Imaging tumor angiogenesis: functional assessment using MDCT or MRI?
- First Online:
- Cite this article as:
- Goh, V. & Padhani, A.R. Abdom Imaging (2006) 31: 194. doi:10.1007/s00261-005-0387-4
- 207 Views
Functional imaging using multidetector row computed tomography and dynamic contrast-enhanced magnetic resonance imaging are increasingly advocated for assessment of tumor vascularity because these techniques provide excellent anatomic imaging and reliable quantitative perfusion data and are easily incorporated into routine examinations. However, differences in acquisition techniques, mathematical analysis, measurement parameters, and propensity to artifacts influence the choice of imaging modality, which is explored in this review.
KeywordsComputed tomography, perfusion imagingMagnetic resonance, functional imagingAngiogenesis
Tumor angiogenesis is well recognized as an essential process for tumor growth, proliferation, and metastasis . Tumor growth cannot be sustained without an adequate blood supply, and drugs targeted at tumor vasculature are an attractive therapeutic strategy , which has already yielded real patient benefits . However, clinical assessment of new therapeutics has been problematic, not least because tumor shrinkage may not occur. For example, although Hurwitz et al.  reported a 5-month improvement in overall survival in patients with metastatic colorectal cancer that was treated with conventional chemotherapy and bevacizumab (Avastin, Genentech, San Francisco, CA, USA), an antivascular endothelial growth factor agent, this was accompanied by an increase of only 10% in objective response rate. Thus conventional assessment of therapeutic efficacy, based on change in size [4, 5], may be of limited value in assessing clinical efficacy; time to progression is probably the best method of assessing the efficacy of antiangiogenesis drugs because it reflects stability of disease. However, the disadvantage of using time to progression as an endpoint for clinical trials is that such studies are expensive and need large numbers of patients. Further, these patients may be treated with ineffective drugs for prolonged periods. In vivo biomarkers of angiogenesis, including functional imaging techniques, are increasingly being advocated [6, 7], particularly for early clinical studies (phases I and II), where they have been shown to provide pharmacodynamic information related to drug action [8, 9]. In this setting, such pharmacodynamic information can help in the dose selection and scheduling to support go/no-go decisions on new therapeutic compounds and to provide confidence to go forward to larger phase III clinical studies with efficacy endpoints. Imaging techniques allow evaluation of the functional status of tumor vasculature, with the tumor in situ, which may be an advantage over the “gold standard” of direct histologic quantification of angiogenesis because data from implanted tumor xenografts have shown that there can be a discrepancy between “true” functional and histologic microvessel density: 20% to 85% of microvessels are perfused at any given time depending on the microenvironment .
Of all the functional imaging techniques capable of providing qualitative and quantitative data of tumor vascularity, functional imaging using multidetector row computed tomography (f-MDCT), also known as perfusion CT, and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) have been promulgated for several reasons. These imaging techniques are widely used in routine oncologic imaging, and functional studies can be incorporated relatively easily with routine examinations. Anatomic depiction is excellent with both modalities, and reliable measurements can be obtained with good spatial resolution. Crucially, these measurements have been shown to correlate with histologic markers of angiogenesis [11–16].
However, there are differences between the two modalities, which will influence the choice of technique. CT and MRI techniques can provide qualitative and quantitative assessments of tumor vascularity, but quantification by DCE-MRI is technically more challenging than that by f-MDCT because of the lack of a direct relation between MRI signal intensity and contrast agent concentration, particularly in large vessels. This is related to the fact that contrast enhancement on MRI is derived from the effect of the contrast medium on the surrounding local environment, which induces a magnification effect (detected by MRI) that can be unpredictable at times (the contrast medium per se is not detected).
Contrast agent kinetics
Functional MDCT and DCE-MRI assessment is based on intravenous contrast medium enhancement. As a bolus of contrast agent passes through a vascular bed, with the exception of the brain, testes, and retina, contrast agent rapidly passes from the intravascular compartment into the extravascular extracellular compartment. The rate at which this occurs is determined by several factors including the rate of tissue delivery, vessel surface area, and the leakiness or permeability of these vessels. There is subsequent return of contrast from the extravascular compartment into the intravascular compartment over time, and contrast is eventually excreted by the kidneys, although some contrast agents do have significant hepatic excretion. Differences in contrast agent kinetics between normal tissue and tumor are exploited by both techniques to provide lesion- and tissue-specific information.
The relation between contrast concentration and enhancement is straightforward with CT; there is a direct linear relation between enhancement change and iodine concentration. For example, at 120 kV, an enhancement change of 25 HU is equivalent to 1 mg/mL of iodine . As a result, the arterial input, which is required for quantitative analysis, can be measured directly from a conveniently placed artery. Thus absolute quantification of perfusion is possible using f-MDCT and this has been hailed as the major advantage over DCE-MRI.
In contrast, the relation between MR signal intensity change and contrast agent concentration is not so easily defined and is indeed nonlinear. Paramagnetic contrast media produce magnetic field inhomogeneities within the vascular space and in the immediate vicinity. This results in a decrease in relaxation time of the surrounding tissues. The MR signal intensity change is dependent not only on contrast medium dose but also on the imaging sequence used (T2* versus T1 weighted), the machine setup, native relaxation rate of tissues, and intrinsic tumor heterogeneity. The arterial input, which is required for quantitative analysis, cannot always be measured directly partly due to velocity-induced signal intensity changes within vessels but also due to a nonlinear relation between contrast concentration and MRI signal intensity, particularly at the higher concentrations of contrast medium found in vessels (vide supra). Quantitative measurements are obtainable, although the underlying mathematics is more complex making a number of assumptions, and it has been repeatedly noted that quantification using pharmacokinetic modeling should be performed only if a direct relation between signal intensity and contrast agent concentration can be demonstrated throughout the measured range .
Typical f-MDCT and DCE-MRI acquisition parameters
MDCT Single level (2D)
MDCT Volume (3D)
T1W DCE-MRI (2D/3D)
T2*W DCE-MRI (2D)
>300 mg/mL iodine
>300 mg/mL iodine
0.5 mmol/mL gadolinium
0.5 mmol/mL gadolinium
3–5 mL/s bolus
2 mL/s infusion
3 mL/s bolus
4–6 mL/s bolus
4 × 5 mm
1 s for 1–2 min
5 s for 1–2 min
5–12 s for 5–7 min
1–2 s for 1–2 min
SNR of technique
Signal change observed/magnitude of effect
Deconvolution, distributed parameter model
General multicompartment pharmacokinetic model
Central volume theorem
Parameter typically measured
Blood flow, blood volume, mean transit time, permeability
Permeability, relative blood volume
Transfer constants, leakage space, blood volume and flow
Relative blood flow, relative blood volume, mean transit time
Although macromolecular contrast agents may allow assessment of macromolecular hyperpermeability (a specific tumor angiogenic characteristic) and provide a more specific approach of assessing vascular volume, these agents are not currently in clinical use, although the first MRI macromolecular contrast agent is expected to be licensed in early 2006. Preliminary data from animal studies using macromolecular contrast-enhanced f-MDCT have been promising. In rats with chemically induced primary liver tumors, early changes in hepatic flow allowed detection of hepatocellular carcinoma before development of overt lesions . A large body of DCE-MRI literature using macromolecular agents has shown its utility in being able to assess microvessel permeability and fractional plasma volume . Clinical studies of macromolecular contrast-enhanced MRI in human tumors are eagerly awaited .
Functional MDCT techniques are most commonly sequential single-level acquisitions, although sequential volume acquisitions are technically possible with current technology. Tumor coverage with single-level techniques will depend on the number of detectors. For example, tumor coverage with a four-detector row scanner (Lightspeed Plus, GE Healthcare Technologies, Waukesha, WI, USA) is 2 cm, consisting of four contiguous 5-mm axial images; tumor coverage with a 64-detector row scanner (VCT, GE Healthcare Technologies) is 4 cm, consisting of eight contiguous 5-mm axial images.
At least one baseline noncontrast image acquisition is required for f-MDCT techniques, before image acquisition after intravenous contrast medium administration. Acquisition sampling and timing will depend on several factors including the perfusion parameter being measured, the mathematical analysis method, and technical considerations. Typical parameters are listed in Table 1. A wide variation in acquisition techniques is found in clinical practice; this in part is due to the different commercial software packages available, which themselves impose strict data acquisition requirements. These software packages implement different quantitative analysis methods, which include compartmental analysis, deconvolution, Patlak analysis, and the distributed parameter model. Like DCE-MRI, there remains a lack of consensus regarding optimal acquisition technique, type of analysis method, and surrogate measurement to use .
Given the variety of f-MDCT methods that can be used clinically, studies tend to be tailored, taking into account the desired perfusion parameter, mathematical analysis method to be used, and anatomic location of the pathology. A bolus tracking technique with a high temporal resolution of one acquisition per second for 40 to 65 s is typically used for blood flow, blood volume, and transit time measurements. In contrast, permeability measurement requires longer but less frequent data acquisitions, in the order of 2 to 3 min with a sampling rate of up to one per 5 s. A hybrid acquisition technique can be tailored to measure all four perfusion parameters. For example, blood flow, blood volume, mean transit time, and permeability can be assessed using the deconvolution and modified distributed parameter model (Perfusion 3.0, GE Healthcare Technologies) by performing an acquisition with a high sampling rate (one per second) for the initial 60 s and then a lower sampling rate (up to one per 5 s) for the next 120 s.
For anatomic sites such as the lung or liver, where excessive movement along the long axis of the body may occur because of respiration, breath-hold acquisitions are useful to decrease misregistration and resulting mathematical modeling failures. Breath-hold acquisitions are typically in the order of 40 s (this is usually possible, particularly with the assistance of oxygen breathing). Multiple acquisitions can be performed, e.g., to assess liver tumor perfusion, although care must be taken to ensure that the same tumor level is examined during such acquisitions. Motion misregistration appears to be less of a problem for DCE-MRI because motion can be compensated for by the use of navigator technology developed for cardiac applications or by imaging in the plane of motion and then to use anatomic registration techniques before data analysis.
As with f-MDCT, different DCE-MRI techniques are available for use in clinical practice. As with f-MDCT, DCE-MRI is essentially a single-level technique, although several tumor levels may be scanned during an examination; techniques are tailored according to the desired measurement parameters. As with f-MDCT, qualitative and quantitative measurements can be obtained, although direct quantification is more challenging with DCE-MRI. T1- and T2*-weighted sequences are most commonly used. Typical acquisition parameters are listed in Table 1. The choice of sequence and parameters will depend on anatomic coverage, acquisition times, susceptibility to artifacts resulting from magnetic field inhomogeneities, and need for quantification.
T1-weighted sequences (T1-weighted gradient echo, saturation recovery/inversion recovery snapshot sequences, or echo planar sequences) are sensitive to contrast within the extravascular extracellular compartment and provide information on microvessel perfusion, permeability, and extracellular leakage space. Shortening of the T1 relaxation time is the mechanism of contrast enhancement; the signal intensity is dependent on the native T1 relaxation of tissue, dose of contrast agent, imaging sequence and parameters used, machine gain, and scaling factors. Qualitative measurements obtained with such sequences, including the shape of the signal enhancement/time curve, gradient of the upslope of the curve, maximum signal intensity, and washout gradient, work well in diagnostic applications but, in general, are not recommended for longitudinal/multicenter studies because these semiquantitative measurements are influenced by scanner settings. Quantification using pharmacokinetic modeling is possible so long as appropriate calibrations are performed, thus enabling the relation between signal intensity and contrast agent concentration over the measured range of values to be determined.
T2*-weighted sequences (susceptibility weighted echoplanar spine echo or gradient echo sequences) are equivalent to the bolus tracking type sequences used in f-MDCT techniques and provide information of perfusion and blood volume. However, unlike MDCT, these are relative parameters. Direct quantification is achievable in the cranial circulation by measuring the changing contrast concentration within the feeding artery, but this is technically challenging. Direct measurement of the arterial input is not easily achievable in the extracranial tumor circulation due to factors such as artifact from nonlaminar flow, high vascular permeability, and T1 signal enhancing effects of contrast media leakage into the extravascular space in the first-pass counteracting T2* effects.
Validation and measurement reproducibility
Functional MDCT has been validated against a variety of techniques including microspheres, xenon CT, and O15-labeled H2O positron emission tomography [23–28]. Functional MDCT and DCE-MRI have been correlated against histologic markers of angiogenesis in a variety of tumors including lung, kidney, colorectal, breast, and cervical cancers [11–16, 29], although some MRI studies have found no correlation between T1 kinetic parameter estimates and microvessel density (MVD) . Other characteristics that have been correlated with enhancement patterns include the degree of stromal cellularity and fibrosis, tissue oxygenation, and tumor proliferation. Recently, vascular endothelial growth factor, a potent vascular permeability and angiogenic factor, has been implicated as an additional explanatory factor that determines MR signal enhancement .
Measurement reproducibility is acceptable for both techniques, which is important for the use of these techniques for response assessment purposes. A variability of 13% to 35% has been quoted for CT measurements in normal brain tissue in animal and human studies [30–33] and for colorectal and for lung cancer in human studies [34, 35]. In general, the variability quoted for DCE-MRI techniques is greater [36, 37], probably reflecting the inability of these techniques to adequately compensate for input function because vascular concentration of MRI contrast medium is not easily measured. This degree of measurement variability is well within levels of expected therapeutic effect of current antiangiogenic and antivascular drugs that are undergoing clinical evaluation, e.g., bevacizumab (Avastin, Genentech) [38, 39] and combretastatin (OXiGENE, Waltham, MA, USA) [8, 35].
Although greater inter- than intra-observer variability has been demonstrated for MDCT measurements, this variability has been acceptable for therapeutic assessment: intraclass correlation coefficients greater than 0.7 have been quoted for cranial  and colorectal  cancer measurements. A similar level of observer variability has been found for DCE-MRI measurements in breast cancer (A. R. Padhani, personal communication). This measurement variability is well within the levels of expected therapeutic change of current antiangiogenic and antivascular drugs that are undergoing clinical assessment.
DCE-MRI and f-MDCT techniques have been used in different clinical situations. Both techniques have been used in clinical trials of antiangiogenic and antivascular targeting agents to provide biological evidence of drug action [8, 9, 35, 38, 42, 43]. These techniques have been used to monitor the effects of other treatments, including radiotherapy and conventional chemotherapy, and to predict response. Functional MDCT has been used to monitor radiotherapy effects in rectal cancer  and prostate cancer  and as a predictor of response in head and neck cancer [46, 47]. DCE-MRI has been used to monitor the effects of neoadjuvant chemotherapy in breast and bladder cancers and bone sarcomas [48–52] and radiotherapy in rectal and cervical cancers [53–55]. DCE-MRI changes after one or two cycles of chemotherapy have been shown to predict therapeutic response . Both techniques have also been used as prognostic indicators: MDCT has been used in hepatic metastatic disease [57, 58] and lymphoma , and DCE-MRI has been used in cervical cancer [60, 55].
Which imaging technique?
The choice between f-MDCT and DCE-MRI will be determined by several key factors including local availability and expertise, tumor site, desired perfusion parameter, and the need to decrease radiation burden. DCE-MRI is currently established as a reliable technique for assessing angiogenesis, although quantification remains problematic. The widespread availability of MDCT may be a major determinant in future use. MDCT is already used extensively in oncology for diagnosis and therapeutic assessment. The availability of commercial software and ability for direct quantification makes assessment straightforward; multicenter assessment is easily achievable in comparison with DCE-MRI (where quality assurance challenges are greater). To date no studies have compared the performance of f-MDCT and DCE-MRI in tumor assessment, although a recent study has compared the performance of MDCT with that of DCE-MRI in the evaluation of solitary pulmonary nodules; the investigators concluded that there was no significant difference between techniques with respect to performances .
There are anatomic regions where f-MDCT is preferable to DCE-MRI, mainly due to the presence of artifacts that would interfere with MRI evaluations. These include the upper abdomen, in particular the root of the visceral vessels (in the region of the pancreas), the mediastinum, and at the pulmonary hila. Phase-encoded artifacts arising from vascular pulsatility, exaggerated by concentrated contrast medium, can render DCE-MRI uninterpretable. Conversely, for brain examinations, DCE-MRI should be the preferred technique because the radiation burden from MDCT, particularly for serial examinations, may become unacceptable.
In summary, functional imaging techniques are being used increasingly as a biomarker of tumor angiogenesis. MDCT and MRI techniques have been used successfully in clinical trials of antivascular and antiangiogenic agents. Ultimately the mechanism of action of the therapeutic agent, tumor site, patient characteristics, and local availability and expertise will influence choice of technique.