Detecting Consciousness with MEG: A selective attention paradigm for communicating with immobile (locked-in) children and adults Egeth, M1, Rey, M1, Monroe, J1, and Roberts, T1 1

The Children’s Hospital of Philadelphia, 19146 Philadelphia, USA [email protected]

ABSTRACT How can we tell if an unmoving person is conscious? Owen et al’s1 method of looking for normal fMRI brain responses to complex tasks (imagine playing tennis vs. imagine walking through a house) has been criticized on the grounds that processing does not necessary imply consciousness. Egeth2 proposes that communication requires volitional brain activity and thereby implies consciousness: if a person can alter his brain state in order to answer a question, such as by imagining playing tennis to answer yes or imagining a house to answer no, and proceeds to answer questions normally, then we would be hard pressed to deny the attribute of consciousness to that person. In order to use communication to establish consciousness, rather than finding complex mental tasks for patients to engage in, we can instead search for relatively simple tasks that a patient can use as a code to communicate “yes.” Many extant tasks could be adapted for this purpose; here, we ask neurotypical subjects to alternately listen to and ignore blocks of soft tones (Experiment 1), and one subject attempts to selectively listen in order to deliberately communicate a message (Experiment 2). Experiment 3 describes a communications paradigm intended specifically for children, Essentially, if we can detect that a person has made a choice, we can infer that the person is conscious. KEYWORDS: MEG, BCI, Turing Test, Consciousness, Vegetative, Coma, Comatose, Locked-In, Auditory, Attention

INTRODUCTION Brain-computer interfaces (BCIs) have had exciting but limited success allowing locked-in (conscious but immobile) adults to communicate, and none have been reported to work for locked-in children. Most BCIs require extensive practice. In addition, in spite of Owen et al’s1 report that some might be conscious; the field has not reported successful communication with people diagnosed as comatose or vegetative. Here, we hope to provide new, user-friendly methods of communication that do not require practice, that children could use, and that might be used to identify consciousness among seemingly unconscious comatose or vegetative people. We focus on early evoked neural responses that can tell us where a subject has chosen to pay attention and which children are known to exhibit.

METHODS Experiment 1 MEG recordings from a CTF 275-channel whole-head biomagnetometer were collected. Three subjects experienced 16 blocks of 50 200ms 200Hz tones in the right ear with an 800ms +-10ms ISI and 10 seconds between blocks; subjects were asked to close their eyes during the experiment to minimize eye-blink artifacts. Before each block, subjects were asked to either ignore or listen to the tones; the ignore/attend instruction alternated with each block, starting with ignore, resulting in 400 trials of each condition per subject. Experiment 2 The third subject from Experiment 1 returned for a subsequent session to attempt to communicate a concept (a vegetable of her choosing) through yes/no answers. The experimenter asked “yes or no” questions from the neural network 20-questions website,, which plays the game “Animal, Vegetable or Mineral?” Each question was followed by two blocks of 50 tones. The subject was instructed to listen to the first block of tones to indicate “yes” or ignore the tones to indicate “no,” and then to switch tasks for the 2nd block. This method provides a contrast condition for each question: if the first block is attended, the answer is “yes”; if the second block is attended, the answer is “no.” CTF real-time MEG data outputs were configured to display averages for “Condition 1” (the first block) and “Condition 2” (the second block). Subject 3 showed more time-locked activity during the attended condition in Experiment 1; for Experiment 2, the experimenter (M.E.) provided with “yes” and “no” responses by making an estimate after each block whether Condition 1 or Condition 2 showed more such activity and, thus, revealed the subject to be paying attention. Experiment 3 Subject 1 from Experiment 1 listened to dichotic streams of a female saying “yes” (left ear) and a male saying “no” (right ear) with loudness deviants randomly distributed


within each stream; the subject alternated blocks attending to the yeses and the nos. This method creates a more natural code by which to answer “yes” and “no:” instead of attending to tones, subjects can listen to the yeses to communicate “yes” or listen to the nos to communicate “no.” Attentional mediation of the mismatch field is expected to distinguish between a subject attending to the yeses versus attending to the nos; such effects have been reported for both adults and children.3 RESULTS Experiment 1 All subjects showed recognizable time-locked auditory responses to tones (M100) for both attended and ignored conditions. In addition, all subjects showed time-locked differences to attended versus ignored tones in right posterior brain regions, both prior to and following the tone (see Fig. 1 for difference waves and Fig 2. for one subject’s “attended” and “ignored” time-locked waveforms.)

Experiment 2 After starting with “vegetable” in the game “Animal, Vegetable or Mineral,” provided “yes or no” questions in an iterative manner based on the experimenter’s inputs (see Table 1). After 18 yes/no responses provided by the experimenter, guessed “lime” and “ginger root.” However, the subject reported choosing “carrot” and that the procedure misidentified responses 4 and 18. In spite of the failure to deduce “carrot,” 16 out of 18 correctly transmitted yes/no responses represents above-chance communication through the MEG interface (p < .01). Figure 2 shows Subject 3’s averaged attended and unattended response by brain region from Experiment 1 in a format similar to the real-time outputs used in Experiment 2 as well as averaged responses from a representative question in Experiment 2. W AVEFORMS GROUPED BY BRAIN REGION

DIFFERENCE WAVES, ATTENDED MINUS IGNORED 100ms pre-tone to 0ms 50 to 150ms post-tone onset

Fig 1. Three subjects show differential neural activity while listening to vs. ignoring 200Hz tones in the right ear, both before the tone (left) and after the onset of the tone (right). For each line, deviation from straight indicates a difference in magnetic flux between ignore and attend conditions. The most consistent difference appears to be right posterior activity (in circles).

Fig 2. Subject 3’s averaged response to tones by brain region from 100ms pre-tone to 500ms post-tone from Experiment 1 (top) and one question from Experiment 2 (bottom). Attending during Condition 1 means “yes,” which, based on the amount of time-locked activity around 100ms (grey boxes) is what the subject appears to have done.


Table 1. Yes/No questions from and responses based on real-time analysis of MEG data. Errors are marked with an asterisk. ______________________________________________________

1. Is it green? 2. Is it classified as a Vegetable? 3. Can it bend without breaking? 4. Is it used to make salads? 5. Can it discharge waste? 6. Do you hold it when you use it? 7. Is it smaller than a loaf of bread? 8. Does it have lots of seeds? 9. Does it taste sweet? 10. Is it hard? 11. Does it have leaves? 12. Can it be used more than once? 13. Can you order it in a restaurant? 14. Does it have a hard outer shell? 15. Is it smooth? 16. Is it usually sliced or carved? 17. Does it taste good with butter? 18. Can it be used to season foods?

(no) (yes) (no) (no) * (no) (yes) (yes) (no) (no) (yes) (no) (no) (yes) (no) (yes) (yes) (no) (yes) *


Experiment 3 Following 1000 attended and 1000 unattended simultaneous presentations of the words “yes” and “no” with 20% randomly distributed loudness deviants (-5dB) within each stream, neural differences to attended vs. unattended deviant “yeses” appeared bilaterally in frontal and temporal regions (see Figure 3). RESPONSE TO DEVIANT YESSES While attending to yes While attending to no

DISCUSSION Numerous attentional tasks are known to elicit functional brain differences; many of them could be adapted for locked-in communication. Here, we find neural difference within-individuals for a small number of subjects who complete a basic task from the attention literature – attend to an auditory stimulus. Three subjects showed similar neural differences for selectively paying attention to a tone (Exp. 1); one subject was able to convey, if somewhat imperfectly, a number of choices of the answers “yes” and “no” (Exp. 2); and one subject showed distinct neural differences while using a user-friendly paradigm designed to elicit measurable neural differences in children based on their direction of attention (Exp. 3). Completely ignoring a tone which is the only sound (Exp. 1 and 2) might be impossible. For “ignore” strategies, subjects reported listing the names of states and mentally reciting memorized texts. The key here is that subjects during the “ignore” conditions are, at least, not listening as closely as during the “attend” conditions. Exp. 3 shows an alternate method that does not require subjects to actively ignore a salient stimulus but, rather, to choose which of two stimuli to pay attention to. CONCLUSION Our new selective attention procedures, especially the yes/no procedure described in Exp. 3, are, to our knowledge, the first procedures designed to detect a choice made by a locked-in child, and here we find our new paradigms to be effective with neurotypical adults. Following Owen et al’s method of establishing the presence of consciousness, the more complex a mental task a patient can perform, the more likely the patient is to be conscious. However, it is not clear to what degree unconscious processes might account for complex brain activity. Instead, here we find simple tasks that a patient could use to communicate: detecting that a patient has made a choice, even if the choice is between two simple mental tasks, would efficiently and definitively establish the presence of consciousness. ACKNOWLEDGEMENT Thank you to the team of MEG technicians at CHOP. REFERENCES

Fig 3. Subject 1 shows differential neural activity for deviant exemplars of the word “yes” depending on whether he is listening to the word “yes” in his left ear or listening to the word “no” in his right ear. Sensor groups depicted are the same as in Fig.2 (central, frontal, etc.)

[1] Owen, A.M., Coleman, M.R., Davis, M.H., Boly, M., Laureys, S. & Pickard, J.D. (2006). Detecting awareness in the vegetative state. Science 313:1402. [2] Egeth, M. (2008). A “Turing Test” and BCI for locked-in children and adults. Medical Hypotheses 70(5), 1067. [3] Gomes, H., Duff, M., Barnhardt, J., Barrett, S. & Ritter, W. (2007). Development of auditory selective attention: event-related potential measures of channel selection and target detection. Psychophysiology 44(5): 711-727. .

Detecting Consciousness with MEG

simple tasks that a patient can use as a code to communicate. “yes.” Many extant ... user-friendly methods of communication that do not require practice, that ...

417KB Sizes 1 Downloads 732 Views

Recommend Documents

like to make references to other texts or critics as this will add some depth to your analysis. Make sure ... Discuss the theme of Surfacing. Which character from ...

Detecting Communities with Common Interests on Twitter
Jun 28, 2012 - Twitter, Social Networks, Community Detection, Graph Mining. 1. INTRODUCTION ... category, we selected the six most popular celebrities based on their number of ... 10. 12. 14. 16. 18. Control Group. Film & TVMusic Hosting News Bloggin

assignment for the elective English course entitled Literarycriticism and Theory (MEG - 05) ... Aims: This TMA is concerned mainly with assessing your application and understanding of the course ... question and how you propose to develop it. ... Wha

Mar 31, 2015 - to other texts or critics as this will add some depth to your analysis. ... Discuss the title of The Color Purple in relation to its theme. 20. 3.

Decoding invariant visual information with MEG sensor ...
Data were filtered using Brainstorm Software [17] with a signal space projection and band-pass filtered from 0.1-100 Hz. For more details on the experimental paradigm and data acquisition, please refer to [10]. Fig. 1. Top row: six grayscale images u

Detecting highly overlapping communities with Model ...
1Our C++ implementation of MOSES is available at ..... a) Edge expansion: In the initial phase of the algorithm, .... software. For the specification of overlapping NMI, see the appendix of .... development of the model.

Design Considerations for Detecting Bicycles with ...
Inductive loop detectors are widely used for vehicle detection. Histori- cally, these ... engineering. They have ... Engineering,. Purdue University, West Lafayette, IN 47907. 1 .... that a depth of 5 cm provides the closest fit to the measured data.

Design Considerations for Detecting Bicycles with ...
well studied, and there are design guidelines concerning how it should be constructed .... loops spaced 4.5 m on center, the bicycle interacts with only one loop at a time. ... from the model are compared with measured loop detector data. The.

Explaining Consciousness
tive in the way normal for a waking state. I call this phenomenon creature ..... III. Sensory Consciousness. On this argument, sensory consciousness is simply a ...

Explaining Consciousness - David Rosenthal
Cartesian theory of mind, on which a mental ..... gest that the mind cannot accommodate very ...... Although we recognize on reflection that mediation does in fact ...

Detecting Like-minded Communities with Common ...
ABSTRACT. The popularity and prevalence of online social networks (OSN) have made them efficient platforms for advertising and mar- keting campaigns. One important problem in target adver- tising and viral marketing on OSNs is the efficient identifi-

A Spike-Detecting AQM to deal with Elephants
Mar 20, 2012 - As mice flows do not have much data, they almost always complete in ... Therefore, in this work, we also analyze performance of the RED. AQM. ... Priority-based scheduling gives priority to packets of one type over pack-.

Software review Detecting horizontal gene transfer with T-REX and ...
Software review ... and a flood of biological data is produced by means of high-throughout sequencing techniques, ... supervised analysis detects the possible.

Detecting Primary User Emulation Attack with Two ...
use the network service when the spectrum is not used by the primary users. ... detection has a seriously disadvantage, that is the high false alarm probability ...

Detecting Rumors from Microblogs with Recurrent ...
ploit deep data representations for efficient rumor detection. We posit that given ... units in an RNN form a direct cycle and create an internal state of the network ...

Explaining Consciousness
What is puzzling about consciousness must therefore be a .... For present purposes, I'll speak interchange- ..... hypothesis actually allows us to explain what is.