Using magnetic resonance imaging and spectroscopy in cancer diagnostics and monitoring

Abstract

Nuclear Magnetic Resonance (MR) based imaging has become an integrated domain in today's oncology research and clinical management of cancer patients. MR is a unique imaging modality among numerous other imaging modalities by providing access to anatomical, physiological, biochemical and molecular details of tumour with excellent spatial and temporal resolutions. In this review we will cover established and investigational MR imaging (MRI) and MR spectroscopy (MRS) techniques used for cancer imaging and demonstrate wealth of information on tumour biology and clinical applications MR techniques offer for oncology research both in preclinical and clinical settings. Emphasis is given not only to the variety of information which may be obtained but also the complementary nature of the techniques. This ability to determine tumour type, grade, invasiveness, degree of hypoxia, microvacular characteristics, and metabolite phenotype, has already profoundly transformed oncology research and patient management. It is evident from the data reviewed that MR techniques will play a key role in uncovering molecular fingerprints of cancer, developing targeted treatment strategies and assessing responsiveness to treatment for personalized patient management, thereby allowing rapid translation of imaging research conclusions into the benefit of clinical oncology.

General Introduction to Imaging Techniques

Imaging is an essential component of tumor assessment both in preclinical and clinical settings. Whereas traditionally, it has provided information largely on the tumor size, location and its relationship with adjacent structures, imaging is increasingly providing information on the biological properties of the tumor. This gives an unparalleled opportunity to investigate tumor biology directly in vivo. Since modern therapeutic strategies for cancer management are aimed largely at targeting known biological properties of the tumor, imaging is now providing noninvasive information to guide both drug development and aid therapeutic management in the clinic. This review focuses on MRI and MRS techniques which are routinely used in the clinical environment and also widely used in cancer research. They are able to probe a wide range of tumor properties making them attractive for providing an in vivo biological profile of the tumor. It is important to understand how MRI and MRS fit within the spectrum of imaging modalities which are available for evaluating tumors and these are summarized with respect to their physical properties in Table 1.

This review demonstrates the wide range of MR applications available to oncology research and patient management. MR scanners are widely available today and they provide a truly integrated “hybrid” platform for in vivo imaging, i.e., MRI and MRS can be accomplished by the MR equipment with only minor hardware modifications in the same imaging session. MR techniques allow the anatomical, physiological and biochemical data to be acquired non-invasively from virtually any part of the body. In addition to the endogenous molecules exploited for imaging exogenous contrast agents are available to bring further biological targets under our scrutiny. The most recent technical developments include targeted-contrast agents1,2 and dynamic nuclear polarization (DNP) for hyperpolarizing stable isotopes.3,4 We will review the use of MR techniques in imaging multiple aspects of cancer both in preclinical and clinical settings. Our goal is several-fold: (a) to highlight the wealth of information available from cancer anatomy, physiology and metabolism by multimodal MR (Fig. 1) (b) to demonstrate the unprecedented use of MR in cancer management and (c) to project the future of cancer imaging by MR and other non-invasive imaging techniques.

MR of Cancer

It is beyond the scope of this review to explain the physics of MR and technical principles of its descendants, such as MRI and MRS,5 instead we give a short description of the very basics behind in vivo MR signals. Inherent physical properties of water protons, such as relaxation and Brownian motion, together with chemical interactions of water with macromolecules (chiefly with proteins and polypeptides, but also with lipid-containing species) and exchange between cells and subcellular compartments, provide the basis for MRI in living systems. The MRI signal can also be rendered sensitive to key physiological parameters, such as blood flow. Furthermore, the immediate magnetic environment of water protons strongly influences the signal in specific MRI scans, such as T₂ and T₂* MRI. MRI pulse sequences can be designed to probe selectively many of the chemical and/or physical properties of water, such as relaxation or diffusion, to yield the contrast into MR images.

MRS exploits the property that atomic nuclei in different electric environments resonate at slightly different frequencies. Since the electric environments are dictated by the surrounding chemical structure, the resulting resonant frequencies are a characteristic of the chemical groups and molecules present. Since the intensity of the resonance is dependent on the concentration of the metabolite, MRS can provide concentrations of a number of different metabolites non-invasively in vivo, providing a very powerful tool. There are, however, several challenges to using this technique. The resonant frequencies of the different metabolites are very close together and a homogeneous magnetic field is required to ensure that they can be resolved. Bone and air interfaces are therefore difficult regions to collect high quality data from and whole coverage of a given body part is not usually obtained. Metabolites are present in very small concentrations compared with the amount of water in tissue and so signals are very weak and low spatial resolution is obtained clinically, typically in cm³, and only the metabolites present in millimolar concentrations are detectable.

Since the early applications of MR,6 it has become apparent that MRI can provide not only exquisite detail on structure but also probe tissue properties. Today's requirements for any imaging modality for oncology applications is toward more specific molecular, biochemical and cellular details from cancer in situ for accurate diagnosis, prognosis and treatment monitoring. The surge of new multimodal MR techniques over the past decade or so has assured a rapid progress and ever increasing number of new applications of MR as illustrated in this review and elsewhere.7–9 Advanced MR techniques have already made personalized medicine a reality in cancer treatment by revealing responders and non-responders early on before the tumor volume changes enabling clinicians to adjust anticancer therapies accordingly.8,10 Elusive goals in clinical practice are being achieved, such as determining cancer prognosis from MR scan at the time of primary diagnosis and prior to surgical tissue biopsies. Below we will give details of cancer applications of several MR techniques, yet the list below is not exhaustive and we advise the readers to consult recent literature for further information.

Use of Conventional MRI in Oncology: From Diagnosis and Beyond

The concept of conventional (or standard) MRI has been commonly adopted in the literature embracing longitudinal relaxation (T₁), transverse relaxation (T₂) and proton density images. The underpinning mechanism for image intensity, and consequently, for contrast in conventional MR images is chiefly governed by relaxation processes of water protons as well as the total water content in tissue. For example, free water appears dark in T₁ (see lateral ventricles in Fig. 2A) and bright in T₂ (see lateral ventricles in Fig. 2B) images, and species with short T1 (such as white matter in brain) are bright in T₁ scans (Fig. 2A), whereas species with short T₂ (such as iron-containing basal ganglia) are dark in T₂ scans (Fig. 2B). The proton density image represents water content, as the image is acquired with minimal T₁ and T₂ weighting to eliminate signal losses due to T₁ and T₂ relaxations.

The conventional MRI techniques form the workhorse for tumor assessment in clinical radiology, but also in monitoring tumor volume in preclinical models of cancer. These images are acquired at high anatomical resolution in minutes and these factors make them attractive for tumor volume assessment. The main biological source for MR contrast between healthy tissue and tumor is the water content. An example is shown in Figure 2 for a brain tumor where tumor itself appear isointense in T₁ (Fig. 2A) but bright, with heterogeneous signal, in T₂ images (Fig. 2B). Peritumor edema gives a strong signal in T₂ images (Fig. 2B) due to high tissue water content and mobility. Similarly, cysts containing pure water-like liquid and hemorrhages within tumors are revealed with high anatomical precision.

It should be stressed that the conventional MRI techniques allow biophysical factors other than water content to be exploited for image contrast. For instance, T₂ images acquired using the so-called Car-Purcell (CP) technique render the image signal sensitive to dynamic properties of water, governed by diffusion and susceptibility dephasing. Using the CP acquisition for T₂ images one can eliminate signal from free water rendering areas of tumor MR visible where water dynamics is changing due to ongoing cell death in response to cytotoxic therapy11 (Fig. 3). A previous study indicates that the bright, the so-called dynamic dephasing signal, appears in responding glioma tissue before tumor shrinks.11

Diffusion MR for Cancer Imaging: Improved Treatment Monitoring

Brownian motion of molecules, termed as diffusion, is governed by the thermal motion of molecules. Diffusion of water in vivo can be probed for MRI contrast and since implementation into MRI in 199012 this contrast has been demonstrated to possess numerous applications, in particular in monitoring tumor treatment in vivo. Diffusion weighted imaging (DWI) both in preclinical cancer models and cancer patients has demonstrated that the value of DWI in cancer monitoring is several-fold.9,13 This is partially due to the fact that diffusion contrast sensitizes MR images, among others, to tumor cellularity and microenvironment.13–15

DWI has demonstrated heterogeneity in the apparent diffusion coefficient (ADC) within solid tumors (for an ADC image, see Fig. 2C) in numerous organs. ADC images are generated from two or more sub-images acquired with different diffusion weighting factors. Low ADC in tumor is associated with high cell density,16,17 fibrosis and other alterations in the tumor microenvironment leading to increased tortuosity in the extracellular space.18 This is understood by the restrictions imposed by cell structures, high intracellular-to-extracellular volume ratio and elevated macromolecule content in tumor per unit volume.13 The converse is also true, i.e., lowered cell density results in elevated ADC.17

Evidence is also emerging that DWI performed prior to treatment can predict progression of some cancers.15,18 In prostate cancer, ADC varies according to the location of tumor. In peripheral zone tumors, high ADC is associated with benign tumors, and thus, ADC is able to predict tumor aggressiveness.19 In locally advanced breast cancer, low tumor ADC was shown to predict good response to neoadjuvant chemotherapy.20 Using ADC as an imaging biomarker it has been shown that malignant and benign breast tumors can be differentiated with sensitivity and specificity of 89 and 77%, respectively.21 Similarly, ADC can be used to separate malignant and benign hepatic lesions.22 In addition, diffusion MRI has shown great potential for separating benign from malignant head and neck tumors and thereby guiding biopsing.23 Recent technical advancements in body MR have indicated that DWI may provide valuable information for lymph node staging,24 which has recently been possible in situ only by radionucleotide techniques, such as PET. Affected nodes show ADCs that differ from unaffected nodes. In lymphomatous nodes ADC has been found to be low, whereas metastases in nodes appear to result in high ADC values.9 Whole body DWI MRI with background suppression has been recently introduced to cancer staging.25,26 The whole body DWI MRI has been shown to be as accurate as PET with CT (PET/CT) to staging lymphomas25 and parenchymal neoplasms in the body.26 It is anticipated that whole body DWI MRI integrated with morphological MRI scans will greatly increase diagnostic accuracy of cancers by MRI.

One of the most promising applications for DWI is monitoring of treatment response to radiation and/or anticancer agents. As mentioned above, ADC is influenced by cell density and consequently, an increase in ADC after anticancer therapy may indicate response before the tumor volume begins to shrink.17,27 Chenevert et al.14 introduced the use of histogram analysis of ADC in brain tumors to predict responders. In cases with successful treatment the ADC histogram shifts to the right, in contrast to non-responders where no shift or shift to the left is evident. The approach proved to be more sensitive to detect effective treatment response than either by average tumor ADC or shrinkage of tumor.14,17 Treated with standard anticancer drugs and/or radiation ADC increases in responding tumors.15 In contrast, a recent study indicates that ischemia-like decrease in ADC to anti-angiogenic therapy with bevazicumab is detected in brain tumors.28 Rieger and coworkers observed that decreased ADC was evident in 72% of bevazicumab treated malignant gliomas as early as four weeks after introducing the drug. A biopsy from one low ADC lesion showed atypical necrosis and upregulation of hypoxia-inducible factor 1-α. These data indicate that bevazicumab by blocking neovascularisation may effectively decrease the supply of nutrients and starve the glioblastomas of oxygen and other energy substrates.

A clinically burdening issue concerns distinguishing between recurrence and treatmen-related necrosis in patient follow up. DWI may provide clinically valuable information to address this issue. It has been observed that the magnitude of ADC will be informative, low ADC in treated brain tumor is indicative of a recurrence, whereas elevated ADC reflects edema and/or necrosis due to treatment.29 The accuracy of ADC to discriminate between tumor recurrence and treatment necrosis is improved by combining it with MRS.30 However, published reports so far show no fully consistent pattern for ADC in recurrent tumors and more work is required to clarify the position for ADC in detection of tumor recurrence.31

A recently developed diffusion MRI technique, diffusion tensor imaging (DTI),32,33 provides orientation-specific information for water diffusion. DTI is now in common use to track axons and nerve fibers in the brain.33 Interestingly, recent data show that DTI can help to delineate the influence of tumors on surrounding brain parenchyma and to differentiate tumor from peritumor brain tissue.34 DTI has been shown to be very useful for planning brain tumor resections by enabling the position of fiber tracts to be visualized with respect to the tumor.35 Figure 2D gives a DTI image from a brain tumor patient. Wang and coworkers reported recently that quantitative diffusion parameters obtained by DTI allow separation of glioblastomas from brain metastates and primary lymphomas.36

Incorporation of DWI into oncology MRI protocols has prompted the need for quantitative images rather than observer interpreted film reading as traditionally performed in Radiology. In addition to quantitative images, forming so called parametric maps, which then can be analyzed using multivariate procedures, such as pattern recognition, principal component analysis, independent component analysis and neural networks.

Magnetization Transfer and Amide Proton Transfer Rate MRI

Magnetization transfer (MT) through either dipolar or non-dipolar mechanisms is a common phenomenon in vivo due to differing molecular mobilities of water protons in different physico-chemical environments.37 MT MRI contrast has been shown to be beneficial for clinical diagnosis of tumors. In brain tumors MT contrast, more specifically the so-called MT ratio (MTR), is lower than in gray or white matter, a state that is thought to reflect altered water-to-macromolecule ratio in tumor parenchyma.38,39 Preliminary MTR data indicate that it may be useful also for brain38 and breast tumor40 grading. MT has also been introduced to delineate lung tumors for radiotherapy planning.41

A special case of MT contrast is the recently introduced amide proton transfer ratio (APTR).42 APTR probes magnetization exchange between amide protons in macromolecules and bulk water protons and is accomplished by setting the off-resonance saturation on the specific resonance frequency of amide protons.42 APTR allows the assessment of protein and peptide content of tumors43 and it gains contributions from pH as well.42 APTR MRI has been applied to imaging of brain tumors indicating that this new MRI technique may be able to distinguish between tumor recurrence and treatment-related necrosis, an important clinical issue in management of glioma patients.44

In the context of MT MRI, it is worth mentioning rotating frame relaxation MRI, the T_1ρ MRI.45 Strictly speaking T_1ρ MRI is a form of relaxation governed MR contrasts, however, it is often associated with MT MRI.46 Using T_1ρ and MT MRI in combination it was observed that malignant and benign head and neck tumors can be separated with these MR contrasts with specificity exceeding 90%.46 Similarly, T_1ρ MRI has been shown to possess unprecedented sensitivity to highlight early treatment response in preclinical tumor models.47 Noninvasive information from tumors in situ afforded by MT-based and T_1ρ MR techniques is substantially adding to the overall picture depicted above for conventional and diffusion MRI.

ASL and Blood Volume MRI: Probing Hemodynamics and Vascular Reactivity

MRI offers a truly non-invasive access both into blood flow and blood volume in vivo. An MRI technique called arterial spin labeling (ASL) can be used to image and quantify tissue perfusion (∼blood flow).48,49 ASL MRI is accomplished by applying rf-pulses to the arterial blood proximal (i.e., to the heart side) to the organ under study to magnetically label the water in the blood entering the tissue. The labeled blood exchanges with tissue water at the capillary level, resulting in decrease in (T₁)-MRI signal intensity that is proportional to perfusion.49 One of the strengths of ASL MRI is that one can obtain absolute blood flow with similar accuracy as with PET, but with better spatial resolution than obtained with PET.

ASL is used chiefly in the assessment of brain and kidney tumors due to favorable hemodynamic properties in these organs for the MRI perfusion technique.50 ASL of orthotopic animal brain tumor models has shown uniformly low perfusion in tumor stroma compared with brain parenchyma.51 ASL studies of human brain tumors, however, show heterogeneous perfusion patterns.52–54 This is not unexpected in the light of known heterogeneity of oxygenation in many brain tumor types. Interestingly, ASL perfusion measurements indicate that astrocytic tumors can be graded as low or high grade tumors by means of quantitative perfusion. Low grade gliomas show lower maximum perfusion than high grade gliomas,52–54 yet the global perfusion in all glioma classes does not differ.54 High grade gliomas show higher blood flow than CNS lymphomas.52 These studies are very promising toward the use of ASL MRI for both presurgical brain tumor grading as well as for monitoring treatment responses, for instance in the case of anti-angiogenic drugs.50

Blood volume, the key index of hemodynamics, can be determined by MRI either using exogenous contrast agent (see the section below) or with a blood MR signal nulling technique, the so called VASO MRI (VAscular Space Occupancy).55 So far, VASO MRI has been used to image blood volume in brain tumors. Lu et al.56 studied 39 glioma patients with VASO MRI, where the VASO signal was acquired in a manner rendering it sensitive to both blood volume and vascular permeability. Lu et al. reported that low grade gliomas have lower blood volume than high grade ones, the discrimination accuracy of 97% was found. Interestingly, blood volume as determined by VASO MRI showed a tendency the discriminate between low and high grade gliomas. The current VASO MRI protocols allow only for partial coverage of the whole brain, but nevertheless the VASO data, as part of multimodal MR assessment of brain tumors, are promising for presurgical evaluation of brain cancers.

DSC and DCE MRI: Revealing Quantitative Properties of Vasculature

Properties of tumor vasculature, including microperfusion, vascular permeability, vessel density and vessel size, are crucial indices of neovascularisation (or angiogenesis) in cancer tissue.57 MRI provides a window to image these indices and thus, angiogenesis in cancer in vivo.58–60 The techniques used to this end require injection of exogenous contrast agents that are commonly used in clinical imaging with MR data acquisition accomplished using fast imaging techniques, such as echo planar imaging (EPI). The two MRI techniques, the so-called dynamic susceptibility contrast (DSC)61 and dynamic contrast enhanced (DCE)62 MRI are targeted to quantify properties of microperfusion and vasculature. DSC is commonly accomplished by acquiring MRI signal by an apparent transverse relaxation (T₂*) MRI technique61 in rapid succession following rapid iv bolus of Gd-based contrast agent. The passage of the bolus leads to a transient decrease in T₂* signal, leakage of the contrast agent results in persistently lowered signal that may decrease further as a function of time. The hemodynamic parameters derived from DSC MRI include (relative) blood flow, (relative) blood volume and mean transit time.61

DCE MRI62 involves acquisition of MRI signal using a T₁-weigthed MRI acquired either with EPI or rapid gradient echo techniques, passage of Gd-bolus results in a transient signal increase in tissue with mature (healthy) microvasculature, but in tumors with immature and leaky vasculature a persistent signal increase results. From the DCE signal one can derive quantities describing blood-brain-barrier transfer constant, the microvascular permeability (capillary wall permeability-surface area product), size of extravascular-extracellular space, and capillary blood flow.62 Figure 4 illustrates a glioblastoma multiforme case imaged with DCE MRI with quantitative maps generated for microvascular and microenvironmental parameters.

DSC and DCE MRI have an unprecedented position in evaluation of vascular properties and hemodynamics in tumors in vivo. For instance, in brain lesions the degree of vascular leakiness is an important imaging surrogate in assessment of the nature of the lesion. Malignant lesions show a high degree of leakiness due to rapid angiogenesis with immature microvasculature, whereas benign lesions show vascular leakiness less commonly.62 Similarly, it is a common observation that cancer recurrence shows a larger degree of leakiness than treatment related necroses, but it should be borne in minds that contrast enhanced MRI and assessment of microvascular properties by MRI alone may not be accurate enough to allow discrimination of recurrence and treatment-related necrosis. Nevertheless, using advanced analytical routines for DSC MRI, Galban et al.63 reported that relative blood volume was reduced in patients with progressive disease despite radiation and anticancer therapy relative to lesions with successful treatment response.

Owing to the wide availability and sensitivity of DCE and DSC MRI to neovascular pathologies in tumors, these imaging techniques have become work horses for indirect assessment of angiogenesis in many cancer types.8,60 At the same time DCE and DSC MRI have gained wide use in evaluating response in tumors in situ to anti-vascular agents, involving both anti-vascular endothelial growth factor (VEGF) and vascular disruption agents.64,65 The strength of MRI techniques lays in the multitude of information obtained from tumors by MRI. Quantitative hemodynamic and/or microvasular data obtained by DCE MRI from breast,66 renal67 and head and neck tumors68 has shown to provide prognostic information prior to therapy. It has been observed that while anti-VEGF treatment decreases vascular permeability and blood volume, indicating normalization of tumor vasculature, edema in and around tumor also decreases.65 An example of bevazicumab treatment response was given above, suggesting that this anti-angiogenic drug can lead to an ischemia-like state in responding gliomas.29

MRI can also be used to estimate vessel size index (VSI, sometimes referred also to as vessel caliber index) using either endogenous69 or exogenous contrast agents.70,71 VSI for cancer imaging exploits contrast agent injection to measure the average diameter of microvasculature in vivo. At the same time MRI allows for the estimation of vessel density.72 For VSI MRI data are collected from tissue following injection of either iron oxide- or Gd-based contrast agents.71 The former contrast agents have long plasma half lives, but are not yet commonly licensed for clinical indications. Instead, Gd-based contrast agents are in common clinical use and can thus be used for VSI both in clinical and preclinical settings.73 VSI provides an excellent measurement for microvascular morphology, however, VSI MRI may not give an accurate enough measure for microvascular morphology.74 Using VSI MRI in a preclinical colorectal cancer model it was shown that anti-VEGF and anti-neurophilin-1 drugs both reduce VSI and vessel density.72 In a murine glioma model VSI increased, while blood volume and vascular permeability decreased in response to cediranib treatment,73 which is a VEGF-inhibitor. Instead, cediranib treatment of human gliomas leads to decrease in VSI (see ref. 73 for references), with reduced blood volume and vascular permeability underscoring the importance of using appropriate references for treatment response evaluation.

DSC and DCE MRI are unparalled tools for imaging tumor vascular characteristics and angiogenesis. They provide a major advance for individualized treatment response monitoring, with numerous key parameters revealed in vivo for diagnosis and treatment response assessment.

Tumor Hypoxia Probed by MRI

MRI and MRS provide indirect ways of assessing tumor oxygenation. Above it was described how ASL, VASO, DSC and DCE allow the derivation of hemodynamic variables. Of course, perfusion of tissue is the key factor supplying oxygen (and other substrates) for cellular metabolism in vivo. Blood oxygenation level dependent (BOLD) MRI contrast75 is generated by a change in local deoxy-hemoglobin concentration (Hb). The (Hb)/(HbO₂) ratio will reflect the balance of oxygen delivery (∼perfusion) and consumption (∼mitochondrial respiration)76,77 Hb is a paramagnetic species causing dephasing of the transverse magnetization. As a result, both T₂* and T₂ in parenchyma shorten in response to increased (Hb)/(HbO₂) ratio, the former being more sensitive than the latter due to inherent physical difference between these MR variables.

BOLD contrast has been examined as a potential means to indirectly evaluate changes in tumor oxygenation in vivo.78 The BOLD signal gains complex contributions from blood oxygenation and blood volume and their contributions vary as a function of magnetic field strength and image acquisition parameters, such as echo time.79 These factors together with inherent variation in intratumor T₂*/T₂ signal due to edema and increased water permeability of tumor vasculature have hampered quantification of tumor oxygenation in absolute terms from the BOLD signal.78 To measure absolute partial pressure of O₂ in tissue (ptO₂) one would need either invasive techniques, MR contrast agents80 or electron paramagnetic resonance (EPR).81

Several studies have been undertaken to evaluate tumor oxygenation with BOLD MRI. These studies have commonly exploited systemic respiratory challenges by allowing the subjects to breathe either pure O₂ or carbogen which is composed of 95–98% O₂ and 2–5% CO₂82–84 Inspiring 100% O₂ will maximally saturate HbO₂ and also increase dissolved O₂ in blood plasma. Inspiring carbogen not only maximally saturates HbO₂, but also induces vasodilation and increased blood flow due to effects of CO₂. Owing to noninvasiveness of the BOLD MRI and good tolerance of respiratory challenges used in preclinical settings, substantial efforts have been directed to human applications of these techniques.85 In addition to these factors the on-going transition from 1.5–3 T for standard human imaging makes the exploitation of BOLD contrast for tumor imaging more attractive.86 The largest body of BOLD studies in human cancers is in CNS tumors. This is understandable in many regards, as glioblastomas are one of the most hypoxic and at the same time most fatal human cancers and therefore, a commonly available imaging technique for pretreatment assessment of brain tumor hypoxia would be invaluable.57 Studies84,86 indicate that gliomas show heterogeneous T₂* signal changes to carbogen challenge. The heterogeneity has been taken to indicate the inherent patchy nature of perfusion and/or vascular reactivity in these tumors; areas showing no response to carbogen are considered to be severely hypoxic. In a study on meningiomas, it was observed that BOLD response to carbogen was seen only in areas with leakage of injected Gd-based contrast agent, demonstrating that BOLD response is an index of vascular reactivity (and thus, perfusion in the tumor stroma).82 The BOLD MRI data from respiratory challenges are promising pointing to an imminent potential of BOLD MRI in monitoring tumor vasoreactivity and oxygenation in a clinical setting. In animal models drugs have also been used to cause vasodilation, such as nicotinamide.78 The conclusions from this work is that well perfused and therefore oxygenated tumor regions can be imaged, but any quantitative data for ptO₂, in a similar manner as obtained by EPR, is difficult to obtain.78

Recently, a combined MRI/EPR imaging scanner was constructed, making it possible to obtain concurrent MR images and ¹H MR spectra from a tumor with quantitative ptO₂ images by EPR.87 This hybrid imaging scanner was applied to address the interrelationships between tumor perfusion, lactate concentration and oxygenation in a subcutaneous SCC tumor. The data showed surprisingly that substantial hypoxia was present in well perfused parts of the tumor and that high levels of lactate was found in tumor tissue with normal ptO₂. This type of scanner is currently available only for preclinical applications.

²³Na MRI in oncology

²³Na is a quadrupolar MR nucleus which has been exploited also for in vivo MRI in oncology. In breast tumors, as revealed by contrast enhanced MRI, tissue sodium concentration was found to be elevated by over 60% relative to the glandular tissue when the tumor was malignant.88 Bartha and associates quantified ¹H MRS metabolites and tissue sodium by ²³Na MRI in low grade gliomas.89 They observed that sodium was elevated in gliomas relative to healthy brain parenchyma and that sodium MR data improved delineation of glioma tissue beyond that by MRS. Rat 9L gliomas have been studied by ²³Na MRI during nitrosourea treatment.90 It has been observed that ²³Na signal and ADC changes occur in responders well before tumors started to shrink.90 In summary, ²³Na MRI provides complementary information to the ¹H MRI and may prove to be useful for multi-nuclear imaging of cancer, but more work is required to consolidate its position in clinics.

Multinuclear MRS for Cancer Imaging: ¹H MRS

¹H MRS has provided a wealth of information both clinically and in pre-clinical models. While in principle, any atomic nucleus with a magnetic moment can be used for MRS, to date almost all the clinical data has been collected using ¹H MRS owing to its high inherent sensitivity. The most studied anatomical region is the brain, where N-acetyl aspartate, total choline, creatine, myo-inositol, lactate, taurine, glutamate+glutamine, lipids and macromolecules resonate, are commonly measured. More sophisticated analysis and increasing field strength can provide information on additional metabolites. High levels of total choline (for Cholineto-creatine ratio map in a brain tumor see Fig. 2F) and its ratio to N-acetyl aspartate is a hallmark of most tumors and has been used to distinguish brain tumors from other lesions.91,92 Due to inherent high glycolytic activity (i.e., the Warburg effect) several tumor types have high lactate concentration independent of the presence of hypoxia,93 in fact, ¹H MRS data from a preclinical glioma points to production of lactate in well oxygenated part of tumor.94

¹H MRS has also been used to identify the most active regions of large complex tumors thereby aiding tumor biopsy95,96 and for delineating tumor margins which can extend beyond the enhancing regions seen on conventional MRI.95 The entire MRS profile has been shown to be a strong characteristic of tumor type and has been studied extensively as an aid to non-invasive diagnosis, the best results coming from the use of pattern recognition97–100 Large multi-center studies have shown that this can be robust even when evaluated prospectively101 and clinical decision support systems based on MRS have been developed.102,103 High total choline and mobile lipids together with low myo-inositol (for an example brain tumor ¹H MRS see Fig. 2) have been identified as markers of high grade particularly in gliomas but also in other CNS tumors.104,105 More detailed analysis of ¹H MRS has revealed other important markers of tumor aggressiveness, such as glycine.106 Differences have also been found in the primary tumors of medulloblastomas with and without metastatic disease at diagnosis.107 High total choline and lipids have been identified as biomarkers of poor survival in pediatric tumors.108 Biomarkers of early response to treatment are also important to identify and there is some indication that MRS can provide this under some circumstances,109 however, the promising results seen in animal and cell line studies have yet to be mirrored in the clinical arena.

Prostate cancer is a disease in which ¹H MRS is being used increasingly to identify regions at high risk of being malignant. Identifying these regions on conventional imaging is difficult and multiple biopsies are usually taken. High resolution, 3-dimensional (3D), multivoxel MRS can give good coverage of the prostate, commonly with an endorectal coil but increasingly with coil arrays placed on the abdomen. Choline and creatine may not be resolvable in a conventional clinical scanner, but a high combined total choline+creatine/citrate ratio is a good correlate of malignancy and can be used to identify the best regions to biopsy.110 In comparison with histopathology, up to 91% specificity and 95% sensitivity was achieved when MRI was combined with 3D MRSI.111 The Gleeson score obtained from multiple prostate biopsies together with prostate specific antigen levels in the blood are used to determine clinical management such as the prospect of surgical cure112 and several studies have now shown a correlation between the MRS parameters and Gleeson score.113,114 Furthermore the magnitude of the change in total choline+creatine/citrate correlates with the how aggressive the tumors are.115 With this success, attention is increasingly turning to performing clinical trials which can evaluate the use and predictive nature of these biomarkers.116

In breast cancer, response of neoadjuvant chemotherapy is an important indicator of survival and MRS has been investigated as a marker of early response.117–119 Total choline is high in breast cancer and reduces in patients with MRI and histopathological response.120 Furthermore, a reduction of total choline between pre-treatment and within 24 h of the first chemotherapy dose correlates significantly with changes in tumor size.121,122

¹H MRS provides direct access to key aspects of tumor hypoxia. Lactate, the end product of anaerobic glycolysis, is ready detected, however, as stated above interpretation of the presence of lactate in tumor is not unproblematic. Lactate is produced by oxygenated tumor cells in vivo94 and hence, lactate is found by ¹H MRS with non-hypoxic ptO₂.87 Interestingly, ¹H MRS detectable lipids123 may aid in defining hypoxic tumor stroma.124¹H MRS detects lipids present in lipid vesicles.123 In a C6 glioma, displaying strong ¹H MRS lipid signals, lipid vesicles were found in histologically verified hypoxic tumor stroma.124 However, ¹H MRS lipids are present in necrotic tumors with no other detectable metabolites125 and therefore, presence of other metabolites, such as cholines and lactate, may be indicative to hypoxia in viable tumor stroma rather than necrosis.

³¹P MRS.

³¹P MRS was used to perform many of the earliest clinical MRS studies,126 but has become relatively neglected and at present is only used in clinical centers with a specific interest. It has the advantage over ¹H MRS that the individual choline metabolites, phosphocholine and glycerophosphocholine, can be separately detected and these are more robust markers of tumor aggressiveness than the total choline. The studies which have been performed have therefore tended to concentrate on MRS as a biomarker of response rather than diagnosis.127,128 ATP, ADP and inorganic phosphate can also be detected and used to give a non-invasive measurement of pH and this can be a good indicator of tumor activity and response.7³¹P MRS suffers from poor sensitivity and voxel volumes of around 27 cm³ in clinical settings are required. Technical advances in this pursuit such as the availability of higher field strength scanners and the development of new coils may lead to further clinical applications.

¹³C MRS.

While MRS is only sensitive enough to identify the metabolites which are present in the highest of concentrations, there has been much interest in imaging molecules which have been injected into the patient. ¹³C labeled molecules can be detected with MRS and there have been many studies investigating ¹³C labeled glucose and its metabolites.129 The method has the advantage over PET that that the parent molecule and its metabolites can be identified separately. A major advance has come in this field with the use of DNP in which the labeled molecules attain a very high magnetic moment in a process where they are cooled near to the absolute zero.3 This increases the sensitivity of detection by 10,000-fold and allows detection of small amounts of tracer also by a clinical scanner. Impressive experiments using ¹³C labeled pyruvate in animal systems shows the metabolism and/or exchange to lactate in real time and shows that this is altered in cancer cells.4 Disordered energy metabolism is a hallmark of cancer cells and this technique offers a direct means of measuring it in vivo. Animal models have also been studied by ¹³C labeled fumarate which gives specific information on necrosis130 and ¹³C labeled bicarbonate which gives information on pH.131 These techniques have high potential and clinical studies are already underway using ¹³C labeled pyruvate to investigate prostate cancer.

¹⁹F MRS.

¹⁹F is an inherently sensitive MR nucleus being only slightly inferior in sensitivity to ¹H. ¹⁹F concentration in vivo is very low so that there is no background MR signal in vivo. These two factors have been turned to advantage in MR detection of exogenous ¹⁹F containing compounds, such as indicators for tissue oxygenation and anticancer drugs. Several ¹⁹F tagged compounds have been introduced to quantify ptO₂ as indicated above.80¹⁹F MRS has proven to be a unique tool to monitor uptake, metabolism and wash-out of the anticancer drug 5-fluorouracil (5-FU). 5-FU uptake kinetics has been examined by ¹⁹F MRS in Lewis lung carcinoma132 and RIF-1 tumors.133 It was shown in the RIF-1 tumors that inhalation of 5% CO₂ improved uptake of 5-FU and also treatment response to the drug.133 5-FU is converted into cytotoxic metabolites in target tissue, process that can be visualized by ¹⁹F MRS.132¹⁹F MRS has been used clinically to monitor 5-FU uptake by liver metastases134 and to guide dose escalation of 5-FU for optimal therapeutic effect.135 It was observed that increased uptake by liver metastases correlated with clinical response.134¹⁹F MR provides access to tissue oxygenation and guides drug assessment, both applications with potential clinical impact.

MRS of cancer specimens in vitro.

As well as in vivo imaging of metabolites, MRS can also be used to obtain metabolite profiles of whole tissue, extracts of tissue and body fluids. High resolution magic angle spinning MR provides excellent metabolite identification of tissue samples as small as 5 mg while keeping them intact for further experiments. This has been used to validate in vivo MRS results136 and obtain a more detailed metabolite profile of tumor tissue in both patients and animals thereby delineating the metabolic pathways in more detail.137–141 The technique has also been used for investigating intact cells from cell culture experiments. These approaches have established the importance of phosphocholine, glycerophosphocholine, glutamate, glutamine, glycine and lactate individually as biomarkers of tumor response to treatment across a range of tumor cells and treatment approaches.142–144 However, no single metabolite is a universal marker of tumor response and attention has turned to using combinations of metabolites for which there is emerging evidence that metabolite profiles exist which are specific to certain types of cell stress and cell death.142 Studies of animal tissue and whole cells have also shown that polyunsaturated fatty acids are an early marker of cell stress and cell death but have proven difficult to image clinically.145

Molecular Imaging with MRI in Cancer Imaging

To complement the multitude of information available from endogenous MR contrasts as well as to increase the molecular specificity of MRI signals in vivo, targeted contrast agents tagged with an MRI detectable moiety have been introduced. The targeted contrast agents have greatly increased molecular specificity of biological targets to be imaged by MRI, such as receptors146 and products of transgenes.147,148 Most of the contrast agents offer good specificity to the given target by enhancing either T₁ and/or T₂ relaxation. The former agents provide positive contrast in T₁-images, and the latter ones negative contrast in T₂-images.

Since the revelation of transgene expression by targeted MRI contrast agents147 bespoke contrast agents have been used to reveal several aspects of cancer in vivo. Zhao et al.149 generated a synaptotagmin I-based contrast reagent to detect externalized phosphatidyl-serine (PS) in apoptotic lymphomas by MRI. The same group has recently modified the reagent to increase its affinity to PS and to give positive MRI contrast with a view to potential clinical application.150 Figure 5 illustrates treatment response in a murine EL-4 lymphoma model to etoposide+cyclophosphamide as revealed with a PS-specific MR molecular contrast agent. Strijkers et al. have developed multi-modal contrast agents for direct detection of angiogenesis activity in tumors.151 Artemov and coworkers imaged HER-2/neu receptors with MRI in a preclinical breast cancer model.146 Hyaluronidase activity was detected by MRI in an ovarian tumor model to assess peritumor angiogenic balance.152 Until today, targeted contrast agents used for molecular imaging by MRI have only been used in preclinical applications, however, the proof of principle studies in preclinical settings have opened avenues for novel targets for MRI in the clinical assessment of cancer patients.

Collective Picture of MR in Monitoring Cancer

Advancement in imaging has fundamentally changed our way of visualizing cancer in vivo. MR techniques have been the key players in this pursuit and there is continuing evolution of new MR techniques and applications. The wealth of information provided by MRI and MRS on tumor biology is having an increasing clinical impact. Today's images from cancer, such as those generated from multi-modal MR scans, are viewed as multi-parametric quantitative maps rather than just individual qualitative films and they are revealing cancer in vivo in a new comprehensive manner. With the shift toward reliance on quantitative parameters, there are very real challenges to clinical application. Studies with impressive results are invariably from a single center with considerable MR research experience and expertise. In order to have a broad clinical impact, the same investigation must provide the same result in any center, irrespective of the level of research expertise. It is emerging that this requires a complex blend of adherence to common acquisition protocols, strict quality control and robust processing and analysis methods. Where these have been achieved, multi-center results have been similar to those from single centers, however, translating this to a non-research clinical environment remains a significant challenge. Achieving this goal, in particular for some of the more innovative techniques, will require a significant level of investment and cooperation from the scanner manufacturers as well as robust evidence from mutlicenter clinical trials. Despite these challenges, compelling evidence is massing that the increasing availability of these imaging modalities is leading to radical changes in the thinking of cancer biology by sampling it directly in vivo.

Abbreviations

ADC	=	apparent diffusion coefficient
APTR	=	amide proton transfer ratio
ASL	=	arterial spin labeling
BOLD	=	blood oxygenation level dependent
CP	=	Carr-Purcell
CNS	=	central nervous system
CT	=	computed tomography
DCE	=	dynamic contrast enhanced
DNP	=	dynamic nuclear polarization
DSC	=	dynamic susceptibility contrast
DTI	=	diffusion tensor imaging
DWI	=	diffusion weighted imaging
EPR	=	electron paramagnetic resonance
MR	=	(nuclear) magnetic resonance
MRI	=	magnetic resonance imaging
MRS	=	magnetic resonance spectroscopy
MTR	=	magnetization transfer ratio
MTT	=	mean transit time
PET	=	positron emission tomography
PS	=	phosphatidyl serine
ptO2	=	partial tissue oxygen tension
T1	=	longitudinal relaxation time
T2	=	transverse relaxation time
T2*	=	apparent transverse relaxation time
VASO	=	vascular space occupancy
VEGF	=	vascular endothelial growth factor
3D	=	three dimensional
5-FU	=	5-fluorouracil

Figures and Tables

Figure 1 Schematic summary of in vivo properties of tumors that are probed by MR techniques.

Figure 2 Axial MR images from a patient with a right thalamic high grade glioma acquired at 1.5T (A) T₁-weighted; (B) T₂-weighted; (C) DWI; (D) fractional anisotropy image (E) ₁H MRS from the tumor core with major peaks assigned as follows: Choline-containing compounds (3.23ppm), total creatine (at 3.03 ppm) and mobile lipids (at 2.0, 1.3 and 0.9 ppm); (F) Choline-containing compounds-to-total creatine (Cho/Cr) ratio image from multivoxel MRS represented as a color map on a T₁-weighted image acquired after injection of a Gd-contrast agent.

Figure 3 A BT4C rat glioma imaged at 4.7 T using Carr-Purcell (CP) multi-echo MRI. A BT4C glioma bearing rat was treated with herpes simplex virus thymidine kinase-ganciclovir gene therapy. (A) Shows a CP MR image acquired with short and panel (B) with long interpulse interval. Panel (C) is a normalized difference image of images shown in panels (A) and (B) yielding the so-called dynamic dephasing contrast. It should be noted that in this image (C) free water gives no signal and thus both ventricles and a tumor cyst appear dark. Instead, the tumor tissue undergoing cytotoxic cell death shows bright signal that is not obvious from either of the images in (A) and (B). Courtesy of Dr. Olli H.J. Grohn, University of Eastern Finland, Kuopio, Finland.

Figure 4 Postcontrast image (T_1C), endothelial transfer constant (K_trans), extracellular extravascular space (v_e) and blood plasma volume (v_p) maps obtained using DCE MRI in a patient with a glioblastoma multiforme. Courtesy of Drs. Geoffrey Parker and Samantha Mills, University of Manchester, UK.

Figure 5 Transverse T₁ maps of drug-treated and untreated EL-4 tumors in animals injected with PS-active (GST-C2A-Gd) and PS-inactive (GST-C2A-Gd) contrast agents. Color scale indicates T₁ values for image voxels. In this example, contrast agents were matched for relaxation rate. Images were acquired immediately before injection of contrast agent (a T₁ map acquired from a tumor before injection is shown on the left-hand side) and at 24 h after injection. Reference capillary was placed adjacent to the tumors, which were implanted on lower areas of backs of animals. Position of the tumor is indicated on the gray-scale image. (A) Etoposide+cyclophoshamide-treated tumor in animal injected with PS-active GST-C2A-Gd (TA). (B) Drug combination-treated tumor in animal injected with PS-inactive GST-C2A-Gd (TI). (C) Untreated tumor in animal injected with PS-active GSTC2A-Gd (UA). (D) Untreated tumor in animal injected with PS-inactive GST-C2A-Gd (UI). Drug combination-treated tumor in an animal injected with PS-active contrast agent shows greater accumulation at 24 h after injection (A). Courtesy of Dr. Kevin Brindle, University of Cambridge, UK.

Table 1 Comparison of characteristics for in vivo imaging techniques

Technique	Physics	What is revealed	Resolution (preclinical/clinical)	Duration of scans	Costs
X-RAY/CT (computerised tomography)	X-ray absorption	Organ anatomy, excellent for calcified tissues	Micrometer/millimeter isotropic	Minutes	Intermediate
ULTRASOUND	Echoes generated to high-frequency sound	Organ anatomy, tissue interfaces, contrast agents for some applications	Micrometer/millimeter isotropic	Minutes	Low
SPECT (single positron emitting computerised tomography)	Radioactive tracers	Receptor binding, uptake and metabolism of exogenous compounds, heamodynamics	Millimeter/sub-centimeter in plane (10–20 mm slice)	Tens of minutes	Intermediate
PET (positron emission tomography)	Positron labeled isotope tracers	Receptor binding, gene activity, uptake and metabolism of exogenous compounds, hemodynamics	Submillimeter/millimeter in plane (5–10 mm slice)	Tens of minutes	High
OPTICAL/NIR (near infrared)	Light scattering/light absorption	Cell surfaces, Hemoglobin, NADH and NADP, cytochromes	Micrometer/submillimeter (poor depth penetration)	Seconds to minutes	Low
EPR (electron paramagnetic resonance)	Electron paramagnetic resonance, sensing unpaired electrons	Paramagnetic species, partial oxygen pressure (require injection of exogenous reporter)	Millimeter/several millimeter, For oxygenation few hundered micrometer around the probe	Tens of minutes	Intermediate
MRI (magnetic resonance imaging)	Nuclear magnetic resonance, mainly water protons	Anatomy, hemodynamics, edema, vascular properties (use of exogenous contrast agents)	Micrometer/submillimer in plane (0.3–1 mm slice)	Tens of minutes	High
MRS (magnetic resonance spectroscopy)	Nuclear magnetic resonance, exploiting chemical shift of nuclei	Endogenous metabolites, energy state, pH, Mg^2+ activity Metabolic fluxes using ^13C or ^15N labeled substrates	Few microliter / 0.5 mL for ^1H 100 microliter / 25 mL for ^31P	Tens of minutes	High

Acknowledgments

Support from Medical Research Council, the Engineering and Physical Sciences Research Council, the Cancer Imaging Programme funded by Cancer Research UK, EPSRC, MRC and National Institute for Health Research for research in the author's laboratories that is cited in the review is greatly appreciated. Dr. Martin Wilson (University of Birmingham) kindly constructed Figure 2.