TwinTree Insert

16-03 Dynamic Image-Processing


or a long time, one of the main objectives researchers wanted to achieve was the improvement of MR image contrast by means of possible electronic con­trast agents without the application of any exogenous medium. Today these methods are also known as fingerprinting or biomarkers (see also Chapter 4).

Pure relaxation-time ima­ges and, at a later stage, segmented images were the first re­search targets of this kind of MR image-processing.

However, this notion of the existence of electronic contrast agents that can be em­ploy­ed to highlight pathologies through image-processing derives from a wrong hypo­thesis, namely that there is hidden information about tissue struc­tu­res or pro­ces­ses in the original images or raw data.

The additional information given by a pharmaceutical contrast agent, i.e., vas­cu­la­ri­ty, mem­bra­ne per­mea­bi­li­ty, etc., is not given by any plain imaging modality and there­fore can­not be elec­tro­ni­cal­ly enhanced; if the contrast-to-noise ratio is zero in the raw data, no contrast enhancement by image-processing will be possible.


spaceholder redOn the other hand, it might be possible to enhance minimum quantities of con­­trast agent uptake, which create only minor contrast changes if appropriate image pro­ces­sing techniques are applied.

Since the temporal uptake pattern of contrast agents in vivo can vary between heal­thy and diseased tissue, differences in uptake can be of diagnostic value. In sig­nal intensity-versus-time curves, some lesions reveal steeper slope, higher ma­xi­mum sig­nal intensity, and faster wash-out of the contrast agent than the sur­round­ing tissue [⇒ Gribbestad 1994, ⇒ Kaiser 1990, ⇒ Tofts 1991].

A factor contributing to the uncertainty of quantitative data acquired with dy­­na­­mic ima­ges is the wide scatter of blood volumes and transit times in different or­gans, or even within similar organ structures.

In the brain, blood volumes and transit times tend to be higher in the pons, ce­re­bel­lum, and medulla than in the midbrain and forebrain, which suggest some ge­ne­ral re­gi­o­nal differences in mi­cro­vas­cu­la­ri­ty [⇒ Nakagawa 1995]. These differences can be substantial and may over­lap with the values of pathologies. Proper know­ledge of the normal range of ana­to­my and physiology is essential for the final as­sess­ment of data or calculated pa­ra­met­ric images.


spaceholder redAs a supplement and replacement of analog contrast-enhanced cine fluo­ro­sco­py, dy­na­mic di­gi­tal imag­ing to highlight these additional contrast parameters was de­ve­lop­ed for nuclear medicine and CT. Dynamic PET scans for regional ce­re­bral blood flow measurements, for instance, can be performed at intervals of several seconds. How­ever, radioisotope methods lack good temporal and spatial resolution, whereas CT is fast and shows anatomical structures in detail.

Conventional MR imaging is a slow imaging technique (Figure 16-05). All of the clas­si­cal imag­ing sequences have long examination times. In the late 1980s ra­pid imag­ing pulse sequences became available. They made it possible to follow the dynamic signal-intensity changes after contrast agent injection.


Figure 16-05:
A graphic depiction of the time scale of different events in man versus MR imaging methods. Imaging speed must be 3-10 times faster than the event to be monitored or avoided.
* = of contrast agent.


The prerequisite of sufficient spatial and time resolution was given with rapid imag­ing, and the uptake of contrast agents could be readily monitored by MR imaging (cf. bolus injection in CE-MRA).

Maximum vascular concentration of an ECF-space contrast agent will occur du­ring the first pass through the body in the vascular phase. Maximum tissue con­cen­tra­tion, as well as enhancement rate and onset, depend on a number of not yet com­ple­te­ly under­stood factors such as vascularity, membrane per­mea­bi­li­ty, and leakage of the blood-brain or the tumor barrier, and venous outflow.

Thus, maximal tissue concentration may occur between several seconds and mi­nu­tes after the injection of the contrast agent (Figure 16-06). The agent then diffuses into the extracellular space until an equilibrium is reached during the plateau phase.

Concentration decreases during the wash-out phase which fol­lows.


Figure 16-06:
Bolus injection of contrast agent: rapid in­tra­vas­cu­lar uptake after bolus injection leading to first-pass phe­­no­me­non versus relatively slow extracellular uptake in easily accessible regions.
SI = signal intensiy.


The enhancement processes can be monitored by time-intensity curves. Although both the vascular and the extracellular phase of uptake go hand in hand, one must dis­tin­gu­ish mo­ni­tor­ing the dynamic signal-intensity changes of the first pass of a con­trast agent bolus such as in brain and heart imaging and the slower tissue uptake as it is commonly done in breast lesions.


spaceholder redThe time resolution required to study the dynamics of the contrast agent up­take in an organ depends on the requirements of the individual investigation. Better time resolution will improve the characterization of the bolus while pas­sing through the body region-of-interest, in particular if these images are to be treated by mathe­ma­ti­cal techniques to analyze bolus behavior [⇒ Jones 1993].

Such mathematical treatment that, at the end, can lead to parametric images is be­yond the capability of the human brain, which is unable to make a cor­re­la­tion if the num­ber of features exceeds three.

Such differences can only be distinguished either by a qualitative description of the change in signal intensity observed over time or by the quantification of all avail­able or certain selected parameters by fitting the concentration-time cour­se to a pharmacokinetic model [⇒ Larsson 1994, ⇒ Tofts 1991].

Explanations of the parameters are given in Figure 16-07, and an overview of qua­li­ta­ti­ve or semi-quantitative parameters commonly determined by time ver­sus in­ten­si­ty curves is given in Table 16-02.


Figure 16-07:
Parameters of bolus injection of a T1 contrast agent. The maximum wash-in slope is considered one of the most important parameters, indicative of, e.g., tumor vascularity. It can be calculated as follows: (SIend - SIstart / SIbaseline × Δt) × 100 [% / s].
SI = signal intensiy.


Table 16-02:
Some parameters in dynamic imaging. Note: there is no general agreement on terms and no­men­cla­ture — and there is no agreement on the diagnostic implica­tion of these pa­ra­me­ters.


A number of soft­ware systems offer an integrated, standardized way of performing both image processing and image analysis on all kinds of dynamic images in­de­pen­dent of the imaging equipment on which the images were acquired in a stan­dar­di­zed way [⇒ Torheim 1997, 1999].


spaceholder redIn this context, it is important to remember that MR imaging has no standard units for signal intensity: it cannot be measured in units comparable to Houns­field units in CT. Several attempts have been made to create a quantitative scale, but none of these methods creates an absolute standard.

Thus, curves depicting signal intensity versus time remain only qualitative re­­pre­­sen­­ta­­tions of contrast passage through tissues.

To acquire quantitative data, con­cen­tra­tion versus time curves have to be fitted by the principles of tracer dilution ki­ne­tics. Time-intensity curves of bolus injections have to be corrected, i.e., clean­ed of the influence of recirculation to achieve bet­ter raw data for parameter cal­cu­la­tion, e.g., by gamma-variate fitting [⇒ Gonzalez 2008, ⇒ Sebastiani 1996, ⇒ Thompson 1964].

The concentration (C) of the contrast agent in each pixel, at each time point, can be calculated from the known relaxivity of the contrast agent and the re­­spec­­ti­­ve re­­la­­xa­­tion times, T1 and T1₀:


C = (R1 - R1₀) / r1

where R1 = 1/T1 at the measured point during the dynamic study, R1₀ = 1/ T1₀ prior to injection, and r1 is the longitudinal relaxivity of the contrast agent at the given magnetic field strength and 37° C.


The resulting concentration-time curves can be fitted to pharmacokinetic mo­dels.

From this, values of permeability and leakage volume are de­ri­ved. For extended image processing and data analysis, independent soft­ware systems running on per­so­nal com­pu­ters have been developed.