Perfusion & Angiography

Imaging the Microvasculature and the Macrovasculature

MRI can interrogate blood at two very different spatial scales. Perfusion imaging quantifies microvascular delivery of blood to tissue, on the order of microliters of blood per gram of tissue per minute, by tracking either an exogenous contrast agent or magnetically labeled water as it passes through the capillary bed. MR angiography (MRA) instead depicts the lumen of named arteries and veins, exploiting the bulk motion of spins through the imaging plane. Although they share an underlying theme, that flowing blood carries magnetization, the two families differ profoundly in what they measure, what they assume, and how they are quantified.

This lesson develops three perfusion methods, dynamic susceptibility contrast (DSC), dynamic contrast enhanced (DCE) imaging, and arterial spin labeling (ASL), and three angiographic methods, time-of-flight (TOF), phase-contrast (PC), and contrast-enhanced (CE) MRA. For each we derive the governing relationship and attach representative numbers so you can reason quantitatively about acquisition choices and pitfalls.

Dynamic Susceptibility Contrast (DSC)

In DSC a compact bolus of gadolinium chelate (typically 0.1 mmol per kg injected at 4 to 5 mL per second) is tracked with rapid T2-weighted* gradient-echo or spin-echo EPI, one whole-brain volume every 1 to 2 seconds. While gadolinium is confined to the vasculature, the large susceptibility difference between intravascular contrast and surrounding tissue creates microscopic field gradients that dephase spins throughout the voxel, producing a transient signal drop rather than the T1 brightening one might expect. The key insight is that the susceptibility effect extends well beyond the vessel wall, so even though blood is a small volume fraction, the whole voxel darkens.

Because the T2* signal decays exponentially, the relaxation-rate change is proportional to tracer concentration. We convert the measured signal $S(t)$ to a concentration-time curve via the change in $R_2^* = 1/T_2^*$:

With a concentration curve in hand, tracer-kinetic theory gives the hemodynamic maps. CBV is the area under the tissue concentration curve normalized to the area under an arterial input function (AIF), a concentration curve sampled in a feeding artery such as the middle cerebral or internal carotid:

CBF requires deconvolution. The tissue curve is the convolution of the AIF with the tissue residue function $R(t)$, the fraction of tracer still present at time $t$ after an ideal instantaneous input, scaled by flow: $C_{\mathrm{tissue}}(t) = \mathrm{CBF} \cdot (C_{\mathrm{AIF}} \otimes R)(t)$. Recovering $\mathrm{CBF} \cdot R(t)$ by deconvolution (commonly singular value decomposition, SVD) gives CBF as the peak height of that function, while MTT follows from the central volume theorem as $\mathrm{CBV}/\mathrm{CBF}$.

Dynamic Contrast Enhanced (DCE) Imaging

Where DSC exploits the T2* susceptibility effect of intravascular gadolinium, DCE uses heavily T1-weighted spoiled gradient-echo imaging to watch contrast leak from plasma into the extravascular extracellular space (EES). The acquisition is slower (a few seconds per volume) but is acquired over several minutes, capturing the full wash-in and wash-out of a leaky vessel. DCE is the workhorse of body perfusion, prostate, breast, and tumor permeability assessment, where capillary leak rather than blood volume is the question of interest.

The standard analysis is the Tofts model, a two-compartment exchange between blood plasma and the EES. The rate of contrast accumulation in tissue is governed by the volume transfer constant $K^{\mathrm{trans}}$:

$K^{\mathrm{trans}}$ (units of per minute, often 0.1 to 0.5 in aggressive tumors versus near zero in normal brain) reflects the product of capillary permeability and surface area. In the permeability-limited regime it equals the permeability-surface-area product $PS$ per unit volume; in the flow-limited regime, where leak is fast relative to delivery, $K^{\mathrm{trans}}$ approaches plasma flow. A defining strength of DCE is the explicit separation of these microvascular parameters; a defining weakness is its sensitivity to the chosen AIF and to accurate baseline $T_1$ mapping, since signal must first be converted to absolute concentration through the spoiled gradient-echo signal equation.

| Feature | DSC | DCE | | --- | --- | --- | | Weighting | T2* (susceptibility) | T1 (relaxivity) | | Signal change | Transient drop | Progressive rise | | Temporal resolution | 1 to 2 s | 3 to 8 s | | Primary parameters | CBV, CBF, MTT | Ktrans, ve, vp | | Best for | Brain perfusion, stroke | Permeability, tumor, body | | Main confound | BBB leakage (T1 effect) | AIF and baseline T1 errors |

Arterial Spin Labeling (ASL): Contrast-Free Perfusion

ASL needs no exogenous tracer. Instead, inflowing arterial blood water is magnetically labeled by an inversion or saturation RF pulse applied below the imaging slab. After a post-labeling delay (PLD) that allows labeled spins to flow into the tissue and exchange with the extravascular pool, a label image is acquired. A second control image is acquired identically but without effective labeling. The static tissue is the same in both, so the control minus label subtraction isolates the tiny signal contributed by delivered blood water.

The catch is magnitude. The perfusion signal is only about 0.5 to 1.5 percent of the tissue signal, so dozens of label-control pairs must be averaged. Quantification uses a single-compartment kinetic model relating the fractional difference signal to CBF:

Two labeling schemes dominate. Pulsed ASL (PASL) inverts a thick slab of blood with a single adiabatic pulse, giving a sharp but short-lived bolus with a poorly defined trailing edge. Pseudo-continuous ASL (pCASL) uses a long train of short RF pulses synchronized with a gradient to invert blood continuously as it flows through a labeling plane, achieving higher labeling efficiency and SNR. pCASL is now the consensus recommendation for clinical brain perfusion. The exponential $T_{1b}$ terms in Eq. 10.4 make ASL inherently sensitive to arterial transit time: if the PLD is shorter than the time blood needs to arrive, labeled spins are still in the arteries and CBF is underestimated, a pitfall in elderly patients and steno-occlusive disease.

Use the viewer below to inspect a real pCASL perfusion volume alongside a time-of-flight MRA, bridging the perfusion and angiography halves of this lesson.

MR Angiography: The Physics of Flow

Time-of-Flight (TOF): Flow-Related Enhancement

TOF MRA relies on flow-related enhancement. A spoiled gradient-echo sequence with a short TR and a moderate-to-high flip angle repeatedly excites a slice or slab. Stationary tissue is excited again and again before it can recover longitudinal magnetization, so it becomes saturated and dark. Fresh, fully relaxed blood flowing into the slice has experienced no prior RF and therefore appears bright. The degree of saturation of stationary spins follows the steady-state spoiled gradient-echo behavior:

Maximum vessel-to-background contrast requires that blood fully refreshes the slice between excitations: ideally the slice thickness divided by blood velocity is less than or equal to TR. This is why 2D TOF uses thin slices and is excellent for slow venous flow, while 3D TOF images a thick slab with high spatial resolution but suffers progressive saturation of blood as it travels deeper into the slab. Two engineering fixes address the deep-slab problem: a ramped (TONE) RF pulse that applies a lower flip angle at the slab entry and a higher flip angle at the exit to equalize signal, and multiple overlapping thin-slab acquisition (MOTSA), which divides the slab into thinner, less-saturating chunks.

Phase-Contrast (PC): Velocity Encoding

Phase-contrast MRA encodes velocity directly into the phase of the signal. A balanced bipolar gradient imparts zero net phase to stationary spins but a net phase to moving spins proportional to their velocity along the gradient. Acquiring the data twice with opposite bipolar polarities and subtracting cancels background phase and yields a phase map linear in velocity:

The user-selected velocity encoding (VENC) is the velocity that produces a phase of plus or minus 180 degrees, mapping the full available phase range. Choosing VENC is a quantitative balance: set it too high and slow flow occupies only a sliver of the phase range, giving noisy velocities; set it too low and faster flow wraps past 180 degrees, producing aliasing that appears as abrupt black-white reversals within a vessel. A typical carotid VENC is 60 to 80 cm per s; cerebrospinal fluid flow studies use a few cm per s. PC also provides quantitative flow by integrating velocity over the lumen cross-section, which TOF cannot.

Contrast-Enhanced (CE) MRA

CE-MRA abandons flow phenomena and instead shortens blood $T_1$ directly. A gadolinium bolus drops intravascular $T_1$ from about 1600 ms to well under 100 ms, so a fast 3D spoiled gradient-echo sequence with a very short TR renders blood markedly brighter than all background tissue regardless of flow direction or velocity. Because contrast is fleeting, success hinges on bolus timing: the center of k-space, which dictates image contrast, must be sampled during peak arterial enhancement and before venous return. Timing is achieved with a test bolus, automated bolus detection (fluoroscopic triggering), or time-resolved acquisitions such as TWIST or TRICKS that trade spatial for temporal resolution.

Flow Phenomena: A Unifying Lens

Three recurring flow effects explain both the strengths and the artifacts of all these methods. Inflow (flow-related) enhancement is the brightening of unsaturated blood entering a slice, the very basis of TOF and a cause of incidental vessel brightness on any short-TR gradient-echo scan. Flow voids are the signal loss seen when spins excited by the 90-degree pulse of a spin-echo sequence move out of the slice before the 180-degree refocusing pulse, leaving rapidly flowing blood black on conventional spin-echo images. Intravoxel dephasing occurs when a spread of velocities within one voxel, from turbulence or shear near vessel walls, produces a distribution of phases that sum destructively, attenuating signal; this is mitigated by short TE and by flow-compensating gradient designs that null the first moment for constant-velocity flow.

| Method | Contrast mechanism | Needs gadolinium | Quantitative output | | --- | --- | --- | --- | | DSC | T2* susceptibility drop | Yes | CBV, CBF, MTT | | DCE | T1 relaxivity rise | Yes | Ktrans, ve | | ASL | Tagged blood water | No | Absolute CBF | | TOF | Inflow enhancement | No | Lumen depiction only | | PC | Velocity-induced phase | No | Velocity and flow | | CE-MRA | T1 shortening of blood | Yes | Lumen depiction only |

Mastering perfusion and angiography is ultimately about asking what is being measured. Is it the volume of blood in a voxel (CBV), the rate of its delivery (CBF), the leakiness of its vessels ($K^{\mathrm{trans}}$), its velocity ($\Delta\phi$), or simply where the lumen runs? Matching the question to the mechanism, and knowing the assumptions each method makes, is the difference between a number on a map and a clinically defensible measurement.