11-04 Functional Imaging

Functional imaging is a misleading term because it is mainly used for the de­pic­tion of changes of local blood supply in the brain activated by specific stimuli. Commonly, dynamic or cine imaging of other organs or, e.g., joints are not de­scri­bed as functional MR imaging (fMRI). However, the term is not sharply de­fi­ned, and sometimes diffusion, perfusion as well as brain activation studies are subsumed under fMRI. Functional MR imaging of the brain, in contrast to EEG and MEG, does not provide a direct measure of neural activity.

In 1990, Belliveau and colleagues published the first observation of the sti­mu­la­tion of the human visual cortex by magnetic resonance imaging [⇒ Belliveau]. They watched the first pass effect of a contrast agent after bolus injection to de­mon­stra­te changes in cortical perfusion upon activation with a photic stimulus. The use of bolus tracking to study changes in perfusion was an exact analog to previous experiments using the observation of radioactive tracers with PET or SPECT. It required the injection of a contrast agent in two consecutive scans, one with and one without stimulus.

The performance of such an experiment with MR, instead of nuclear medicine techniques, offers the immediate advantage of vastly superior spatial and tem­po­ral resolution and the lack of radioactive tracers (cf. cerebral blood-volume studies and regional cerebral blood-volume described in Chapter 16). The need for dual contrast agent injection, nevertheless, poses a problem, especially for studies of brain activation in normal individuals.

This disadvantage was resolved by the demonstration of brain activation using the BOLD (blood-oxygen-level dependent) contrast mechanism first described by Ogawa [⇒ Ogawa]. This very ele­gant technique has led to a fast proliferation of fMRI in various centers over the last few years.

11-04-01 BOLD-Contrast

The basis for BOLD-contrast was described by Pauling and Coryell in 1936 [⇒ Pauling]. It relies on the fact that paramagnetic deoxyhemoglobin – by comparison to diamagnetic oxyhemoglobin – possesses a strong magnetic mo­ment. Thus, by interaction of the bulk magnetization of deoxygenated blood with the external field, local field variations in and around blood vessels are created. These susceptibility effects can be measured using appropriate MR imaging se­quen­ces.

The only source of energy of normal brain cells is the oxidation of glucose. Sin­ce the glucose storage capacity of brain cells is negligible, the brain very heavily depends on a constant supply of glucose and oxygen via the capillary bed. This increased demand appears to lead to an increased amount of blood flowing to the activated area. This in turn decreases the local susceptibility effect, which can be visualized using appropriate susceptibility-sensitive imaging techniques (Figure 11-13).

Figure 11-13: BOLD-contrast. The presence of deoxyhemoglobin in a capillary causes a susceptibility difference between the blood vessel and the neighboring tissue. It induces a dephasing of the spins, thus a decrease in T2* and signal loss on T2-/T2*-weighted images.

Susceptibility differences are greater at higher fields, and thus higher fields are desirable for this kind of studies.

For the first brain activation studies Kwong applied gradient-echo-planar ima­ging (GRE-EPI) [⇒ Kwong]. The EPI sequence uses multiple gradient refocusing to acquire all data necessary for image reconstruction after a single excitation pul­se. In spite of its not very well defined signal behavior, EPI has turned out to be a very efficient technique for brain activation studies due to its short ac­qui­si­tion time.

Conventional gradient-echo imaging with long echo times (40-60 ms, de­pen­ding on field strength) has also turned out to be a useful technique for fMRI [⇒ Frahm]. Its advantage over EPI lies in the fact that it allows the acquisition of high-resolution images, whereas the resolution in EPI is determined roughly by the number of echoes which can be acquired within the T2 of brain parenchyma.

Conventional gradient-echo imaging does, however, suffer from a number of severe drawbacks. The long acquisition time per image restricts the application to a single slice and thus requires prior knowledge about the area of activation. Partial volume effects can lead to difficulties in data interpretation.

Gradient-echo techniques also are very sensitive to inflow. Since vascular flow – especially in large veins – also changes upon stimulation, this can lead to the measurement of activation effects many centimeters from the area of activation [⇒ Segebarth]. These vascular signal changes can be much larger than the actual parenchymal effects, which seldom exceed 2-3%.

The image quality of all susceptibility-sensitive techniques is strongly de­pen­dent on macroscopic susceptibility problems occurring especially at soft tissue – bone – air interfaces, leading to magnetic field inhomogeneities over several cen­ti­me­ters. These long-range effects will cause image distortions when oc­cur­ring in the direction of the readout gradient, which is normally of no practical con­se­quen­ce.

Field inhomogeneities across the selected slice, however, will lead to signal attenuation and thus severely affect the image quality. The use of thin slices (or 3D data acquisition) is, therefore, to be preferred for fMRI. The strength of the stimulation effect will, of course, not be dependent on the slice thickness due to the small range of the BOLD effect (Table 11-02).

Table 11-02: BOLD-contrast. Limitations of monitoring of brain activity with BOLD imaging.

Applications. The first experiments performed with fMRI used the well- known paradigm of photic stimulation with an alternating checkerboard pattern or a flicker display. This is known to lead to significant changes in perfusion and thus serves as a test tool for sequence development.

Meanwhile, quite a number of experiments have been performed, which led to new insights in neurocognitive research. Apart from activation in the primary vi­su­al cortex, activation of associated areas was demonstrated using a number of paradigms to test cognitive processing of motion, texture, color, object re­cog­ni­tion, sound, memory, and others (Figure 11-14). Various paradigms using motor activation have been successfully examined. Numerous groups have in­ves­ti­ga­ted language processing using a number of well established paradigms. In ad­di­tion to activation of the cerebral cortex, the involvement of the cerebellum in lear­ning tasks has been demonstrated. Subcortical activation has been found, for example, in the nucleus geniculatus (upon visual stimulation).

Nomenclature. In some articles on fMRI, NMR/MRI terms are used wrongly or confusingly, for instance T2* for apparent T2 (T2app), and T1* (a term which is not appropriate because T1 is not affected by susceptibility effects) for an apparent T1 (T1app or T1influx).

Figure 11-14: Working memory test: typical activation pat­tern in the parietal cortex; cognitive/ speech processing dorsolaterally.

Critical remarks. Unfortunately, BOLD imaging at common (high and ul­tra­high) fields strengths for investigative fMRI, such as 1.5 or 3.0 T, has a very low sensitivity and signal-to-noise ratio. The signal changes related to cerebral ac­ti­va­tion are close to the noise level and therefore numerous signal processing tech­ni­ques are used to overcome it. More so, T2* to estimate blood oxygen sa­tu­ra­tion is only one singled-out factor; oxygen supply and saturation are de­pen­dent on several additional and independent parameters, among them lung and heart function, vessel size, and hematocrit. At the present stage this means that some groups look again – after the first description in 1990 by Belliveau et coll. [⇒ Belliveau] – into exogenous agents , e.g., manganese, to highlight he­mo­dy­na­mic changes in the brain [⇒ Chen, ⇒ Christen, ⇒ Kim and Ogawa, ⇒ Leite].

fMRI tickles the imagination of researchers, as well as the laity of all medical and paramedical disciplines including neuro­economics and neuro­marketing, and the population at large because it shows the brain at work and reacting to the en­vi­ron­ment in beautiful color images.

In a very catching paper, Bennett and collaborators show that it is pivotal in fMRI to apply statistics properly and scrupulously because random noise in the EPI time series may yield spurious results if multiple testing is not controlled for [⇒ Bennett]. To date, the results presented in thirty percent of all fMRI papers, most likely even far more, are wrong.

Pitfalls and snags of the technique and of the interpretation of the outcome are copious. The article by Haller and Bartsch might be worthwhile reading [⇒ Haller] – as well as the comments mentioned below. More than 25 years after its first description fMRI remains an immature technique. Among a long list of other things, Ogawa pointed out in a critical review in 2012 [⇒ Kim and Ogawa]:

"Dynamic properties and magnitudes of BOLD functional responses are dependent on many physiological parameters as well as baseline conditions. In patients with neuro­vascular disorders, the BOLD response could be sluggish, or even decreased relative to baseline. This should not be interpreted simply as a decrease in neural activity, because neuro­vascular coupling may be hampered. Resting­state fMRI studies are widely performed, but its physiological source needs to be systematically investigated."

Some BOLD or, perhaps, bold comments
about functional imaging and applications of fMRI.

BOLD, bolder, the boldest ...
Another comment

fMRI and "Big Science": a billion euro debacle?
A (final?) comment