5- T1-Weighted DCE MRI-MN MJ2

T1-Weighted DCE MRI
Mohammad-Reza Nazem-Zadeh
14.1 Introduction
 DCE (Dynamic Contrast Enhanced) MRI
 measures tissue haemodynamics through tissue’s longitudinal relaxation time (T1 )
 monitors concentration of an intravenously administered injection of CA in vivo
 Quantitative analysis of signal intensity–time series
 provides measurements of tissue perfusion
 cerebral blood flow (CBF)
 cerebral blood volume (CBV)
 characterizes the brain microvasculature
 characterizes blood–brain barrier (BBB) integrity (may be disrupted by pathologies)
 provides measurements of permeability and distribution volume of CA in extracellular
extravascular space (EES)
14.1.1 Blood–Brain barrier (BBB)
 A highly selective multicellular semipermeable membrane structure
 separates the circulating blood from the extracellular fluid in the central nervous system.
 formed by endothelial cells
 Endothelial cells of the BBB are unique compared to other tissues
 continuous intercellular tight junctions
 lack of fenestrations (small pores)
 extremely low rates of transcytosis (Abbott et al., 2006; Obermeier et al., 2013).
 actively regulates the influx and efflux of exogenous and endogenous molecules
 maintains a microenvironment that allows neuronal circuits to function properly
14.1.2 BBB disruption
 BBB development and maintenance is governed by cellular and non-cellular elements
 astrocytes, pericytes, microglial cells, neurons and extracellular matrix components
 interact with endothelial cells, forming an interactive cellular complex.
 BBB breakdown
 leads to unregulated exchange of molecules, ions and/or cells
 agitates the normal function of neuronal processes
 enables brain interstitium to become locally accessible to low-molecular weight MRI Contrast
 accumulation of CAs causes an increase in T1 signal in tissue
 provides an alternative tool to probe brain pathologies in vivo
 e.g. intracranial neoplasm, ischaemia, meningitis, type II diabetes, multiple sclerosis (MS), small vessel
disease, etc.
14.1.3 Basis of contrast enhancement
following bolus injection
 Types of low-molecular-weight CAs used in clinical applications in brain are
 In healthy brain, CAs do not pass to the brain interstitium, except in a few
regions such as choroid plexus (Gibby, 2000).
 gadolinium-based CAs
 linear chemical structures
 cyclic structures
 Until recently, these agents were all considered to be very safe for human
14.1.3 Basis of contrast enhancement
following bolus injection
 Since 2010 the European Medicines Agency (EMA) have recommended against
the use (at least minimization) in patients with severe kidney problems or patients
undergoing liver transplants
 at risk of a condition known as “nephrogenic systemic fibrosis” by gadolinium
contained in these agents.
 In March 2017 the EMA recommended suspension of the marketing authorisation
for the four linear agents
 Evidence: Small amounts of gadolinium accumulate in brain following multiple CA
 Future studies will employ cyclic agents only.
14.1.3 Basis of contrast enhancement
following bolus injection
 When bolus of CA passes into blood circulation temporarily confined within vascular
space for many cardiac cycles: a phase known as the first pass.
 CA diffuses from blood pool into EES of the brain
 due to CA concentration difference between vascular space and brain interstitium,
 Rate and extent depend on underlying pathophysiology in
 tissue perfusion, capillary permeability, surface area of leaking vessels
 CA does not enter cells in brain tissue, distribution volume is the EES
 Accumulation of CA in brain EES causes T1 shortening and an increase in the MRI signal
 Over a period of time (many minutes to hours), CA diffuses back into vasculature
 Accumulation and washout of the CA is observed as a change in MR signal intensity.
 Kinetics (e.g. rate of enhancement, peak enhancement and signal decay) of the
distribution of CA depend on the pathology
14.2 Analysis of DCE-MRI data to
measure BBB disruption
 Simple visual assessment of spatiotemporal enhancement patterns
 Semi-quantitative and quantitative techniques through to the use of
complex tracer kinetic models
 a predefined ROI
 voxel-by-voxel analysis (generating parametric maps)
14.2.1 Visual assessment of
enhancement patterns
 Based upon subjective evaluation of signal–time curve describing
enhancement characteristics in post-contrast T1 -weighted images at
predefined time points
 4 general curve patterns:
 (1) no enhancement
 (2) slow enhancement
 signal rises slowly within ROI for the scan duration
 (3) fast enhancement followed by a plateau
 (4) rapid enhancement followed by a washout phase
14.2.1 Visual assessment of
enhancement patterns
 Studies:
 newly formed MS lesions
 initial enhancement patterns in lesions were distinguished as nodular -homogeneous
hyperintensity throughout the lesion, closed hyperintense ring
 Tumour
 a hypointense center, and open hyperintense rim with semilunar configuration
 a dynamic enhancement pattern was defined as
 centrifugal (enhancing from the centre to the periphery)
 centripetal (enhancing from the periphery to the centre)
14.2.2 Semi-Quantitative analysis of
enhancement series
 The simplest approach: empirical description of the enhancement curves
using a series of metrics
 peak of the enhancement curve
 initial wash-in slope
 time to maximum signal
 washout slope
 initial area under the enhancement curve (IAUC) over a predefined period of
time (e.g. for the first 90 s)
 cannot differentiate physiological from physical factors affecting signal
14.2.2 Semi-Quantitative analysis of
enhancement series
 MRI features shaping the signal intensity–time curves
 imaging sequence
 scanner parameters
 tuning and scaling factors of the machine
 native T1 of the tissue of interest
 dose and mode of administration (e.g. bolus injection, infusion) of CA
 patient’s systemic status (e.g. cardiac output)
 non-linear relationship between the CA’s concentration to signal intensity changes
 different sites or multiple MRI sessions
14.2.2 Semi-Quantitative analysis of
enhancement series
 Need to normalize and take into account the differences in amplitude or slope of
the enhancement because of the choice these factors
 normalization with respect to a reference tissue (healthy)
 using concentration–time series instead of signal intensity.
 Even then, interpretation of these parameters remains vague.
 IAUC or initial wash-in slope reflect kinetics governed by a combination of blood flow,
endothelial permeability, blood volume and EES volume
 Signal enhancement curves even within a single voxel may represent signal from a
blood vessel, EES or a combination
 Difficult to differentiate signal coming from each compartment
 Difficult to thus investigate interaction between those compartments
 This is where tracer kinetic modelling may be beneficial and provide insights into
14.2.3 Quantitative analysis using tracer
kinetic modelling
 Use of tracer kinetic models describing passage of a CA bolus through the
tissue of interest.
 Models aim to portray temporal features observed in a concentration–time
curve, through parameters that reflect physiological processes in vivo such
 BBB leakiness
 volume of the EES
 Volume of vascular spaces
14.2.3 Quantitative analysis using tracer
kinetic modelling
 Most of tracer kinetic models: tissue is made up of a number of compartments
 The number and assumptions on compartments, depend on the properties of CA
and tissue physiology or pathology
 intact [disrupted] BBB: low molecular weight CAs mostly confined within
vascular compartment: vb [diffuse into the EES: ve]
 The rate of diffusion depends on local blood flow (Fb), BBB permeability (P)
and surface area (S) (Tofts, 1997; Tofts et al., 1999).
A general model of tissue
The bold black arrows
illustrate the direction of
blood flow through the
vessel (i) with fractional
volume vp.
Grey arrows represent bidirectional CA leakage
(green circles) from the
vascular to the EES (ii)
volume ve) and vice
versa. Most CAs do not
enter the intracellular
space (iii gold discs).
14.2.3 Quantitative analysis using tracer
kinetic modelling
 A transfer constant describing the CA exchange between vascular and EES
 Its interpretation depends on a combination of BBB permeability (P) surface
area (S) product (PS) and CBF.
 At any time, the overall CA concentration within a voxel or ROI:
weighted sum of concentrations within each individual compartment
14.2.3 Quantitative analysis using tracer
kinetic modelling
 Several other factors not considered
 plug flow in vessels
 CA mixing within compartments
 More complex modelling approaches attempt to take these factors into account
14.2.3 Quantitative analysis using tracer
kinetic modelling
 Basis of compartmental tracer kinetic modelling: describe CA exchange
between compartments with a simple rate equation
 diffusive transport of a substance across a semipermeable membrane.
 describes CA flux to be driven by concentration gradient between two
compartments, with a rate constant that depends on the properties (e.g.
restrictiveness) of the membrane.
 Flux J of CA between two compartments:
 P takes into account any form of passive or active transport (e.g. paracellular and
transcellular) across semipermeable barrier.
14.2.3 Quantitative analysis using tracer
kinetic modelling
 Foundation of compartmental modelling
 Shows how compartments interact with one another
 Assumes that a compartment response to an influx is linear and stationary, (proportional to the
dose administered) and independent of the arrival time
 Characterizing CA concentration in each compartment as the sum of inlets and outlets, which
transport the CA in and out of the compartment, an outlet from one compartment can form the
inlet of another.
 No CA created or destroyed inside the compartment.
 CA kinetics in each compartment: expressed in form of differential equations describing
flux at inlets and outlets
14.2.3 Quantitative analysis using tracer
kinetic modelling
 When solved, Equation 14.1 is used to describe concentration profile in
combined system
 Complexity of compartmental models and number of parameters depend on
several factors
 quality of data acquired
 sampling frequency
 length of acquisition
 pathophysiology of disease
25 The Tofts model
 The most widely used model, distribution of an inert gas in lungs (Kety, 1951), quantitative
analysis in brain (Brix et al., 1991; Larsson et al., 1990; Tofts and Kermode, 1991).
 Substituting terminology commonly used in DCE-MRI in Equation 14.2, the rate of
accumulation and washout of a CA in EES, under the assumption of instantaneous and
well-mixed tracer: general form of the rate equation
 The physiological interpretation of Ktrans, depends on the balance between BBB
permeability and blood flow in tissue.
 In the general case of mixed perfusion and permeability regime, Ktrans is given by
26 The Tofts model
 Fp: plasma flow (volume of blood plasma per minute per unit volume of tissue
 E: fraction of tracer extracted to ve in a single transit through the capillary bed, given by
(a) DCE images from a
patient with meningioma
collected at t = 0, 8 and 260 s
following administration of
Gd-DTPA. BBB disruption
identified, where CA rapidly
leaks through barrier and
appears to distribute relatively
homogeneously within lesion.
Corresponding Ktrans map on
right demonstrates the intrinsic
heterogeneity of the lesion.
(b) DCE images from a
patient with a Grade 4 glioma
at t = 0, 16 and 260 s postinjection of Gd-DTPA, with the
respective Ktrans map.
Different characteristics of
two brain lesions are visible
from both enhancement
profiles and parametric maps.
28 The Tofts model
 Conventional Tofts model (Tofts and Kermode, 1991) discribed the
tracer kinetics in patients with MS, assumed a negligible contribution
to Ct from the vascular space (vp~0).
 Standard quantities and symbols published (Tofts et al., 1999).
29 The Tofts model
 Solving Equation 14.3 and substituting to Equation 14.1 a onecompartment model for weakly vascularized lesions is derived
 An example of the application of Tofts model is shown in Figure 14.3,
for two enhancing lesions in the brain that exhibit different
enhancement profiles.
For the slowly enhancing lesion (c) 2CXM: Fb = 0.24 ml min−1 ml−1, PS =0.02 ml min−1 ml−1, ve = 0.09,
vp = 0.18; ETM: Ktrans = 0.21 min−1, ve = 0.38, vp = 0.00; Tofts: Ktrans = 0.21 min−1, ve = 0.38.
FIGURE 14.3 Quantitative analysis using Tofts model, extended Tofts model (ETM) and twocompartment exchange model (2CXM) on dynamic data from enhancing lesions in the brain:
(a) high spatial resolution GE images acquired in the sagittal plane and used as a reference
(b) one dynamic post-contrast image for each enhanced lesion following a bolus-injection of Gadovist
(c) illustrate the application of different models in the enhancing lesions.
For the fast enhancing lesion (f) 2CXM: Fb = 0.67 ml min−1 ml−1, PS = 0.22 ml min−1 ml−1, ve = 0.11, vp =
0.06; ETM: Ktrans = 0.27 min−1, ve = 0.28, vp = 0.02; Tofts: Ktrans = 0.28 min−1, ve = 0.29. The data were
acquired on a 3T Philips MRI system.
FIGURE 14.3 Quantitative analysis using Tofts model, extended Tofts model (ETM) and twocompartment exchange model (2CXM) on dynamic data from enhancing lesions in the brain:
(d) high spatial resolution GE images acquired in the sagittal plane and used as a reference
(e) one dynamic post-contrast image for each enhanced lesion following a bolus-injection of Gadovist
(f) illustrate the application of different models in the enhancing lesions.
33 The extended Tofts model
 vp~0 assumption is true in pathologies where the contribution of vascular space
to overall concentration remains small
 In diseases such as tumours vp contribution of has to be considered.
 In highly perfused regime, Fp~∞ (Sourbron and Buckley, 2011), extended Tofts
model simply includes the CA concentration in the blood plasma
 The effect of the vascular component of the extended Tofts model is shown in
Figure 14.3.
34 The Patlak model
 The two models described above assume a bi-directional transport of CA between EES
and vascular space.
 Further simplification: assuming a unidirectional transport of tracer: from blood plasma into
the EES.
 Acceptable if the duration of the dynamic scan is short and the CA flux into ve is not
sufficient to fill that space.
 The Patlak model (Patlak et al., 1983) describes this scenario and can be expressed as
 also be expressed in linear form, by dividing each term with Cp(t), as it allows simple and
fast data post-processing.
35 The Two-Compartment
exchange model - 2CXM
 The models presented so far represent boundary regimes of the more
general 2CXM:
 (1) weakly vascularized
 (2) highly perfused lesions.
 The 2CXM can be applied to general case of a mixed perfusion and
permeability regime (Brix et al., 2004; Sourbron and Buckley, 2011, 2013).
36 The Two-Compartment
exchange model
 2CXM allows separate estimates of PS and Fp to be extracted from the data.
 The solution: convolution (⊗) of an impulse response function H(t) with arterial
plasma concentration (arterial input function, AIF).
 For a 2CXM, H(t) is bi-exponential with terms (α, β, A, Fp) that relate to
physiological parameters
37 The Two-Compartment
exchange model
 Figure 14.3 demonstrates how the more complex model can fit lesions that
follow different kinetic profiles.
 The choice of a model that best describes the acquired data depends on
several factors.
 generate inaccurate and inconsistent results, sources of errors, inappropriate
implementation of models
14.3 Data acquisition requirements
 T1-weighted images
 to detect the short-range effect of CA as it transits through the tissue on the T1
relaxation properties of water in tissue (or a lesion).
 The acquisition protocol plays a major role in determining the kinetic profile
of the CA and quantitative technique used for analysis.
 Some of the requirements for appropriate data acquisition:
14.3.1 From signal to concentration
 To monitor the behavior of a CA using quantitative analysis
 provide a link between changes in signal (relative to baseline) and the
concentration of the tracer.
 A simplified approach: to assume signal changes are directly proportional
to CA concentration
 Linearity breaks down at high concentrations.
 Degree of non-linearity depends upon acquisition parameters (e.g. flip
angle, repetition time, native T1 of the lesion).
 Usual approach: to measure baseline T1 relaxation and monitor T1 change
during dynamic acquisition.
 CA concentration C(t) may be related to change in T1 relaxation rate as
14.3.1 From signal to concentration
 T1 measurement must be fast and accurate, in order to sample the
concentration of the contrast agent at a high temporal rate.
 The measurement techniques need to be reliable over a wide range
of T1.
 1- Measure the baseline T10 using an established T1 mapping
 2- Dynamic T1 in a fast T1 weighted sequence
14.3.1 From signal to concentration
 T1 quantification: inversion recovery prepared imaging sequences or multiple
acquisitions with different repetition times (TR) or flip angles (θ).
 The most common: acquisition techniques in DCE-MRI: 3D-gradient echo
sequences, due to their short acquisition times.
 The signal intensity obtained from a commonly used gradient echo
sequence with spoiling of the transverse magnetization is
 where Ω depends on the scanner settings
14.3.2 Spatio-Temporal requirements
 Is crucial for tracer kinetic modelling and the choice of model as well as
the accuracy of quantitative analysis
 particular importance when sampling the passage of the CA bolus
through a supplying artery (known as the AIF),
 Temporal undersampling: introduce uncertainties into parameter
estimates , errors within 10% of the true value
 AIF should be sampled every 1 s and the tissue at least every 4 s.
14.3.2 Spatio-Temporal requirements
 The sampling should be shorted than the timescale of the processes to
be measured
 The mean transit time of Gd-DTPA through the capillary bed
 in grey and white matter: the order of 1–2 seconds
 in brain tumours (e.g. metastases, meningioma and lymphoma): the order of tens of
 obtain a measurement of CBF a temporal resolution of the order of seconds
 Conversely, the mean transit time of the same CAs in EES of tumors: the order of
 If the focus is BBB permeability the temporal resolution may be relaxed.
 For BBB permeability and ve measurements longer acquisitions are required
 Particularly in slowly enhancing lesions such as MS.
14.3.2 Spatio-Temporal requirements
 The need for high temporal resolution lead to adopt 2D or 3D
gradient echo techniques, parallel imaging, and other acceleration
 DCE data are usually acquired at the expense of spatial resolution.
 A compromise between temporal and spatial resolution
 Imaging volume needs to include a relatively large feeding artery or
vein from where the blood signal can be sampled.
14.3.3 Image contrast-to-noise ratio
 CNR is key for DCE-MRI data.
 A measure of change in signal intensity following contrast
administration referenced to baseline noise
 Depends on the tissue type (native T1)
 quality of the parameter estimates depends upon the quality of
signal–time series.
14.3.4 Arterial input function
 Accurate representation of the AIF is the most challenging part.
 Accuracy:
 delivered CA
 temporal sampling
 choice of model
 spatial resolution, partial volume effects in generated AIF
 propagates errors to the parameter estimates.
 problematic in DCE-MRI of the brain because the feeding arteries are
relatively small.
14.3.4 Arterial input function
 the AIF should be sampled directly at the inlet to the
 In practice it is performed further upstream in the arterial tree
where the diameter of the vasculature is larger.
 Majority of the studies: an AIF is sampled from the internal
carotid artery
 In some cases signal from a venous outflow. Figure 14.4.
14.3.4 Arterial input function
 Slice positioning is crucial in obtaining an accurate AIF.
 MRI planning: toacquire the signal from a feeding artery simultaneously
with lesion sampling.
 Positioning the most caudal slices orthogonally to the internal carotid artery or
acquiring separate slices at the level of the artery may be used to reduce partial
volume effects.
 The simplest way to account for that: assume an AIF derived from time
course of the mean blood plasma concentration from a number of
 Another proposed solution compares the tissue of interest curve to that
of a reference region, thereby eliminating need for direct AIF
14.3.5 Contrast agent administration
 Factor with an impact on the kinetic profile of the pathology
 1- administered dose,
 2- injection protocol.
 3- pharmacokinetic properties of the tracer
 CA partially bind to albumin should be avoided when BBB
permeability is the subject of investigation
 Choose The dose carefully depend on the dynamic range of
the MR sequence.
14.3.5 Contrast agent administration
The duration of the injection:
CAs are administered either as a bolus or as an infusion
 Bolus administration
 most common, higher sensitivity
 desirable duration is less than 10 s to generate a sharp AIF and
reduce error in the parameter estimates
 but this comes at the expense of requirement of a high
temporal resolution to capture the AIF.
14.4 Sources of error
14.4.1 Signal non-linearity
 Common assumption is that of a linear relationship. this is not true.
 Many studies adopt this approach, either due to acquisition
limitations. or the ease of using a direct relation between signal
changes and concentration.
 The non-uniformity of the transmit radiofrequency depends on
several factors
 location in the anatomy
 field strength
 non-ideal slab profile
 Can introduce errors both in the T1 measurement and in signal conversion to
dynamic T1
14.4.1 Signal non-linearity
 These errors will propagate into the analysis and create
a bias in the pharmacokinetic parameters.
 Another possible error may come from T2 * effects
 may have a significant impact at high CA concentrations, even when a
very short echo time is used.
14.4.2 Water exchange
 Water in different environments (i.e. intracellular, interstitial and intravascular)
has different MR properties, such as multiple T1 relaxation times.
 Water is not stationary, moves continuously from one environment to another, T1
measured is an average weighted by the time water resides in each subspace.
 The ratio of (water exchange rate between compartments) / (difference in
intrinsic T1 relaxation rates in each subspace)
 High: A single T1 value is a valid representation of system
 Low: (relaxation rate difference increases), the relaxation properties of each
subspace should be taken into account.
 Important for the first pass of the bolus, when CA remains intravascular with little or
no permeation into interstitium
 less relevant if the lesion under investigation is a tumour where the degree of first
pass extraction is higher.
 If significant, will lead to underestimated CBF and CBV.
 Methods to account for vascular– interstitial water exchange, in both modeling and imaging.
14.4.3 AIF Measurement
 A common issue: Inflow artefact
 causes an enhanced signal in arterial blood in precontrast images, underestimaging
the generated enhancement curve
 propagates into blood T1 Measurement
 leads to errors in CBF and blood-volume estimates.
 Error can be minimized
 by optimizing the acquisition protocol (e.g. choose a 3D acquisition over a 2D, use a
nonselective saturation prepulse)
 by appropriate positioning of the excitation slab such that the sampled artery runs for a long
distance through it.
 Another requirement: to sample the signal from a location close to the tissue in order to
minimize dispersion errors
 However, arteries in the brain are very small
 can lead to partial-volume errors, particularly when a high temporal sampling strategy is
required, at the expense of spatial resolution.
14.4.4 Haematocrit
 Concentration estimates in blood reflect the blood plasma
compartment, rather than whole blood, since the CA does
not enter blood cells.
 The AIF measurement needs to be corrected to describe
concentration in plasma (cp) rather than whole blood (cb),
using the blood haematocrit in cp = cb/(1–Hct).
 Measured Hct usually reflects the large vessel Hct.
 Small vessel Hct is also needed since it is a more accurate
representation of plasma volume in capillaries
14.4.5 Temporal resolution and scan
 To model CBF and CBV, a high temporal resolution is essential
and duration of acquisition can be reduced.
 If permeability and EES are the subjects of interest, longer
acquisitions are required and sampling rate can be relaxed.
 Simple model uses fewer parameters to describe the underlying
physiology  parameters are contaminated from processes
not incorporated into the model,
 resulting in systematic errors in pharmacokinetic estimates.
14.4.5 Temporal resolution and scan
 Temporal resolution requirements are particularly
problematic in brain pathologies DCE-MRI of a higher
spatial resolution is needed.
 A possible solution to this conflicting requirement is to
incorporate both high spatial and high temporal
resolution acquisitions in one single dynamic protocol.
14.4.6 Other sources of error
 1- point-spread functions due to k-space undersampling or
aliasing effects and are dictated by the imaging parameters.
 2- motion induced artefacts
 can be corrected by image registration and signal intensity drift,
which becomes more problematic in long acquisitions.
 3- The ROI selection process can introduce bias in the parameter
14.4.7 Reproducibility
 Crucial to incorporate reproducibility studies of normal and
pathological tissues into clinical research
 in order to assess the expected physiological variation, prior to
monitor treatment effects or the progression of a pathology.
 Numerous practical issues that need careful consideration and
design prior to a DCE-MRI study.
 The difficulty comes from the fact that most of these issues
have conflicting requirements and a compromise is inevitable.
14.5 Clinical applications

similar documents