Functional magnetic resonance: biomarkers of response in breast cancer

Functional magnetic resonance (MR) encompasses a spectrum of techniques that depict physiological and molecular processes before morphological changes are visible on conventional imaging. As understanding of the pathophysiological and biomolecular processes involved in breast malignancies evolves, newer functional MR techniques can be employed that define early predictive and surrogate biomarkers for monitoring response to chemotherapy. Neoadjuvant chemotherapy is increasingly used in women with primary breast malignancies to down-stage the tumour and enable successful breast conservation surgery. It also plays a role in the treatment of undetected micrometastases. Cardinal physiological features of tumours that occur as a result of interactions between cancer cells, stromal cells and secreted factors and cytokines and how they change with treatment provide the opportunity to detect changes in the tumour microenvironment prior to any morphological change. Through sequential imaging, tumour response can be assessed and non-responders can be identified early to enable alternative therapies to be considered. This review summarises the functional magnetic resonance biomarkers of response in patients with breast cancer that are currently available and under development. We describe the current state of each biomarker and explore their potential clinical uses and limitations in assessing treatment response. With the aid of selected interesting cases, biomarkers related to dynamic contrast-enhanced MRI, diffusion-weighted MRI, T2*/BOLD and MR spectroscopy are described and illustrated. The potential of newer approaches, such as MR elastography, are also reviewed.


Introduction
Magnetic resonance imaging (MRI) of the breast has an established role in assessing response to neoadjuvant chemotherapy and provides better monitoring of the chemotherapeutic eff ect than clinical breast exami nation, mammography and ultrasound, especially in non-mass lesions and tumours that have fragmented into many foci [1]. As the overall response rate off ered by neoadjuvant chemotherapy ranges from 60% to 80%, with complete pathological response rates being around 10% to 20% [2], identifi cation of early response is important in planning subsequent management. More complex functional MRI techniques off er to quantify changes in tumour microvasculature, cell density, hypoxia, metabo lism and stiff ness and so provide early predictive and surrogate biological biomarkers for monitoring response to chemotherapy.
As breast tumours respond to chemotherapy, changes occur within the tumour and its microenvironment. Angio genesis, the fundamental physiological process asso ciated with tumour development, is interrupted. Th e composition of the extracellular matrix and stroma is altered, and secreted factors and cytokines, which can aff ect the transport of molecules to and from tumour cells, change the physiology and chemical composition of the tumour. For example, tumour cells become hypoxic and fragment, leaving fi brotic and collagenous tissue that may be quantifi ed using functional magnetic resonance (MR) techniques.
Th is article reviews the functional MR biomarkers of response currently routinely available and under development for assessing treatment response. Specifi cally, dynamic contrast-enhanced (DCE)-MRI, diff usionweighted (DW)-MRI, intrinsic susceptibility-weighted MRI, MR spectroscopy (MRS) and MR elastography are described with a focus on the current state of each technique and its limitations as a response biomarker.

Dynamic contrast enhanced MRI
DCE-MRI of the breast involves an intravenous injection of a low molecular weight T1-shortening paramagnetic compound (a gadolinium chelate) at doses between 0.1 and 0.2 mmol/kg. Agents currently licensed for use Abstract Functional magnetic resonance (MR) encompasses a spectrum of techniques that depict physiological and molecular processes before morphological changes are visible on conventional imaging. As understanding of the pathophysiological and biomolecular processes involved in breast malignancies evolves, newer functional MR techniques can be employed that defi ne early predictive and surrogate biomarkers for monitoring response to chemotherapy. Neoadjuvant chemotherapy is increasingly used in women with primary breast malignancies to down-stage the tumour and enable successful breast conservation surgery. It also plays a role in the treatment of undetected micrometastases. Cardinal physiological features of tumours that occur as a result of interactions between cancer cells, stromal cells and secreted factors and cytokines and how they change with treatment provide the opportunity to detect changes in the tumour microenvironment prior to any morphological change. Through sequential imaging, tumour response can be assessed and non-responders can be identifi ed early to enable alternative therapies to be considered. This review summarises the functional magnetic resonance biomarkers of response in patients with breast cancer that are currently available and under development. We describe the current state of each biomarker and explore their potential clinical uses and limitations in assessing treatment response. With the aid of selected interesting cases, biomarkers related to dynamic contrast-enhanced MRI, diff usion-weighted MRI, T2*/BOLD and MR spectroscopy are described and illustrated. The potential of newer approaches, such as MR elastography, are also reviewed.
include gadopentetate dimeglumine (Magnevist), gadodia mide (Omniscan), gadobenate dimeglumine (Multihance), gadoteriodol (ProHance), gadofosveset trisodium (Vasovist), gadoxetic acid (Eovist) and gadoversetamide (OptiMARK). Once injected, gadolinium circulates in the blood stream before passing into the extravascular extracellular or interstitial space. Th e concentration of gadolinium equilibrates between the intravascular and extravascular compartments over time and is eventually excreted by the kidneys. Post contrast images provide additional information to the unenhanced sequences by exploiting diff erences in temporal enhancement characteristics between malignant and normal or benign tissues. Th e diff erential uptake and washout of gado linium in each of these tissues results in an increased signal on T1-weighted (T 1 W) images. Along with morpho logical assessment from the unenhanced T 1 W and T 2 W sequences, the use of gadolinium to depict enhancement characteristics of the tissues can improve the sensitivity of MRI for cancer detection to between 89% and 100% [3].
Along with its high sensitivity, the specifi city of breast DCE-MRI, although initially reported to be low, has more recently been shown to equal that of mammography with signifi cantly higher values than ultrasound [4]. Th e blooming sign seen in 63% of malignant compared to 14.7% of benign lesions describes a brisk enhancement with sharply shaped borders at 1 minute after a bolus contrast injection that become progressively unsharp [5]. Further potential adjunctive morphological indicators of malignancy include unifocal oedema (91% of malignant lesions and 45% of benign lesions), centripetal enhancement with a rapidly enhancing outer ring that fi lls in (52% malignant lesions [6]), and the hook sign, a hook-like connection to the underlying pectoral muscle (33% of malignant lesions and 5% of benign lesions [7]). Th e presence of an adjacent vessel on subtraction images is also a promising sign for malignancy (85.9% of malignant and in situ lesions compared to 19.6% of benign lesions [8]). Finally, the addition of morphologic signs from unenhanced T 2 W sequences, such as spiculated margins, homogeneous intermediate signal intensity or stellate appearance, have been described to further improve the specifi city of breast MRI [8].
Th e rate of contrast uptake into breast lesions is nonlinear and diff ers between malignant and benign lesions ( Figure 1) so that enhancement curve characteristics can be used in conjunction with morphologic features to aid diff erential diagnosis. Malignant lesions exhibit stronger and faster enhancement than benign changes or normal tissues. In benign lesions, a slow onset (type I) curve that plateaus after 3 to 5 minutes is described in 83% of cases. In malignant lesions, a rapid onset with plateau (type II) or rapid onset with washout (type III) curve can be found in 91% of cases (57% for type III and 34% for type II) [9].
Semi-quantitative parameters can be calculated from these enhancement curves, including the onset time (from injection to the appearance of contrast in the tissues), maximum signal intensity, gradient or rate of contrast uptake and washout, and initial area under the time signal curve (IAUC).
Quantitative analysis involves pharmacokinetic modelling and requires more complex analysis methods of estimating changes in tissue contrast agent concentration following intravenous injection. Between 12% and 45% contrast leaks into the extravascular extracellular space (v e ) during the fi rst pass and results in measurable T1 shortening of tissues. Th e transfer constant, K trans , describes the transendothelial transport of contrast medium by diff usion from the vascular space to the tumour interstitium and provides a measure of vascular permeability. Over time, gadolinium diff uses back into the vasculature, which can be measured by the rate constant, k ep . Th ese parameters are related by the equation k ep = K trans /v e [10] (Figure 2).
Magnetic fi eld inhomogeneities induced by gadolinium on a T 2 W image can also be exploited to derive relative measures of blood fl ow and volume (rBF and rBV) as well as mean transit time (MTT). Th ese variables are related by the central volume theorem equation (BF = BV/MTT). Th e signal loss on a T 2 W sequence caused by dephasing of spins is related to the concentration of gadolinium and thus to vessel size and density [11].

Figure 1. Time-signal intensity curve for breast lesions.
A type I curve shows progressive enhancement in which the signal continues to increase over the whole dynamic study. A type II curve plateaus off after an initial increase in enhancement. A type III curve demonstrates immediate washout after a rapid increase in enhancement. Th ere is good evidence for use of such quantitative measures in breast cancer. K trans is generally high in tumours, as is k ep (Fig. 3), and a signifi cant reduction of up to a third has been shown in both parameters in patients with locally advanced breast cancer responding early in their course of neoadjuvant chemotherapy [12]. In addition, an increase in v e of nearly a third has been shown in non-responders [13]. rBV and rBF obtained from T 2 *W dynamic susceptibility sequences have also shown reduc tions of around two-thirds in patients responding to treatment [12]; however, T 2 *W functional imaging, while commonly used in MRI of the brain, is almost never used in breast MRI. Alterations in these parameters are likely to relate to changes in microvessel density and function of the microvasculature due to antiangiogenic eff ects of chemotherapy. Th e evidence from phase I and II studies strongly suggests that K trans can be used as a predictive biomarker to determine response to antiangiogenic drugs or vascular disruptive agents, with a change in K trans of greater than 40% considered by many investigators as the threshold required to represent defi nitive disease response [14]. Th ere is some discrepancy in the published data, however, with several other studies demonstrating little or no decrease in K trans or k ep following neoadjuvant chemotherapy [15,16]; in fact, one small study of 29 patients scanned very early after one cycle of chemotherapy showed that early tumour size change is a better response predictor than either K trans or k ep [17].

Time
Th e explanation for these variations in reported data are multifactorial: patient numbers, tumour type, chemotherapeutic agent and time-point of scanning after commencing therapy have all varied. Th e classifi cation of responders was also not consistent, varying from a 65% reduction in the largest tumour volume [13] to the accepted Response Evaluation Criteria in Solid Tumours (RECIST) or International Union Against Cancer criteria [18] of 30% or more reduction in one-dimensional tumour size or a 50% or more reduction in the product of the tumour diameter (assuming a spherical tumour model), respectively [12,15,16]. Th e group who demonstrated change in tumour size to be a better predictor of response compared to K trans or k ep chose an arbitrary reduction of 15% in one-dimensional size [17].
Th ere was also a signifi cant diff erence in data analysis methodology; most studies used manual regions-ofinterest (ROI) on enhanced subtracted images, although one group used semi-automated ROI generation, and in addition analysed a 3 × 3 pixel ROI hot spot [13]. Furthermore, the median or mean of each pharmacokinetic parameter for analysis has not been consistent. An increasing awareness of the heterogeneity of breast tumours makes the median a more appropriate parameter, with the change in the skewness of the distribution of these parameters likely to be as signifi cant as changes in the median value.
Another source of variation is the range of mathematical models used for pharmacokinetic analysis and the choice of arterial input function measurement, which also impacts on the overall results of tumour vascular heterogeneity. Traditionally, use of a nearby blood vessel for arterial contrast was deemed the ideal arterial input function [19], but population-based arterial input functions are more robust [20]. Alternatively, tumour enhance ment relative to that in neighbouring muscle tissue can be evaluated [21], and avoids error from fl ow eff ects in blood vessels. With the introduction of standard ised scanning protocols, automated analysis soft ware and the publication of reproducibility studies, derivation of pharmaco kinetic parameters could become more standardised and robust and be usefully adopted as functional imaging markers in breast cancer.

Diff usion-weighted MRI
DW-MRI develops intrinsic contrast within tissues based on the microscopic motion of water molecules by applying magnetic fi eld gradients during the MRI pulse sequence that sensitize the readout signal to losses from this motion. DW-MRI contrast provides diff erent, and complementary, information to DCE-MRI, being sensitive to factors that aff ect this microscopic water motion, such as cell density, membrane integrity and tissue microstructure. Changes in signal intensity on DW-MRI refl ect the movement of water diff usion over distances of 0 to 30 μm over time periods of 50 to 100 ms [22]. As with other tumours, breast cancers demonstrate restricted diff usion because water molecules cannot move as freely in tissues with a high cell density where extracellular space is limited ( Figure 4); this results in reduced signal loss from Brownian motion and is seen as a high signal intensity lesion on the DW-MRI image ( Figure 5). Magnetic fi eld gradients used to provide diff usion weighting can be varied in their amplitude, duration and spacing, which are jointly refl ected by a 'b' value. Acquisition of DW-MRI data using at least two b values allows calculation of an apparent diff usion coeffi cient (ADC), a derived logarithmic parameter of signal change with b value. At very low b values (<100 s/mm 2 ), the ADC predominantly refl ects larger distances of water move ment likely to represent movement within microvessels. Th is phenomenon is known as intravoxel incoherent motion [23]. By eliminating these low b values, this 'perfusion' eff ect in vascular rich areas can be suppressed and the ADC value can more accurately represent the shorter distances travelled by water protons in the extracellular space, or true diff usion [24].
Th e role of DW-MRI in tumour diagnosis is gradually being explored and it is increasingly shown to aid decision-making [25]. Diff erentiation between malignant and benign breast lesions using DW-MRI has been well reported [26][27][28], with the mean ADC value of malignant lesions being signifi cantly lower than that of benign lesions or normal breast tissue. Th is degree of overlap requires incorporation of an ADC threshold methodology for analysis; a 1.6 × 10 -3 mm 2 /s cutoff gives up to 96% sensitivity and 55% specifi city for tumour identifi cation [27]. More recently, one group has normalised ADC values to the surrounding glandular tissue and demon strated a reduction in overlap between benign and malig nant lesions. Using this method, normalised ADC values for tumour and benign lesions are 0.55 × 10 -3 mm 2 /s and 1.1 × 10 -3 mm 2 /s, respectively) [29], with the optimal thres hold of 1.6 × 10 -3 mm 2 /s normalising to 0.7 × 10 -3 mm 2 /s. Th e visibility of lesions on DW-MRI is better in older women compared to younger women, likely related to the density of the glandular parenchyma. Also, due to the lower spatial resolution off ered by DW-MRI compared to DCE-MRI, the diagnostic performance of DW-MRI is less helpful for non-mass-like lesions such as invasive lobular carcinomas and lesions <1 cm in size [30]. Th e

Restricted diffusion
High signal intensity DWI Low ADC major false-positive lesions reported are intraductal papillomas and fi brocystic disease, which can also result in overestimation of cancer extension [27]. Mucinous carcinomas interestingly demonstrate a signifi cantly higher ADC compared to other types of breast cancer, leading to false-negative reports [31]. DW-MRI also shows promise as an early surrogate biomarker for detecting response to neoadjuvant chemotherapy. Induction of successful apoptosis results in loss of membrane integrity, alteration of membrane barriers to water diff usion and cell shrinkage, increasing extracellular space. Th is translates to a rise in the ADC value of up to 35% and precedes any decrease in tumour size in locally advanced tumours [32][33][34]. Transient early decreases in ADC have also been demonstrated before this increase, and are thought to be related to cell swelling, reduction in blood fl ow or changes in composition of the extracellular space [16].
Th e optimal b values for diff usion-weighted MRI in the breast have not been established; nor indeed have the optimal scanning protocol, imaging parameters and methods of analysis, which all have a bearing on the ADC value. Published data indicate that b values from 0, 600 to 850, and up to 1,000 may be optimal [27,28,30], with at least three values required to ensure robustness of reproducibility of the measurement. Th e ability to obtain these data without the use of extrinsic contrast agents, in a short scanning time, independent of magnetic fi eld strength and operator interpretation is hugely advantageous. Reproducibility studies and quality assurance of methodology crucially need to be established.

T 2 */blood oxygen level-dependent MRI
Blood oxygen level-dependent (BOLD) or intrinsic susceptibility-weighted MRI relies on the paramagnetic property of deoxyhaemoglobin, which creates susceptibility variations in the magnetic fi eld (or microscopic fi eld gradients), which in turn decrease the transverse relaxation rate R 2 * ( = 1/T 2 *) of water in blood and the tissue surrounding blood vessels. An increase in the deoxyhaemoglobin concentration (that is, hypoxia) leads to a decrease in the signal intensity on the T 2 * image and a faster R 2 * [35] (Figure 6). An improvement in oxygenation has the converse eff ect. Deoxyhaemoglobin therefore acts as an intrinsic BOLD contrast agent for imaging tissue hypoxia. Specifi c gradient-recalled echo (GRE) sequen ces are required to detect changes in R 2 *. Variations in R 2 * have largely been evaluated in xenograft and human models using inhaled carbogen (95% oxygen (O 2 ), 5% carbon dioxide (CO 2 )) to intensify the otherwise small changes in signal intensity: the CO 2 induces vasodilation and the O 2 tension is high with 95% O 2 so that subtracted images with and without carbogen reveal regions of hypoxia where signal change is greatest. Unfortunately, the hyperventilation induced by breathing carbogen in humans is poorly tolerated so reliance has been largely on R 2 * measurements during air, or alternatively 100% oxygen, breathing.
A recent study in breast cancer patients has shown R 2 * values to be signifi cantly lower in tumour than normal breast parenchyma prior to the commencement of chemotherapy [36], suggesting that breast tumours are less hypoxic than normal breast tissue, possibly because of their high vascularity. Th is contrasts with other recently published data in prostate cancer [37], where R 2 * is increased, indicating increased hypoxia in these tumours. Th e increased R 2 * in normal breast tissue has also been related to the fi brocollagenous ligaments of Cooper, which maintain normal breast structural inte grity and contribute to higher R 2 * values. In responders following treatment, the R 2 * value has been shown to increase, likely as a result of decreased blood fl ow; however, in this one published study this parameter was not as effi cacious as changes in other DCE-MRI para meters, such as K trans , rBV, and rBF, or even morphological parameters such as tumour size, in indicating response [36]. Th e complexity  and heterogeneity of the micro vascu lature in diff erent tumour types thus need to be understood prior to using such measurements for evaluating changes in tumour oxygenation in response to chemotherapy.

MR spectroscopy
MRS exploits the nuclear spin properties of hydrogen ( 1 H) as well as of other atoms with unpaired protons, such as 31 P, 23 Na and 19 F, in a magnetic fi eld to absorb and emit radiofrequency. Th e acquired frequency spectrum of a range of metabolites provides information about the altered metabolism of cancer cells. With 1 H MRS, effi cient water suppression is mandatory to document proton resonances within molecules known to be increased in cancer, such as choline and lipids; protons within these molecules resonate at slightly diff erent frequencies when placed in a magnetic fi eld because of their immediate molecular environment. In breast cancer, as with other tumours, high levels of cholinecontaining metabolites involved in phospholipid metabolism, and thus cellular membranes in prolifera tion, result in a triplet at 3.22 ppm of free choline, phosphocholine and glycerophosphocholine. Choline is virtually undetectable in normal breast tissue and a peak at 3.25 ppm indicates benign tissue [38]. Several groups have shown that total choline concentration [Cho] can be used as a marker of malignancy, and when combined with DCE-MRI, increases the specifi city of breast MR up to 88% (and to 100% after the inclusion of a single slice T 2 * perfusion measurement) [39]. In vivo 1 H-MRS has also been shown to be useful in monitoring metabolic response to chemo therapy, with increased [Cho] and water/fat ratios in malignant tumours indicative of residual disease [40]. Small patient studies to date show promising results, with reduction of the choline signal following two cycles of neoadjuvant chemotherapy being more sensitive than changes in tumour size at predicting pathological response [41]. Also, by imaging after one cycle of chemo therapy, the same group demonstrated that a reduction in the choline signal may be more sensitive than DW-MRI in demonstrating pathological response [41,42].
In vivo 1 H MRS is a single, large voxel technique, and overall variations in the fat and water composition, particularly in heterogenous tumours such as invasive lobular cancers and ductal carcinoma in situ, reduces sensitivity of [Cho] quantifi cation. Partial volume eff ects in a large voxel also cause problems in the quantifi cation of [Cho], which pose a signifi cant problem after neoadjuvant chemotherapy [43].
Two dimensional localised correlated sepctroscopy (L-COSY MRS) incorporates a second spectral dimension that is indirectly detected through the acquisition of multiple one-dimensional MRS with incrementally longer times to echo (TEs). Cross-correlation of peaks enables identifi cation of lipid species by reducing contami nation from overlapping metabolite resonances. Early reports have shown identifi cation of invasive ductal carcinoma within these spectra versus normal fatty breast tissue with 92.4% sensitivity and 92.7% specifi city [44]. However, the technique is time consuming (20 minutes for a 3 cm voxel) and requires specialist analysis software, so its clinical utility is limited.
Sodium ( 23 Na) MRI has also been shown to be a very sensitive indicator of cellular integrity and cellular energy metabolism [45], with an elevated tissue sodium concentration in neoplastic tissue. 23 Na images can be accurately determined by co-registering high-resolution 1 H images acquired during the same scan. Its potential as a surrogate biomarker of response has been reported with a signifi cantly reduced tissue sodium concentration in responders after one cycle of chemotherapy [46].

MR elastography
MR elastography, an imaging correlate to palpation, is another novel technique that can be easily implemented  T2W image DCE Subtraction image T 2 *W image in the clinic. It assesses the viscoelastic shear properties of lesions by direct MRI visualisation of acoustic waves and quantifi es the decreased elasticity of malignant tumours. Quantifi cation of the diff erential 'stiff ness' between a breast lesion and the background adipose and fi broglandular tissues is achieved by assessing the propagation of mechanical waves, generated by an electromechanical driver, through the breast using a gradient echo phase contrast sequence [47]. Th e tissue stiff ness map (or elastogram) is based on a linear scale, calibrated into kilopascals and represented as a colour map. MR elastography can be performed as an extension to conventional breast MRI and could potentially be incorporated into a standard MRI breast examination. It is already being used clinically for the assessment of patients with chronic liver disease. Th ere have been very few studies of breast MR elastography; early published data on a small population group suggests that MR elastography in combination with DCE-MRI could increase the diagnostic performance of breast MRI and increase its specifi city from 75% with a persistently high sensitivity of 90% [48]. Further investigations of larger cohorts and smaller lesions will be necessary to validate these results.

Discussion
In patients with breast cancer, traditional anatomical imaging using size and morphological criteria for assessing response to neoadjuvant chemotherapy does not provide information on functional changes within tumours. Th e current challenge is to move beyond these anatomical boundaries to develop a more personalised, individual assessment of functional changes within a tumour for response evaluation. For those with the poorest prognosis and most advanced disease, early identi fi cation of nonresponders allows alternative management options to be considered. MRI has the versatility to contribute to the functional assessment of breast tumours and improve diagnostic confi dence, as well as provide early surrogate markers of disease response.
Functional MRI biomarkers of response described need careful validation, ideally against clinical outcome measures, before they can be adopted as established surrogate end points of response. However, currently an insuffi cient number of clinical studies have been reported with this kind of data for this to happen; reported changes are summarised in Table 1. Multicentre validation against histopathological markers, such as for microvessel density, apoptosis, hypoxia and vascularity, would further qualify the use of these functional biomarkers and support their translation into clinical practice. However, histological validation lacks true 'functional' input, so the limitation of this kind of validation needs to be recognised: it may well be that MRI on its own may more accurately refl ect in vivo tumour physiology.
In an attempt to tackle the issues of validation, two phase II multicentre national trials are underway -Neo-COMICE in the United Kingdom and I-SPY 2 in the US, both of which are examining the eff ectiveness of MRI in the early prediction of response to neoadjuvant chemotherapy and the development of surrogate imaging biomarkers. Neo-COMICE, currently in its pilot stage, also aims to evaluate the optimal scanning protocols that determine parameters of greatest predictive value of treat ment response and establish parity of MRI examiations between centres. Currently, there is no consensus on a standardized MR imaging examination or on the role of MRI for assessing response in patients receiving neoadjuvant chemotherapy. Th e adaptive design of the I-SPY 2 trial will allow early data from one set of patients to guide decisions about which treatments may be more useful in the trial in order to eliminate ineff ective treatments and off er patients the best chance of successful therapy [49].
Th e increased availability of higher-fi eld MRI scanners allows higher signal-to-noise ratio and so better spatial resolution to increase the visibility of small cancers. Early reports suggest that the sensitivity and specifi city of DCE-MRI at 3T for malignant breast lesions increase to 95% and 91%, respectively, from 91% and 85% [50]. Going up in fi eld strength from the commonly available 1.5T to 3.0T is not without challenges, however, as the nonuniformity of the magnetic induction fi eld (B 1 ) results in non-homogeneity of fat suppression, which in turn leads to poor enhancement in areas with a very low magnetic induction fi eld and errors in quantifying enhancement ratios. Reports have shown that the B 1 fi eld in one breast can be reduced by as much as 40%, which is suffi cient to reduce the conspicuity of a malignant lesion and reduce the sensitivity to cancer detection [51]. Quantitative functional MRI at 3T requires improved radiofrequency excitation methods and improved analysis to ensure B 1 inhomogeneity is accounted for when calculating DCE metrics.
Future potential in the search for biomarkers of effi cacy for certain therapeutic treatments may lie in correlating baseline gene expression with MRI response using several of the above techniques. Gene arrays and immuno histochemistry analysis of vascular endothelial growth factor pathways could indicate which pattern of gene expression relates to specifi c changes in vascular volume and permeability assessed by MRI, and this is a promising area of research [52].
Ongoing research and recent technical advances indicate that the prospects for substantial improvements in monitoring of therapeutic response as well as for improved early detection and accurate diagnosis of breast cancer with functional MRI are promising, with the rate of development indicating early translation to routine clinical care. A key factor in their success will depend on rigorous quality control and assurance to ensure that the quantitative measurements are robust and reproducible.