Saturday, June 18, 2011

Consciousness vibrations

New method assesses consciousness in people in vegetative and similar states.

The world has recently become aware that some paralyzed or comatose people may be more conscious than previously assumed. The medical profession had assumed that there is always some modality of behavior available to a conscious person- perhaps a finger point, and eyelid raise, or an eye-roll. But fMRI studies show that consciousness can exist in a fully "locked-in" state. Doctors have also assumed, with perhaps more cause, that some sensory modality is always preserved. If not vision, then at least hearing or touch. Whether exceptions exist here is unknown, and may be impossible to tell without much more knowledge and invasive methods of stimulation.

It is a deeply frightening prospect- to wake up from a coma, but be completely paralyzed, unable to tell the world you are there, and thence sentenced to a life of vegetative warehousing or worse. A recent fMRI study demonstrated that when one subject was asked to imagine a tennis game, the scanner could pick up distinct patterns of the characteristic activity, yet nothing else transpired.. no hand motions, no head nods, no meaningful eye blinks.. nothing. This subject's hearing was fine, evidently.

The current paper extends this work by using a different and cheaper method of diagnosis, (EEG scalp surface electrical wave monitoring), and develops out of it a deeper theory of what is going on during consciousness. MRI scans are expensive enough in usual practice (about $4000), and functional MRIs take more still time and expense. EEGs, in contrast, are easier to administer and are normally performed anyhow in cases of doubtful brain function, though they only "see" the brain's surface, and at quite low resolution.

In other ways the data from EEGs are quite rich, however, comprising the superficial brain waves that can come in many frequencies and locations. Could it provide a map of the brain's communications that reflect the long-range connections thought to be characteristic of consciousness? These authors devise a dense matrix of sixty electrodes for the patient's head, with data processing methods that provide a geographic and temporal map of electrical activity. They use two classes of patients- vegetative state (VS) and minimally conscious state (MCS) along with conscious controls, to ask whether the three groups can be reliably differentiated.

The patients had been pre-evaluated by an extensive panel of more conventional criteria, to separate some cognition without communication (MCS) from the non-cognitive VS state. For example, "... if visual pursuit of a mirror is present at least two times in the same direction, the patient is then considered to be MCS". The researchers didn't reclassify anyone based on their EEG analyses. They also wheel in a form of analysis called "Dynamic Causal Modelling" (DCM), which "... allows for for inferences about the neuronal architectures that generate hemodynamic [fMRI] or electromagnetic [EEG] signals." This is the key to their work- a model devised in their own prior work (drenched in Bayes-style statistical methods) that uses the timing and amplitude relations between EEG signals to develop simple models of communications links between major areas of the brain.

These areas are:
The inferior frontal gyrus (IFG), which is involved in decision making and risk assessment. This is in the lower part of the frontal cortex.


The superior temporal gyrus (STG), which is involved in auditory and speech processing. This is in the upper part of the temporal (side) lobe.


The primary auditory cortex (A1), which it says is at the first stage of auditory processing, is very close to STG in the temporal cortex.



The model is that not only do signals flow upward from A1 to STG and thence to IFG, but signals return back from IFG to STG while the patient is conscious, perhaps as the mechanism of attention.

The test the researchers did was very simple, offering the patients auditory tones where strings of monotonous tones were sporadically shifted to new pitches. The signals they were looking for were of the brains "noticing" the shifts.


The top diagram shows tones used, including noticeable deviations. The bottom diagram shows a sample trace (ERP) from one EEG electrode in a control subject, where the shifted sound is reflected/recognized in a later blip (red) which is proposed to be the conscious recognition of ... something different.



Here, all sixty electrodes are mapped out spatially on a notional scalp, and temporally relative to tone shifts. One can see that the controls have far more active resonating activity after such a shift than either the MCS or VS patients. Yet the MCS patients show a discernable peak at 172 milliseconds that the VS patients don't. Below in black are shown the same data thresholded to a significance level of P<0.001 for each population.

The theory is that auditory processing is intact for all the subjects, reflected in early activation (~60ms) of difference detection in the primary auditory areas. Yet consciousness, which is known to lag sensory processing by larger intervals (up to 0.5 second), happens later on, and would necessarily be picked up later on in the EEGs. The question is where exactly the conscious signal lies, what does it mean, and can its detection be automated in a practical EEG test that can be widely applied in hospital settings?

The researchers apply their home-grown DCM model of intra-brain communication, and conclude that by far the best model (of all the theoretically possible connection schemes among the A1, STG and IFG regions) that matches the data is one where the missing ingredient in VS patients is not conduction of auditory information all the way up the chain from A1 to STG to IFG, but where recurrent conduction back from IFG to STG was absent, shown in red below.



They note that by their model, recurrent connections are still active in all cases between the lower STG and A1 levels, but while these may have important roles in auditory processing, they are all unconscious. "These findings stress the importance of recurrent processing in higher-order associative areas in the generation of conscious perception and do not support the view that recurrent processing in sensory cortex can be equated with consciousness."

VS patients lack the long-latency EEG signals that indicate reaching-back of the cortical areas back to the sensory areas. The authors characterize the role of such connections as Bayesian predictive modelling that is executed by the cortical areas and constitutes "inference on the causes of external stimuli". But the inference isn't enough on its cortical pedestal. It needs to continually check back on its inputs to validate its modelling, and perhaps thus create the sensation of "realness" and the sensation of time passing as differentiable events.

Another group puts the thought similarly:
"These findings challenge the pivotal role of the prefrontal cortex in consciousness. Instead, it appears that specific brain areas (or cognitive modules) may support specific cognitive functions but that consciousness is independent of this. Conscious sensations arise only when the brain areas involved engage in recurrent interactions enabling the long-lasting exchange of information between brain regions. Moreover, recent evidence suggests that also the state of consciousness, for example, in vegetative state patients or during sleep and anesthesia, is closely related to the scope and extent of residual recurrent interactions among brain regions."


"The Lilliputians look upon fraud as a greater crime than theft. For, they allege, care and vigilance, with a very common understanding, can protect a man’s goods from thieves, but honesty hath no fence against superior cunning. . . where fraud is permitted or connived at, or hath no law to punish it, the honest dealer is always undone, and the knave gets the advantage."