Professional Documents
Culture Documents
Birbaum Et Al 1999 - Spelling Device For The Paralysed
Birbaum Et Al 1999 - Spelling Device For The Paralysed
Birbaum Et Al 1999 - Spelling Device For The Paralysed
HOFFENTLICH-KOMMEN-SIE-MICH-BESUCHEN,-WENN-DIESER-
training training Copy spelling Free spelling
100 BRIEF-SIE-ERREICHT-HAT-.ICH-DANKE-IHNEN-UND-IHREM-
..
sensory and cognitive functions, he TEAM-UND-BESONDERS-FRAU-KUBLER-SEHR-HERZLICH,-DENN-
..
activate communication devices. We have 80 For the spelling program at level 1, the
developed a new means of communication alphabet was split into two halves (letter
60
for the completely paralysed that uses slow banks) which were presented successively at
cortical potentials (SCPs) of the electro- 40 the bottom of the screen for 4.5 seconds
encephalogram to drive an electronic spell- (subject A) or 6 seconds (subject B). If the
20
ing device. subject selected the letter bank being shown
The neurophysiological basis and behav- 0 by generating a SCP, it was split into two
0 100 140 180 220 260 300 340
ioural functions of SCPs have been new halves, and so on until each of the two
Sessions
described2, and operant conditioning has letter banks had only one letter in it. When
been used to bring them under voluntary one of the two final letters was selected, it
control. We have shaped the voluntary con- Figure 1 Response accuracy of subjects using the was displayed in the top text field of the
trol of SCPs in two locked-in patients with new spelling device. Subject A began with feed- screen and then the selection began again at
amyotrophic lateral sclerosis who were able back training of SCP amplitude (initial and level 1. A ‘go back’ function, which
to learn to operate a spelling device by regu- advanced training; 71.3% correct selections, 75.0% appeared as an option when two successive
lating their brain responses. The spelling correct rejections, based on ‘go-back’ responses), letter banks had not been selected, allowed
device drives a cursor on a video screen, proceeded to copy spelling (copying of letters and the speller to go back one step to the previ-
allowing the subjects to select letters of the then words; 78.7%, 75.3%) and finally to free ous set of letter banks. If the speller was at
alphabet. Previous interfaces between spelling (self-selected letters; 66.4%, 82.9%). Sub- level 1, selecting this function erased the last
brains and computers have used different ject B began with initial training, then switched to a symbol written in the text field, so mistakes
brain responses, such as certain frequency combination of advanced training (77.5%, 68.8%) could be corrected. This procedure was
bins or event-related brain potentials3,4, but and copy spelling (77.5%, 67.6%) and finally to free chosen after extensive pilot work6 involving
these have not yet been tested with locked- spelling (86.2%, 73.7%). different types of speller.
in patients. The accuracy of responses during feed-
Both subjects suffer from advanced by a box being highlighted in either the back training, copy spelling and free
amyotrophic lateral sclerosis and have been upper (negativity) or the lower (positivity) spelling is shown in Fig. 1 for both patients
artificially respirated and fed for four years. half of the screen. Visual feedback of SCPs, over the course of 446 (subject A) and 308
Sustained voluntary control of the muscula- which was updated every 64 ms, was pro- sessions (subject B). Two types of error
ture was not possible in either patient, so vided by a ball that moved towards or away were possible: incorrect rejection was due to
they were unable to use a muscle-driven from the box, depending on the direction the SCP amplitude being too low, and
communication device. They were trained in which the SCP deviated from baseline. incorrect selection occurred when a high
to produce changes voluntarily in their SCPs The response criterion was progressively SCP was made in the presence of a wrong
lasting 2–4 seconds. Each trial consisted of a increased from 5 to 8 mV. letter or row of letters. Both subjects can
2-second baseline and a response period SCPs were extracted from the regular now write messages. The first full message
lasting 2–4 seconds, the length of which was electroencephalogram using a time con- written entirely by the brain of subject A is
adapted to suit the patient. A training day stant of 8 s and a low-pass filter of 40 Hz. shown in Fig. 2.
consisted of 6–12 sessions (depending on They were recorded from the vertex rela- Our data indicate that patients who lack
the condition of the patient), each of which tive to linked mastoids at a sampling rate muscular control can learn to control vari-
comprised about 70 to 100 trials and lasted of 256 Hz. Vertical eye movements were ations in their SCPs sufficiently accurately
about 5–10 minutes. simultaneously recorded with standard to operate an electronic spelling device.
During the response period, the subjects on-line removal of eye-movement arte- Although writing sentences is time con-
were required to produce either negativity facts2. Using an imagery strategy5, both suming — subject A took 16 hours to write
or positivity greater than a specific criterion patients were better able to produce posi- the message in Fig. 2, a rate of about 2
amplitude in random order. The baseline tivity than negativity, so training for nega- characters per minute — it is reliable and
and response periods were signalled by two tivity was soon discontinued. As soon as a precise enough to allow the patient to com-
clearly different tones, and whether positiv- stable performance of at least 75% correct municate with his or her environment.
ity or negativity was required was indicated (hitting the goal for at least 500 ms) These completely paralysed individuals
NATURE | VOL 398 | 25 MARCH 1999 | www.nature.com © 1999 Macmillan Magazines Ltd 297
scientific correspondence
now have the ability to communicate, a
15 2 —6 1
possibility that has not previously existed
Relative sensitivity
d
EAD c
298 © 1999 Macmillan Magazines Ltd NATURE | VOL 398 | 25 MARCH 1999 | www.nature.com