Brain–Computer Interfaces - Index of

164 N. Birbaumer and P. Sauseng
of the BOLD response (BOLD – blood oxygen level dependent) with a delay of
3s from the neuronal response to the hemodynamic change (see [10, 23, 36]). The
laboratory of the author reported the first successful and well-controlled studies of
subcortical fMRI-BCI systems (see [36] for a summary). Figure 4 shows the effect
of operant conditioning of the anterior insula within 3 training sessions, each session
lasting 10 min. Subjects received feedback of their BOLD response in the Regionof-interest
with a red or blue arrow, red arrow indicated increase of BOLD in the
respective area relative to baseline, the blue arrow pointing downwards indicated a
decrease of BOLD referred to baseline. Before and after training subjects were presented
with a selection of emotional pictures from the International Affective Picture
System (IAPS [21]) presenting negative and neutral emotional slides. As shown in
Fig. 4 subjects achieved a surprising degree of control over their brain activity in
an area strongly connected to paralimbic area, being one of phylogenetically oldest
“cortical” areas generating mainly negative emotional states.
Figure 4 also shows that, after training, subjects indicated a specific increase
in aversion to negative emotional slides only. Neutral slides were not affected by
the training of BOLD increase in the anterior insula, proving the anatomical specificity
of the effect. The rest of the brain was controlled for concomitant increases
or decreases and it was demonstrated that no general arousal or general increase or
decrease of brain activity is responsible for the behavioral effects. In addition, two
control groups one with inconsistent feedback and another with instructions to emotional
imagery only did not show a learned increase in BOLD in the anterior insula
nor behavioral emotional valence specific effects on emotion.
Conditioning of subcortical areas such as the anterior cingulate and the amygdala
were also reported. Ongoing studies investigating the possibility of increasing
and decreasing dynamic connectivity between different brain areas by presenting
feedback only if increased connectivity between the selected areas is produced voluntarily
are promising. Most behavioral responses depend not on a single brain
area but on the collaboration or disconnection between particular brain areas;
behavioural effects should be much larger for connectivity training than for the
training of single brain areas alone.
Two clinical applications of the fMRI-BCI were reported: DeCharms et al. [11]
showed effects of anterior cingulate training on chronic pain and our laboratory
showed that criminal psychopaths are able to regulate their underactivated anterior
insula response (see [6]). Whether this has a lasting effect on the behavioral
outcome of criminal psychopaths remains to be demonstrated. However, the direction
of research is highly promising and studies on the conditioning of depression
relevant areas and on schizophrenia are on the way in our laboratory.
The remarkable ease and speed of voluntary control of vascular brain response
such as BOLD suggests a superior learning of instrumental control for these
response categories. EEG, spike trains, and electrocorticographic activity all need
extended training for voluntary control. Vascular responses seem to be easier to
regulate, probably because the brain receives feedback of the dynamic status of the
vascular system, and these visceral perceptions allow for a better regulation of a nonmotor
response which can not be detected by the brain in neuroelectric responses.