6,6’-²H₂]fructose generates greater HDO to precursor ratios over time than [6,6’-²H₂]glucose in a murine model of liver cancer. F CSI spectroscopic reconstruction of [6,6’-²H₂]fructose utilization in liver cancer. Adapted from Chang et al. 2025, AAAS © 2025; Montrazi et al. 2023, Springer Nature © 2023; Zhang et al. 2023, Wiley © 2023. Abbreviations: singly deuterated water (HDO), deuterated water (D₂O).

Another advancement in DMI is the pre-clinical application of steady-state free precession (SSFP) acquisition with chemical shift imaging (CSI) developed by Montrazi et al. 2023⁷¹. SSFP is a highly efficient sequence capable of achieving substantially high SNR per unit time compared to standard MR sequences⁷². Prior applications of SSFP sequences yielded considerable contrast and SNR benefits compared to standard CSI sequences^73,74, however the application of multi-echo SSFP (ME-SSFP) for DMI with [6,6’-²H₂]glucose failed to detect deuterated lactate signal in a pancreatitis model. To gain SNR, researchers developed a weighted average CSI-SSFP sequence (Fig. 3C) and were able to achieve 4x greater SNR than ME-SSFP with a cumulative 20x increase compared to standard CSI sequences. This increase in sensitivity mediated the acquisition of HDO, [6,6’-²H₂]glucose, and ²H-lactate images (Fig. 3D), as well as natural abundance HDO and deuterated fat images which could be used as superior internal standards compared to external phantoms which experience different electronic environments. This feature allows for more straightforward quantification without the use of additional correction factors needed in the case of phantoms⁷⁰.

Beyond glucose imaging, the field has also worked on developing alternative probes to assess glycolytic and metabolically adjacent pathways that may generate even greater contrast. One example demonstrated by Zhang et al. 2023, is the use of [6,6’-²H₂]fructose to detect liver cancer metabolism in a pre-clinical model⁷⁵. Here, investigators compared the applicability of [6,6’-²H₂]fructose as a DMI probe to the widely used [6,6’-²H₂]glucose probe. This study reports that in liver cancer, both probes detect comparable glycolytic metabolism in the TCA cycle through equivalent production of deuterated glx. However, [6,6’-²H₂]glucose detected overall greater glycolytic metabolism through higher production of ²H-lactate. Furthermore, this study observed greater HDO production from [6,6’-²H₂]fructose (Fig. 3E, F) and posited that HDO generated from fructose may be a more specific marker than HDO made from [6,6’-²H₂]glucose since fructose is primarily consumed by the liver^75–77, thus minimizing peripheral contributions that would occur from [6,6’-²H₂]glucose. Additionally, detection of glx from glucose was challenging in the normal liver due to the overlapping resonances of glucose and potential glycogen products, thus limiting its application to probe many liver-related diseases like fatty liver disease and hepatitis⁷⁵. Given the complexity of assessing downstream metabolic activity, recent efforts have led to the pre-clinical development of 2-deoxy-2-[²H₂]-d-glucose (2-DG-d2) imaging which, like FDG-PET, can generate notable contrast through substrate accumulation and retention⁷⁸. However, ²H imaging with 2-DG-d2 may be difficult to clinically translate due to the use of a substrate known to cause acute toxicity at high doses.

DMI certainly has a bright future in the field of glycolytic imaging. One major future goal of these approaches should be to reduce tracer doses. While current human doses are cost equivalent to a dose of FDG⁶⁴, physiologically, the delivery of a bolus of deuterated glucose is equivalent to that of a glucose tolerance test. Several efforts have already worked towards this goal, however improvements in radiofrequency (RF) coil technology and fast imaging sequences are essential. By increasing the sensitivity of detection through polarization transfer/indirect detection, the application of quadrature detection, volume coil transmit and surface coil receive set ups, and cryogenic technology, as well as reducing the time of a single scan, the field will gain the opportunity to substantially reduce the substrate dose and maintain or gain SNR through the acquisition of greater signal averages per unit time. An interesting potential approach for fast DMI is the application of compressed sensing. Although compressed sensing is ideally beneficial in a paradigm with excess SNR, in the case of HDO imaging it may be the key to allowing high signal averaging at considerably small timescales. Additionally, if compressed sensing could be employed along with CSI-SSFP, the field may reach unprecedented capabilities in fast DMI which when combined with computational flux modeling at a pixel-by-pixel basis will pave the way to in vivo flux imaging with deuterium.

Conclusions

Here, we have discussed advancements in imaging glycolysis and associated limitations to highlight the unique ability of molecular imaging to investigate cancer metabolism at preclinical and clinical levels. At the cutting edge of noninvasive imaging with stable isotopes are hyperpolarized MRI (HP MRI) and DMI. HP MRI can achieve remarkable metabolic contrast with high specificity for select enzymatic reactions but is generally limited by time. Given the relatively short life span of hyperpolarized ¹³C probes, HP MRI is only capable of acquiring fast kinetics, but benefits from the long T₂ relaxation times of ¹³C which mediate the ability to encode high spatial resolution images in a person. Given more time, DMI has the ability to sample slower kinetic processes, thus potentially allowing for direct measurement of glycolysis as opposed to targeted steps. However, the T₂ of ²H will ultimately limit the spatial resolution one can achieve and the long scan times required may prohibit the use of other MRI approaches (e.g. spectroscopy) to become widespread. Future developments in DMI will have to improve these constraints by implementing fast imaging approaches that can incorporate spectroscopic imaging. In addition, it is imperative that future metabolic imaging studies establish the safety of newly developed probes all while assessing physiological and metabolic responses to probe administration, comparing performance to other imaging approaches, and defining the utility of each method for assessing specific aspects like spatial resolution, signal sensitivity, and targeting local or systemic metabolism. While our perspective is limited to the discussion of glycolytic imaging paradigms, many of the principles mentioned will also apply to paradigms targeting other metabolic pathways. A cumulative effort across disciplines like chemistry, physics, engineering, and biochemistry will be essential to continue developing the next generation of imaging technology.

Author contributions

Writing—original draft: M.C.C., M.E.M., and K.R.K. Conceptualization: K.R.K. Investigation: M.C.C., M.E.M., and K.R.K. Writing—review and editing: M.C.C., M.E.M., and K.R.K. Funding acquisition: M.E.M. and K.R.K. Supervision: M.E.M. and K.R.K. Formal analysis: M.C.C., M.E.M., and K.R.K. Visualization: M.C.C., M.E.M., and K.R.K.

Data availability

No datasets were generated or analysed during the current study.

Competing interests

The authors declare the following Competing Financial Interests. K.R.K. is a founder of Atish Technologies and a member of the scientific advisory boards of Nvision Imaging Technologies, Imaginostics and Mi2. K.R.K. holds patents related to imaging and modulation of cellular metabolism. M.E.M. and M.C.C. declare no Competing Financial or non-Financial Interests.