MRI Spectroscopy Could Let Doctors Watch Chemotherapy Work Inside Tumors in Real Time

For decades, one of the most frustrating blind spots in cancer medicine has been the simple question of whether a drug actually reaches the tumor it was aimed at. Interventional radiologists can thread a catheter through the arterial system and deliver chemotherapy directly into liver cancers, a procedure known as transarterial chemoembolization, or TACE. They can see the catheter, they can see the tumor, and they can see the embolic material they inject to trap the drug in place. What they cannot see is the drug itself, sitting inside the tumor, doing its work. A new proof-of-concept study from researchers at Philipps-Universität Marburg in Germany suggests that a technique already built into nearly every modern MRI scanner could change that, by detecting the chemical fingerprint of doxorubicin, one of the most widely used chemotherapy agents, directly inside living tissue.

The study, published in the journal CVIR Oncology, explored whether proton magnetic resonance spectroscopy, often abbreviated as 1H-MRS, could identify the key components of a TACE procedure in a laboratory setting. Unlike conventional MRI, which produces anatomical images based on water and fat signals, magnetic resonance spectroscopy reads the chemical composition of tissue. Every molecule containing hydrogen protons resonates at a slightly different frequency, measured in parts per million on a chemical shift scale, and the resulting spectrum acts like a molecular barcode. Clinicians already use this method to measure metabolites in the brain, where markers such as N-acetylaspartate and lactate help diagnose tumors, strokes, and metabolic disease. The Marburg team asked a bolder question: could the same technique pick out a chemotherapy drug and its delivery vehicles amid the dense chemical noise of liver tissue?

To find out, the researchers used a preclinical 7 Tesla MRI scanner, roughly double the field strength of most clinical machines, and performed two-dimensional single-voxel spectroscopy using a STEAM sequence. Each measurement sampled a tiny cube of tissue just five millimeters on each side, with an echo time of 27 milliseconds, a repetition time of 2500 milliseconds, and 128 phase-encoding steps acquired over about five and a half minutes. They tested three substances central to TACE: doxorubicin, the cytostatic drug; Embozene TANDEM drug-eluting beads, microscopic polymer spheres that soak up the drug and release it slowly inside the tumor; and Lipiodol, an oily iodinated contrast agent commonly mixed into the embolic mixture. Each substance was measured first in a clean in-vitro setup, then embedded in pieces of porcine liver to simulate the messy chemical environment of real tissue.

The results were strikingly clear for doxorubicin. The drug’s most prominent spectroscopic signal, produced jointly by a methoxy group and a cycloalkyl group on its molecular structure, appeared at 3.46 plus or minus 0.04 parts per million in solution and at 3.54 plus or minus 0.02 parts per million inside liver tissue, close to the theoretically calculated value of 3.77 parts per million. Crucially, this peak sits in a region of the spectrum that is largely free of signals from endogenous liver metabolites, meaning the drug’s fingerprint stands out against the biological background. When the researchers measured doxorubicin-loaded Embozene beads inside liver tissue, the drug signal remained clearly detectable, and the weak signals from the polymer beads themselves did not interfere with the measurement. In other words, the carrier vanished spectroscopically while the payload remained visible.

Lipiodol told a different story. The oily agent generated enormous signals, ranging from about 0.69 to 2.14 parts per million ex vivo, that dwarfed everything else in the spectrum, including the native lipid peaks of the liver itself. This overwhelming intensity completely swamped the underlying tissue signals, which cuts both ways. On one hand, it means Lipiodol is trivially easy to detect, and its distribution could in principle be mapped spectroscopically. On the other hand, any drug hidden beneath the Lipiodol signal becomes nearly impossible to measure without sophisticated spectral-fitting tools such as LCModel, which are currently designed for metabolites rather than drugs. The authors note that computed tomography already offers validated methods for quantifying Lipiodol deposition after TACE, so a spectroscopic alternative for that specific agent may be unnecessary.

Why does any of this matter? The distribution and duration of intratumoral drug presence after TACE almost certainly varies from patient to patient, and it is reasonable to assume that this variation influences treatment response. Yet no technique in clinical use today can assess drug presence or its spatial and temporal distribution inside the tumor. Researchers have relied on ex-vivo instrumental microanalysis of resected tissue, which captures a static snapshot but misses the actual metabolic processes unfolding at the treatment site. If spectroscopy could be repeated over time in the same patient, it would generate a local drug profile, revealing how long doxorubicin persists in the tumor and whether it spreads uniformly or pools unevenly. That information could support individually adapted TACE protocols, with treatment intervals and drug doses tailored to what the spectrum actually shows rather than what population averages suggest.

The study also addressed a subtle but important pharmacological question: metabolization. Doxorubicin is broken down in the body into several metabolites, most notably doxorubicinol, in which a carbonyl oxygen is hydroxylated, along with aglycones such as doxorubicinone and 7-deoxydoxorubicinone. These transformations have been extensively characterized with techniques like high-performance liquid chromatography and liquid chromatography with tandem mass spectrometry. The encouraging news for spectroscopy is that the signaling methoxy group remains unchanged in all known metabolites, so the measured spectrum should not shift as the drug is metabolized. The researchers also observed that the clinical doxorubicin preparation contained lactose monohydrate as a bulking agent, whose theoretical signal range overlaps the drug’s main peak, potentially amplifying the signal. That quirk hints the method might even extend to other powdered parenteral formulations that use similar excipients.

The authors are candid about the limitations of their setup. Signal strengths varied between measurements, likely due to temperature changes affecting molecular mobility and relaxation times, and to pH-dependent proton exchange in different chemical environments. In-vitro measurements benefit from clean, homogeneous conditions that produce sharp, well-defined peaks, while ex-vivo tissue introduces biological background noise from lipids, proteins, and endogenous metabolites, along with magnetic field inhomogeneities that broaden and distort signals. The 7 Tesla preclinical scanner, though powerful, is still weaker than dedicated nuclear magnetic resonance instruments, which partly explains small deviations between measured and calculated chemical shifts. And the leap from porcine liver in a dish to a living human patient is substantial: in vivo spectroscopy must contend with motion, breathing, and the constraints of clinical field strengths.

Still, the proof of concept lands at a moment of genuine momentum for interventional oncology. TACE is a mainstay therapy for hepatocellular carcinoma, cholangiocellular carcinoma, and certain liver metastases, tumors that are fed predominantly by arterial blood and are therefore ideal targets for transarterial delivery. The regional advantage of intra-arterial chemotherapy, amplified by embolization-induced slowing of blood flow, is well established pharmacologically, but it has never been observable at the level of the individual tumor in real time. If 1H-MRS can be validated in animal models and then in patients, it could become the first bedside tool for confirming that a chemotherapy dose actually landed where it was intended and for tracking how long it lingers. The authors caution that the quality of evaluation will depend on dedicated software tools that still need detailed analysis. But the core message is simple and potentially transformative: the drug’s voice, long drowned out by the noise of tissue, can now be heard, and listening to it may one day make liver cancer therapy a genuinely personalized procedure.

Subject of Research: Proton magnetic resonance spectroscopy for monitoring chemotherapeutic drug delivery after transarterial chemoembolization of liver tumors

Article Title: Feasibility of proton magnetic resonance spectroscopy for the assessment of chemotherapeutic drug delivery after TACE: a proof of concept

Article References: Godzinski, A., Jedelská, J., König, A. M., & Mahnken, A. H. (2025). Feasibility of proton magnetic resonance spectroscopy for the assessment of chemotherapeutic drug delivery after TACE: a proof of concept. CVIR Oncology, 1(1), Article 10. https://doi.org/10.1007/s44343-025-00010-6

Image Credits: AI Generated

DOI: 10.1007/s44343-025-00010-6

Keywords: transarterial chemoembolization, proton magnetic resonance spectroscopy, doxorubicin, liver cancer, drug delivery, drug-eluting beads, Lipiodol, interventional radiology, hepatocellular carcinoma, therapy monitoring, MRI, chemotherapy