Electroencephalography

From Wikipedia, the free encyclopedia

Electroencephalography
Epileptic spike and wave discharges monitored with EEG

Electroencephalography (EEG) is an electrophysiological monitoring method to record electrical activity of the brain. It is typically noninvasive, with the electrodes placed along the scalp, although invasive electrodes are sometimes used such as in electrocorticography. EEG measures voltage fluctuations resulting from ionic current within the neurons of the brain. In clinical contexts, EEG refers to the recording of the brain's spontaneous electrical activity over a period of time, as recorded from multiple electrodes placed on the scalp. Diagnostic applications generally focus either on event-related potentials or on the spectral content of EEG. The former investigates potential fluctuations time locked to an event like stimulus onset or button press. The latter analyses the type of neural oscillations (popularly called "brain waves") that can be observed in EEG signals in the frequency domain.

EEG is most often used to diagnose epilepsy, which causes abnormalities in EEG readings.^[2] It is also used to diagnose sleep disorders, depth of anesthesia, coma, encephalopathies, and brain death. EEG used to be a first-line method of diagnosis for tumors, stroke and other focal brain disorders,^[3][4] but this use has decreased with the advent of high-resolution anatomical imaging techniques such as magnetic resonance imaging (MRI) and computed tomography (CT). Despite limited spatial resolution, EEG continues to be a valuable tool for research and diagnosis. It is one of the few mobile techniques available and offers millisecond-range temporal resolution which is not possible with CT, PET or MRI.

Derivatives of the EEG technique include evoked potentials (EP), which involves averaging the EEG activity time-locked to the presentation of a stimulus of some sort (visual, somatosensory, or auditory). Event-related potentials (ERPs) refer to averaged EEG responses that are time-locked to more complex processing of stimuli; this technique is used in cognitive science, cognitive psychology, and psychophysiological research.

History

The first human EEG recording obtained by Hans Berger in 1924. The upper tracing is EEG, and the lower is a 10 Hz timing signal.

 

Hans Berger

The history of EEG is detailed by Barbara E. Swartz in Electroencephalography and Clinical Neurophysiology.^[5] In 1875, Richard Caton (1842–1926), a physician practicing in Liverpool, presented his findings about electrical phenomena of the exposed cerebral hemispheres of rabbits and monkeys in the British Medical Journal. In 1890, Polish physiologist Adolf Beck published an investigation of spontaneous electrical activity of the brain of rabbits and dogs that included rhythmic oscillations altered by light. Beck started experiments on the electrical brain activity of animals. Beck placed electrodes directly on the surface of brain to test for sensory stimulation. His observation of fluctuating brain activity led to the conclusion of brain waves.^[6]

In 1912, Ukrainian physiologist Vladimir Vladimirovich Pravdich-Neminsky published the first animal EEG and the evoked potential of the mammalian (dog).^[7] In 1914, Napoleon Cybulski and Jelenska-Macieszyna photographed EEG recordings of experimentally induced seizures.

German physiologist and psychiatrist Hans Berger (1873–1941) recorded the first human EEG in 1924.^[8] Expanding on work previously conducted on animals by Richard Caton and others, Berger also invented the electroencephalogram (giving the device its name), an invention described "as one of the most surprising, remarkable, and momentous developments in the history of clinical neurology".^[9] His discoveries were first confirmed by British scientists Edgar Douglas Adrian and B. H. C. Matthews in 1934 and developed by them.

In 1934, Fisher and Lowenback first demonstrated epileptiform spikes. In 1935, Gibbs, Davis and Lennox described interictal spike waves and the three cycles/s pattern of clinical absence seizures, which began the field of clinical electroencephalography. Subsequently, in 1936 Gibbs and Jasper reported the interictal spike as the focal signature of epilepsy. The same year, the first EEG laboratory opened at Massachusetts General Hospital.

Franklin Offner (1911–1999), professor of biophysics at Northwestern University developed a prototype of the EEG that incorporated a piezoelectric inkwriter called a Crystograph (the whole device was typically known as the Offner Dynograph).

In 1947, The American EEG Society was founded and the first International EEG congress was held. In 1953 Aserinsky and Kleitman described REM sleep.

In the 1950s, William Grey Walter developed an adjunct to EEG called EEG topography, which allowed for the mapping of electrical activity across the surface of the brain. This enjoyed a brief period of popularity in the 1980s and seemed especially promising for psychiatry. It was never accepted by neurologists and remains primarily a research tool.

In 1988, report was given on EEG control of a physical object, a robot.^[10][11]

Medical use

An EEG recording setup

A routine clinical EEG recording typically lasts 20–30 minutes (plus preparation time) and usually involves recording from scalp electrodes. Routine EEG is typically used in clinical circumstances to distinguish epileptic seizures from other types of spells, such as psychogenic non-epileptic seizures, syncope (fainting), sub-cortical movement disorders and migraine variants, to differentiate "organic" encephalopathy or delirium from primary psychiatric syndromes such as catatonia, to serve as an adjunct test of brain death, to prognosticate, in certain instances, in patients with coma, and to determine whether to wean anti-epileptic medications.

At times, a routine EEG is not sufficient, particularly when it is necessary to record a patient while he/she is having a seizure. In this case, the patient may be admitted to the hospital for days or even weeks, while EEG is constantly being recorded (along with time-synchronized video and audio recording). A recording of an actual seizure (i.e., an ictal recording, rather than an inter-ictal recording of a possibly epileptic patient at some period between seizures) can give significantly better information about whether or not a spell is an epileptic seizure and the focus in the brain from which the seizure activity emanates.

Epilepsy monitoring is typically done to distinguish epileptic seizures from other types of spells, such as psychogenic non-epileptic seizures, syncope (fainting), sub-cortical movement disorders and migraine variants, to characterize seizures for the purposes of treatment, and to localize the region of brain from which a seizure originates for work-up of possible seizure surgery.

Additionally, EEG may be used to monitor the depth of anesthesia, as an indirect indicator of cerebral perfusion in carotid endarterectomy, or to monitor amobarbital effect during the Wada test.

EEG can also be used in intensive care units for brain function monitoring to monitor for non-convulsive seizures/non-convulsive status epilepticus, to monitor the effect of sedative/anesthesia in patients in medically induced coma (for treatment of refractory seizures or increased intracranial pressure), and to monitor for secondary brain damage in conditions such as subarachnoid hemorrhage (currently a research method).

If a patient with epilepsy is being considered for resective surgery, it is often necessary to localize the focus (source) of the epileptic brain activity with a resolution greater than what is provided by scalp EEG. This is because the cerebrospinal fluid, skull and scalp smear the electrical potentials recorded by scalp EEG. In these cases, neurosurgeons typically implant strips and grids of electrodes (or penetrating depth electrodes) under the dura mater, through either a craniotomy or a burr hole. The recording of these signals is referred to as electrocorticography (ECoG), subdural EEG (sdEEG) or intracranial EEG (icEEG)--all terms for the same thing. The signal recorded from ECoG is on a different scale of activity than the brain activity recorded from scalp EEG. Low voltage, high frequency components that cannot be seen easily (or at all) in scalp EEG can be seen clearly in ECoG. Further, smaller electrodes (which cover a smaller parcel of brain surface) allow even lower voltage, faster components of brain activity to be seen. Some clinical sites record from penetrating microelectrodes.^[1]

EEG is not indicated for diagnosing headache.^[12] Recurring headache is a common pain problem, and this procedure is sometimes used in a search for a diagnosis, but it has no advantage over routine clinical evaluation.^[12]

Research use

EEG, and the related study of ERPs are used extensively in neuroscience, cognitive science, cognitive psychology, neurolinguistics and psychophysiological research. Many EEG techniques used in research are not standardised sufficiently for clinical use. But research on mental disabilities, such as auditory processing disorder (APD), ADD, or ADHD, is becoming more widely known and EEGs are used as research and treatment.

Advantages

Several other methods to study brain function exist, including functional magnetic resonance imaging (fMRI), positron emission tomography, magnetoencephalography (MEG), nuclear magnetic resonance spectroscopy, electrocorticography, single-photon emission computed tomography, near-infrared spectroscopy (NIRS), and event-related optical signal (EROS). Despite the relatively poor spatial sensitivity of EEG, it possesses multiple advantages over some of these techniques:

• Hardware costs are significantly lower than those of most other techniques ^[13]
• EEG prevents limited availability of technologists to provide immediate care in high traffic hospitals.^[14]
• EEG sensors can be used in more places than fMRI, SPECT, PET, MRS, or MEG, as these techniques require bulky and immobile equipment. For example, MEG requires equipment consisting of liquid helium-cooled detectors that can be used only in magnetically shielded rooms, altogether costing upwards of several million dollars;^[15] and fMRI requires the use of a 1-ton magnet in, again, a shielded room.
• EEG has very high temporal resolution, on the order of milliseconds rather than seconds. EEG is commonly recorded at sampling rates between 250 and 2000 Hz in clinical and research settings, but modern EEG data collection systems are capable of recording at sampling rates above 20,000 Hz if desired. MEG and EROS are the only other noninvasive cognitive neuroscience techniques that acquire data at this level of temporal resolution.^[15]
• EEG is relatively tolerant of subject movement, unlike most other neuroimaging techniques. There even exist methods for minimizing, and even eliminating movement artifacts in EEG data ^[16]
• EEG is silent, which allows for better study of the responses to auditory stimuli.
• EEG does not aggravate claustrophobia, unlike fMRI, PET, MRS, SPECT, and sometimes MEG^[17]
• EEG does not involve exposure to high-intensity (>1 tesla) magnetic fields, as in some of the other techniques, especially MRI and MRS. These can cause a variety of undesirable issues with the data, and also prohibit use of these techniques with participants that have metal implants in their body, such as metal-containing pacemakers^[18]
• EEG does not involve exposure to radioligands, unlike positron emission tomography.^[19]
• ERP studies can be conducted with relatively simple paradigms, compared with IE block-design fMRI studies
• Extremely uninvasive, unlike Electrocorticography, which actually requires electrodes to be placed on the surface of the brain.

EEG also has some characteristics that compare favorably with behavioral testing:

• EEG can detect covert processing (i.e., processing that does not require a response)^[20]
• EEG can be used in subjects who are incapable of making a motor response^[21]
• Some ERP components can be detected even when the subject is not attending to the stimuli
• Unlike other means of studying reaction time, ERPs can elucidate stages of processing (rather than just the final end result)^[22]
• EEG is a powerful tool for tracking brain changes during different phases of life. EEG sleep analysis can indicate significant aspects of the timing of brain development, including evaluating adolescent brain maturation.^[23]
• In EEG there is a better understanding of what signal is measured as compared to other research techniques, i.e. the BOLD response in MRI.

Disadvantages

• Low spatial resolution on the scalp. fMRI, for example, can directly display areas of the brain that are active, while EEG requires intense interpretation just to hypothesize what areas are activated by a particular response.^[24]
• EEG poorly measures neural activity that occurs below the upper layers of the brain (the cortex).
• Unlike PET and MRS, cannot identify specific locations in the brain at which various neurotransmitters, drugs, etc. can be found.^[19]
• Often takes a long time to connect a subject to EEG, as it requires precise placement of dozens of electrodes around the head and the use of various gels, saline solutions, and/or pastes to keep them in place (although a cap can be used). While the length of time differs dependent on the specific EEG device used, as a general rule it takes considerably less time to prepare a subject for MEG, fMRI, MRS, and SPECT.
• Signal-to-noise ratio is poor, so sophisticated data analysis and relatively large numbers of subjects are needed to extract useful information from EEG^[25]

With other neuroimaging techniques

Simultaneous EEG recordings and fMRI scans have been obtained successfully,^[26][27] though successful simultaneous recording requires that several technical difficulties be overcome, such as the presence of ballistocardiographic artifact, MRI pulse artifact and the induction of electrical currents in EEG wires that move within the strong magnetic fields of the MRI. While challenging, these have been successfully overcome in a number of studies.^[28]

MRI's produce detailed images created by generating strong magnetic fields that may induce potentially harmful displacement force and torque. These fields produce potentially harmful radio frequency heating and create image artifacts rendering images useless. Due to these potential risks, only certain medical devices can be used in an MR environment.

Similarly, simultaneous recordings with MEG and EEG have also been conducted, which has several advantages over using either technique alone:

• EEG requires accurate information about certain aspects of the skull that can only be estimated, such as skull radius, and conductivities of various skull locations. MEG does not have this issue, and a simultaneous analysis allows this to be corrected for.
• MEG and EEG both detect activity below the surface of the cortex very poorly, and like EEG, the level of error increases with the depth below the surface of the cortex one attempts to examine. However, the errors are very different between the techniques, and combining them thus allows for correction of some of this noise.
• MEG has access to virtually no sources of brain activity below a few centimetres under the cortex. EEG, on the other hand, can receive signals from greater depth, albeit with a high degree of noise. Combining the two makes it easier to determine what in the EEG signal comes from the surface (since MEG is very accurate in examining signals from the surface of the brain), and what comes from deeper in the brain, thus allowing for analysis of deeper brain signals than either EEG or MEG on its own.^[29]

Recently, a combined EEG/MEG (EMEG) approach has been investigated for the purpose of source reconstruction in epilepsy diagnosis.^[30]

EEG has also been combined with positron emission tomography. This provides the advantage of allowing researchers to see what EEG signals are associated with different drug actions in the brain.^[31]

Mechanisms

The brain's electrical charge is maintained by billions of neurons.^[32] Neurons are electrically charged (or "polarized") by membrane transport proteins that pump ions across their membranes. Neurons are constantly exchanging ions with the extracellular milieu, for example to maintain resting potential and to propagate action potentials. Ions of similar charge repel each other, and when many ions are pushed out of many neurons at the same time, they can push their neighbours, who push their neighbours, and so on, in a wave. This process is known as volume conduction. When the wave of ions reaches the electrodes on the scalp, they can push or pull electrons on the metal in the electrodes. Since metal conducts the push and pull of electrons easily, the difference in push or pull voltages between any two electrodes can be measured by a voltmeter. Recording these voltages over time gives us the EEG.^[33]

The electric potential generated by an individual neuron is far too small to be picked up by EEG or MEG.^[34] EEG activity therefore always reflects the summation of the synchronous activity of thousands or millions of neurons that have similar spatial orientation. If the cells do not have similar spatial orientation, their ions do not line up and create waves to be detected. Pyramidal neurons of the cortex are thought to produce the most EEG signal because they are well-aligned and fire together. Because voltage field gradients fall off with the square of distance, activity from deep sources is more difficult to detect than currents near the skull.^[35]

Scalp EEG activity shows oscillations at a variety of frequencies. Several of these oscillations have characteristic frequency ranges, spatial distributions and are associated with different states of brain functioning (e.g., waking and the various sleep stages). These oscillations represent synchronized activity over a network of neurons. The neuronal networks underlying some of these oscillations are understood (e.g., the thalamocortical resonance underlying sleep spindles), while many others are not (e.g., the system that generates the posterior basic rhythm). Research that measures both EEG and neuron spiking finds the relationship between the two is complex, with a combination of EEG power in the gamma band and phase in the delta band relating most strongly to neuron spike activity.^[36]

Method

Computer electroencephalograph Neurovisor-BMM 40

In conventional scalp EEG, the recording is obtained by placing electrodes on the scalp with a conductive gel or paste, usually after preparing the scalp area by light abrasion to reduce impedance due to dead skin cells. Many systems typically use electrodes, each of which is attached to an individual wire. Some systems use caps or nets into which electrodes are embedded; this is particularly common when high-density arrays of electrodes are needed.

Electrode locations and names are specified by the International 10–20 system^[37] for most clinical and research applications (except when high-density arrays are used). This system ensures that the naming of electrodes is consistent across laboratories. In most clinical applications, 19 recording electrodes (plus ground and system reference) are used.^[38] A smaller number of electrodes are typically used when recording EEG from neonates. Additional electrodes can be added to the standard set-up when a clinical or research application demands increased spatial resolution for a particular area of the brain. High-density arrays (typically via cap or net) can contain up to 256 electrodes more-or-less evenly spaced around the scalp.

Each electrode is connected to one input of a differential amplifier (one amplifier per pair of electrodes); a common system reference electrode is connected to the other input of each differential amplifier. These amplifiers amplify the voltage between the active electrode and the reference (typically 1,000–100,000 times, or 60–100 dB of voltage gain). In analog EEG, the signal is then filtered (next paragraph), and the EEG signal is output as the deflection of pens as paper passes underneath. Most EEG systems these days, however, are digital, and the amplified signal is digitized via an analog-to-digital converter, after being passed through an anti-aliasing filter. Analog-to-digital sampling typically occurs at 256–512 Hz in clinical scalp EEG; sampling rates of up to 20 kHz are used in some research applications.

During the recording, a series of activation procedures may be used. These procedures may induce normal or abnormal EEG activity that might not otherwise be seen. These procedures include hyperventilation, photic stimulation (with a strobe light), eye closure, mental activity, sleep and sleep deprivation. During (inpatient) epilepsy monitoring, a patient's typical seizure medications may be withdrawn.

The digital EEG signal is stored electronically and can be filtered for display. Typical settings for the high-pass filter and a low-pass filter are 0.5–1 Hz and 35–70 Hz respectively. The high-pass filter typically filters out slow artifact, such as electrogalvanic signals and movement artifact, whereas the low-pass filter filters out high-frequency artifacts, such as electromyographic signals. An additional notch filter is typically used to remove artifact caused by electrical power lines (60 Hz in the United States and 50 Hz in many other countries).^[1]

The EEG signals can be captured with opensource hardware such as OpenBCI and the signal can be processed by freely available EEG software such as EEGLAB or the Neurophysiological Biomarker Toolbox.

As part of an evaluation for epilepsy surgery, it may be necessary to insert electrodes near the surface of the brain, under the surface of the dura mater. This is accomplished via burr hole or craniotomy. This is referred to variously as "electrocorticography (ECoG)", "intracranial EEG (I-EEG)" or "subdural EEG (SD-EEG)". Depth electrodes may also be placed into brain structures, such as the amygdala or hippocampus, structures, which are common epileptic foci and may not be "seen" clearly by scalp EEG. The electrocorticographic signal is processed in the same manner as digital scalp EEG (above), with a couple of caveats. ECoG is typically recorded at higher sampling rates than scalp EEG because of the requirements of Nyquist theorem—the subdural signal is composed of a higher predominance of higher frequency components. Also, many of the artifacts that affect scalp EEG do not impact ECoG, and therefore display filtering is often not needed.

A typical adult human EEG signal is about 10 µV to 100 µV in amplitude when measured from the scalp^[39] and is about 10–20 mV when measured from subdural electrodes.

Since an EEG voltage signal represents a difference between the voltages at two electrodes, the display of the EEG for the reading encephalographer may be set up in one of several ways. The representation of the EEG channels is referred to as a montage.

Sequential montage

Each channel (i.e., waveform) represents the difference between two adjacent electrodes. The entire montage consists of a series of these channels. For example, the channel "Fp1-F3" represents the difference in voltage between the Fp1 electrode and the F3 electrode. The next channel in the montage, "F3-C3", represents the voltage difference between F3 and C3, and so on through the entire array of electrodes.

Referential montage

Each channel represents the difference between a certain electrode and a designated reference electrode. There is no standard position for this reference; it is, however, at a different position than the "recording" electrodes. Midline positions are often used because they do not amplify the signal in one hemisphere vs. the other. Another popular reference is "linked ears", which is a physical or mathematical average of electrodes attached to both earlobes or mastoids.

Average reference montage

The outputs of all of the amplifiers are summed and averaged, and this averaged signal is used as the common reference for each channel.

Laplacian montage

Each channel represents the difference between an electrode and a weighted average of the surrounding electrodes.^[40]

When analog (paper) EEGs are used, the technologist switches between montages during the recording in order to highlight or better characterize certain features of the EEG. With digital EEG, all signals are typically digitized and stored in a particular (usually referential) montage; since any montage can be constructed mathematically from any other, the EEG can be viewed by the electroencephalographer in any display montage that is desired.

The EEG is read by a clinical neurophysiologist or neurologist (depending on local custom and law regarding medical specialities), optimally one who has specific training in the interpretation of EEGs for clinical purposes. This is done by visual inspection of the waveforms, called graphoelements. The use of computer signal processing of the EEG—so-called quantitative electroencephalography—is somewhat controversial when used for clinical purposes (although there are many research uses).

Limitations

EEG has several limitations. Most important is its poor spatial resolution.^[41] EEG is most sensitive to a particular set of post-synaptic potentials: those generated in superficial layers of the cortex, on the crests of gyri directly abutting the skull and radial to the skull. Dendrites, which are deeper in the cortex, inside sulci, in midline or deep structures (such as the cingulate gyrus or hippocampus), or producing currents that are tangential to the skull, have far less contribution to the EEG signal.
EEG recordings do not directly capture axonal action potentials. An action potential can be accurately represented as a current quadrupole, meaning that the resulting field decreases more rapidly than the ones produced by the current dipole of post-synaptic potentials.^[42] In addition, since EEGs represent averages of thousands of neurons, a large population of cells in synchronous activity is necessary to cause a significant deflection on the recordings. Action potentials are very fast and, as a consequence, the chances of field summation are slim. However, neural backpropagation, as a typically longer dendritic current dipole, can be picked up by EEG electrodes and is a reliable indication of the occurrence of neural output.

Not only do EEGs capture dendritic currents almost exclusively as opposed to axonal currents, they also show a preference for activity on populations of parallel dendrites and transmitting current in the same direction at the same time. Pyramidal neurons of cortical layers II/III and V extend apical dendrites to layer I. Currents moving up or down these processes underlie most of the signals produced by electroencephalography.^[43]

Therefore, EEG provides information with a large bias to select neuron types, and generally should not be used to make claims about global brain activity. The meninges, cerebrospinal fluid and skull "smear" the EEG signal, obscuring its intracranial source.

It is mathematically impossible to reconstruct a unique intracranial current source for a given EEG signal,^[1] as some currents produce potentials that cancel each other out. This is referred to as the inverse problem. However, much work has been done to produce remarkably good estimates of, at least, a localized electric dipole that represents the recorded currents.^{[citation needed]}

EEG vs fMRI, fNIRS and PET

EEG has several strong points as a tool for exploring brain activity. EEGs can detect changes over milliseconds, which is excellent considering an action potential takes approximately 0.5–130 milliseconds to propagate across a single neuron, depending on the type of neuron.^[44] Other methods of looking at brain activity, such as PET and fMRI have time resolution between seconds and minutes. EEG measures the brain's electrical activity directly, while other methods record changes in blood flow (e.g., SPECT, fMRI) or metabolic activity (e.g., PET, NIRS), which are indirect markers of brain electrical activity. EEG can be used simultaneously with fMRI so that high-temporal-resolution data can be recorded at the same time as high-spatial-resolution data, however, since the data derived from each occurs over a different time course, the data sets do not necessarily represent exactly the same brain activity. There are technical difficulties associated with combining these two modalities, including the need to remove the MRI gradient artifact present during MRI acquisition and the ballistocardiographic artifact (resulting from the pulsatile motion of blood and tissue) from the EEG. Furthermore, currents can be induced in moving EEG electrode wires due to the magnetic field of the MRI.

EEG can be used simultaneously with NIRS without major technical difficulties. There is no influence of these modalities on each other and a combined measurement can give useful information about electrical activity as well as local hemodynamics.

EEG vs MEG

EEG reflects correlated synaptic activity caused by post-synaptic potentials of cortical neurons. The ionic currents involved in the generation of fast action potentials may not contribute greatly to the averaged field potentials representing the EEG.^[34][45] More specifically, the scalp electrical potentials that produce EEG are generally thought to be caused by the extracellular ionic currents caused by dendritic electrical activity, whereas the fields producing magnetoencephalographic signals^[15] are associated with intracellular ionic currents.^[46]

EEG can be recorded at the same time as MEG so that data from these complementary high-time-resolution techniques can be combined.

Studies on numerical modeling of EEG and MEG have also been done.^[47]

Normal activity

• 

  One second of EEG signal
  
  
• 

  The sample of human EEG with in resting state. Left: EEG traces (horizontal – time in seconds; vertical – amplitudes, scale 100 μV). Right: power spectra of shown signals (vertical lines – 10 and 20 Hz, scale is linear). 80–90% of people have prominent sinusoidal-like waves with frequencies in 8–12 Hz range – alpha rhythm. Others (like this) lack this type of activity.
  
  
• 

  The sample of human EEG with prominent resting state activity – alpha-rhythm. Left: EEG traces (horizontal – time in seconds; vertical – amplitudes, scale 100 μV). Right: power spectra of shown signals (vertical lines – 10 and 20 Hz, scale is linear). Alpha-rhythm consists of sinusoidal-like waves with frequencies in 8–12 Hz range (11 Hz in this case) more prominent in posterior sites. Alpha range is red at power spectrum graph.
  
  
• 

  The samples of main types of artifacts in human EEG. 1: Electrooculographic artifact caused by the excitation of eyeball's muscles (related to blinking, for example). Big-amplitude, slow, positive wave prominent in frontal electrodes. 2: Electrode's artifact caused by bad contact (and thus bigger impedance) between P3 electrode and skin. 3: Swallowing artifact. 4: Common reference electrode's artifact caused by bad contact between reference electrode and skin. Huge wave similar in all channels.
  

The EEG is typically described in terms of (1) rhythmic activity and (2) transients. The rhythmic activity is divided into bands by frequency. To some degree, these frequency bands are a matter of nomenclature (i.e., any rhythmic activity between 8–12 Hz can be described as "alpha"), but these designations arose because rhythmic activity within a certain frequency range was noted to have a certain distribution over the scalp or a certain biological significance. Frequency bands are usually extracted using spectral methods (for instance Welch) as implemented for instance in freely available EEG software such as EEGLAB or the Neurophysiological Biomarker Toolbox. Computational processing of the EEG is often named quantitative electroencephalography (qEEG).

Most of the cerebral signal observed in the scalp EEG falls in the range of 1–20 Hz (activity below or above this range is likely to be artifactual, under standard clinical recording techniques). Waveforms are subdivided into bandwidths known as alpha, beta, theta, and delta to signify the majority of the EEG used in clinical practice.^[48]
 

Comparison of EEG bands

Band	Frequency (Hz)	Location	Normally	Pathologically
Delta	< 4	frontally in adults, posteriorly in children; high-amplitude waves	adult slow-wave sleep in babies Has been found during some continuous-attention tasks[49]	subcortical lesions diffuse lesions metabolic encephalopathy hydrocephalus deep midline lesions
Theta	4–7	Found in locations not related to task at hand	higher in young children drowsiness in adults and teens idling Associated with inhibition of elicited responses (has been found to spike in situations where a person is actively trying to repress a response or action).[49]	focal subcortical lesions metabolic encephalopathy deep midline disorders some instances of hydrocephalus
Alpha	8–15	posterior regions of head, both sides, higher in amplitude on dominant side. Central sites (c3-c4) at rest	relaxed/reflecting closing the eyes Also associated with inhibition control, seemingly with the purpose of timing inhibitory activity in different locations across the brain.	coma
Beta	16–31	both sides, symmetrical distribution, most evident frontally; low-amplitude waves	range span: active calm → intense → stressed → mild obsessive active thinking, focus, high alert, anxious	benzodiazepines Dup15q syndrome [50]
Gamma	> 32	Somatosensory cortex	Displays during cross-modal sensory processing (perception that combines two different senses, such as sound and sight)[51][52] Also is shown during short-term memory matching of recognized objects, sounds, or tactile sensations	A decrease in gamma-band activity may be associated with cognitive decline, especially when related to the theta band; however, this has not been proven for use as a clinical diagnostic measurement
Mu	8–12	Sensorimotor cortex	Shows rest-state motor neurons.[53]	Mu suppression could indicate that motor mirror neurons are working. Deficits in Mu suppression, and thus in mirror neurons, might play a role in autism.[54]

The practice of using only whole numbers in the definitions comes from practical considerations in the days when only whole cycles could be counted on paper records. This leads to gaps in the definitions, as seen elsewhere on this page. The theoretical definitions have always been more carefully defined to include all frequencies. Unfortunately there is no agreement in standard reference works on what these ranges should be – values for the upper end of alpha and lower end of beta include 12, 13, 14 and 15. If the threshold is taken as 14 Hz, then the slowest beta wave has about the same duration as the longest spike (70 ms), which makes this the most useful value.

EEG Frequency bands: Improved definitions ^[55]

Band	Frequency (Hz)
Delta	< 4
Theta	≥ 4 and < 8
Alpha	≥ 8 and < 14
Beta	≥ 14

Others sometimes divide the bands into sub-bands for the purposes of data analysis.

Human EEG with prominent alpha-rhythm

Wave patterns

Delta waves

• Delta is the frequency range up to 4 Hz. It tends to be the highest in amplitude and the slowest waves. It is seen normally in adults in slow-wave sleep. It is also seen normally in babies. It may occur focally with subcortical lesions and in general distribution with diffuse lesions, metabolic encephalopathy hydrocephalus or deep midline lesions. It is usually most prominent frontally in adults (e.g. FIRDA – frontal intermittent rhythmic delta) and posteriorly in children (e.g. OIRDA – occipital intermittent rhythmic delta).

Theta waves

• Theta is the frequency range from 4 Hz to 7 Hz. Theta is seen normally in young children. It may be seen in drowsiness or arousal in older children and adults; it can also be seen in meditation.^[56] Excess theta for age represents abnormal activity. It can be seen as a focal disturbance in focal subcortical lesions; it can be seen in generalized distribution in diffuse disorder or metabolic encephalopathy or deep midline disorders or some instances of hydrocephalus. On the contrary this range has been associated with reports of relaxed, meditative, and creative states.

Alpha waves

• Alpha is the frequency range from 7 Hz to 13 Hz.^[57] Hans Berger named the first rhythmic EEG activity he saw as the "alpha wave". This was the "posterior basic rhythm" (also called the "posterior dominant rhythm" or the "posterior alpha rhythm"), seen in the posterior regions of the head on both sides, higher in amplitude on the dominant side. It emerges with closing of the eyes and with relaxation, and attenuates with eye opening or mental exertion. The posterior basic rhythm is actually slower than 8 Hz in young children (therefore technically in the theta range).

Sensorimotor rhythm aka mu rhythm

In addition to the posterior basic rhythm, there are other normal alpha rhythms such as the mu rhythm (alpha activity in the contralateral sensory and motor cortical areas) that emerges when the hands and arms are idle; and the "third rhythm" (alpha activity in the temporal or frontal lobes).^[58][59] Alpha can be abnormal; for example, an EEG that has diffuse alpha occurring in coma and is not responsive to external stimuli is referred to as "alpha coma".

Beta waves

• Beta is the frequency range from 14 Hz to about 30 Hz. It is seen usually on both sides in symmetrical distribution and is most evident frontally. Beta activity is closely linked to motor behavior and is generally attenuated during active movements.^[60] Low-amplitude beta with multiple and varying frequencies is often associated with active, busy or anxious thinking and active concentration. Rhythmic beta with a dominant set of frequencies is associated with various pathologies, such as Dup15q syndrome, and drug effects, especially benzodiazepines. It may be absent or reduced in areas of cortical damage. It is the dominant rhythm in patients who are alert or anxious or who have their eyes open.

Gamma waves

• Gamma is the frequency range approximately 30–100 Hz. Gamma rhythms are thought to represent binding of different populations of neurons together into a network for the purpose of carrying out a certain cognitive or motor function.^[1]
• Mu range is 8–13 Hz and partly overlaps with other frequencies. It reflects the synchronous firing of motor neurons in rest state. Mu suppression is thought to reflect motor mirror neuron systems, because when an action is observed, the pattern extinguishes, possibly because of the normal neuronal system and the mirror neuron system "go out of sync" and interfere with each other.^[54]

"Ultra-slow" or "near-DC" activity is recorded using DC amplifiers in some research contexts. It is not typically recorded in a clinical context because the signal at these frequencies is susceptible to a number of artifacts.

Some features of the EEG are transient rather than rhythmic. Spikes and sharp waves may represent seizure activity or interictal activity in individuals with epilepsy or a predisposition toward epilepsy. Other transient features are normal: vertex waves and sleep spindles are seen in normal sleep.
Note that there are types of activity that are statistically uncommon, but not associated with dysfunction or disease. These are often referred to as "normal variants". The mu rhythm is an example of a normal variant.

The normal electroencephalography (EEG) varies by age. The neonatal EEG is quite different from the adult EEG. The EEG in childhood generally has slower frequency oscillations than the adult EEG.

The normal EEG also varies depending on state. The EEG is used along with other measurements (EOG, EMG) to define sleep stages in polysomnography. Stage I sleep (equivalent to drowsiness in some systems) appears on the EEG as drop-out of the posterior basic rhythm. There can be an increase in theta frequencies. Santamaria and Chiappa cataloged a number of the variety of patterns associated with drowsiness. Stage II sleep is characterized by sleep spindles – transient runs of rhythmic activity in the 12–14 Hz range (sometimes referred to as the "sigma" band) that have a frontal-central maximum. Most of the activity in Stage II is in the 3–6 Hz range. Stage III and IV sleep are defined by the presence of delta frequencies and are often referred to collectively as "slow-wave sleep". Stages I–IV comprise non-REM (or "NREM") sleep. The EEG in REM (rapid eye movement) sleep appears somewhat similar to the awake EEG.

EEG under general anesthesia depends on the type of anesthetic employed. With halogenated anesthetics, such as halothane or intravenous agents, such as propofol, a rapid (alpha or low beta), nonreactive EEG pattern is seen over most of the scalp, especially anteriorly; in some older terminology this was known as a WAR (widespread anterior rapid) pattern, contrasted with a WAIS (widespread slow) pattern associated with high doses of opiates. Anesthetic effects on EEG signals are beginning to be understood at the level of drug actions on different kinds of synapses and the circuits that allow synchronized neuronal activity (see: http://www.stanford.edu/group/maciverlab/).

Artifacts

Biological artifacts

Main types of artifacts in human EEG

Electrical signals detected along the scalp by an EEG, but that originate from non-cerebral origin are called artifacts. EEG data is almost always contaminated by such artifacts. The amplitude of artifacts can be quite large relative to the size of amplitude of the cortical signals of interest. This is one of the reasons why it takes considerable experience to correctly interpret EEGs clinically. Some of the most common types of biological artifacts include:

• Eye-induced artifacts (includes eye blinks, eye movements and extra-ocular muscle activity)
• ECG (cardiac) artifacts
• EMG (muscle activation)-induced artifacts
• Glossokinetic artifacts

The most prominent eye-induced artifacts are caused by the potential difference between the cornea and retina, which is quite large compared to cerebral potentials. When the eyes and eyelids are completely still, this corneo-retinal dipole does not affect EEG. However, blinks occur several times per minute, the eyes movements occur several times per second. Eyelid movements, occurring mostly during blinking or vertical eye movements, elicit a large potential seen mostly in the difference between the Electrooculography (EOG) channels above and below the eyes. An established explanation of this potential regards the eyelids as sliding electrodes that short-circuit the positively charged cornea to the extra-ocular skin.^[61][62] Rotation of the eyeballs, and consequently of the corneo-retinal dipole, increases the potential in electrodes towards which the eyes are rotated, and decrease the potentials in the opposing electrodes.^[63] Eye movements called saccades also generate transient electromyographic potentials, known as saccadic spike potentials (SPs).^[64] The spectrum of these SPs overlaps the gamma-band (see Gamma wave), and seriously confounds analysis of induced gamma-band responses,^[65] requiring tailored artifact correction approaches.^[64] Purposeful or reflexive eye blinking also generates electromyographic potentials, but more importantly there is reflexive movement of the eyeball during blinking that gives a characteristic artifactual appearance of the EEG (see Bell's phenomenon).

Eyelid fluttering artifacts of a characteristic type were previously called Kappa rhythm (or Kappa waves). It is usually seen in the prefrontal leads, that is, just over the eyes. Sometimes they are seen with mental activity. They are usually in the Theta (4–7 Hz) or Alpha (7–14 Hz) range. They were named because they were believed to originate from the brain. Later study revealed they were generated by rapid fluttering of the eyelids, sometimes so minute that it was difficult to see. They are in fact noise in the EEG reading, and should not technically be called a rhythm or wave. Therefore, current usage in electroencephalography refers to the phenomenon as an eyelid fluttering artifact, rather than a Kappa rhythm (or wave).^[66]

Some of these artifacts can be useful in various applications. The EOG signals, for instance, can be used to detect^[64] and track eye-movements, which are very important in polysomnography, and is also in conventional EEG for assessing possible changes in alertness, drowsiness or sleep.
ECG artifacts are quite common and can be mistaken for spike activity. Because of this, modern EEG acquisition commonly includes a one-channel ECG from the extremities. This also allows the EEG to identify cardiac arrhythmias that are an important differential diagnosis to syncope or other episodic/attack disorders.

Glossokinetic artifacts are caused by the potential difference between the base and the tip of the tongue. Minor tongue movements can contaminate the EEG, especially in parkinsonian and tremor disorders.

Environmental artifacts

In addition to artifacts generated by the body, many artifacts originate from outside the body. Movement by the patient, or even just settling of the electrodes, may cause electrode pops, spikes originating from a momentary change in the impedance of a given electrode. Poor grounding of the EEG electrodes can cause significant 50 or 60 Hz artifact, depending on the local power system's frequency. A third source of possible interference can be the presence of an IV drip; such devices can cause rhythmic, fast, low-voltage bursts, which may be confused for spikes.

Artifact correction

Recently, independent component analysis (ICA) techniques have been used to correct or remove EEG contaminants.^{[64][67][68][69][70][71]} These techniques attempt to "unmix" the EEG signals into some number of underlying components. There are many source separation algorithms, often assuming various behaviors or natures of EEG. Regardless, the principle behind any particular method usually allow "remixing" only those components that would result in "clean" EEG by nullifying (zeroing) the weight of unwanted components. Fully automated artifact rejection methods, which use ICA, have also been developed.^[72]

In the last few years, by comparing data from paralysed and unparalysed subjects, EEG contamination by muscle has been shown to be far more prevalent than had previously been realized, particularly in the gamma range above 20 Hz.^[73] However, Surface Laplacian has been shown to be effective in eliminating muscle artefact, particularly for central electrodes, which are further from the strongest contaminants.^[74] The combination of Surface Laplacian with automated techniques for removing muscle components using ICA proved particularly effective in a follow up study.^[75]

Abnormal activity

Abnormal activity can broadly be separated into epileptiform and non-epileptiform activity. It can also be separated into focal or diffuse.

Focal epileptiform discharges represent fast, synchronous potentials in a large number of neurons in a somewhat discrete area of the brain. These can occur as interictal activity, between seizures, and represent an area of cortical irritability that may be predisposed to producing epileptic seizures. Interictal discharges are not wholly reliable for determining whether a patient has epilepsy nor where his/her seizure might originate.

Generalized epileptiform discharges often have an anterior maximum, but these are seen synchronously throughout the entire brain. They are strongly suggestive of a generalized epilepsy.

Focal non-epileptiform abnormal activity may occur over areas of the brain where there is focal damage of the cortex or white matter. It often consists of an increase in slow frequency rhythms and/or a loss of normal higher frequency rhythms. It may also appear as focal or unilateral decrease in amplitude of the EEG signal.

Diffuse non-epileptiform abnormal activity may manifest as diffuse abnormally slow rhythms or bilateral slowing of normal rhythms, such as the PBR.

Intracortical Encephalogram electrodes and sub-dural electrodes can be used in tandem to discriminate and discretize artifact from epileptiform and other severe neurological events.

More advanced measures of abnormal EEG signals have also recently received attention as possible biomarkers for different disorders such as Alzheimer's disease.^[76]

Remote communication

The United States Army Research Office budgeted $4 million in 2009 to researchers at the University of California, Irvine to develop EEG processing techniques to identify correlates of imagined speech and intended direction to enable soldiers on the battlefield to communicate via computer-mediated reconstruction of team members' EEG signals, in the form of understandable signals such as words.^[77]

Economics

Inexpensive EEG devices exist for the low-cost research and consumer markets. Recently, a few companies have miniaturized medical grade EEG technology to create versions accessible to the general public. Some of these companies have built commercial EEG devices retailing for less than $100 USD.

• In 2004 OpenEEG released its ModularEEG as open source hardware. Compatible open source software includes a game for balancing a ball.
• In 2007 NeuroSky released the first affordable consumer based EEG along with the game NeuroBoy. This was also the first large scale EEG device to use dry sensor technology.^[78]
• In 2008 OCZ Technology developed device for use in video games relying primarily on electromyography.
• In 2008 the Final Fantasy developer Square Enix announced that it was partnering with NeuroSky to create a game, Judecca.^[79][80]
• In 2009 Mattel partnered with NeuroSky to release the Mindflex, a game that used an EEG to steer a ball through an obstacle course. By far the best selling consumer based EEG to date.^[79][81]
• In 2009 Uncle Milton Industries partnered with NeuroSky to release the Star Wars Force Trainer, a game designed to create the illusion of possessing the Force.^[79][82]
• In 2009 Emotiv released the EPOC, a 14 channel EEG device. The EPOC is the first commercial BCI to not use dry sensor technology, requiring users to apply a saline solution to electrode pads (which need remoistening after an hour or two of use).^[83]
• In 2010, NeuroSky added a blink and electromyography function to the MindSet.^[84]
• In 2011, NeuroSky released the MindWave, an EEG device designed for educational purposes and games.^[85][86] The MindWave won the Guinness Book of World Records award for "Heaviest machine moved using a brain control interface".^[87]
• In 2012, a Japanese gadget project, neurowear, released Necomimi: a headset with motorized cat ears. The headset is a NeuroSky MindWave unit with two motors on the headband where a cat's ears might be. Slipcovers shaped like cat ears sit over the motors so that as the device registers emotional states the ears move to relate. For example, when relaxed, the ears fall to the sides and perk up when excited again.
• In 2014, OpenBCI released an eponymous open source brain-computer interface after a successful kickstarter campaign in 2013. The basic OpenBCI has 8 channels, expandable to 16, and supports EEG, EKG, and EMG. The OpenBCI is based on the Texas Instruments ADS1299 IC and the Arduino or PIC microcontroller, and costs $399 for the basic version. It uses standard metal cup electrodes and conductive paste.
• In 2015, Mind Solutions Inc released the smallest consumer BCI to date, the NeuroSync. This device functions as a dry sensor at a size no larger than a Bluetooth ear piece.^[88]
• In 2015, A Chinese-based company Macrotellect released BrainLink Pro and BrainLink Lite, a consumer grade EEG wearable product providing 20 brain fitness enhancement Apps on Apple and Android App Stores.^[89]

Future research

The EEG has been used for many purposes besides the conventional uses of clinical diagnosis and conventional cognitive neuroscience. An early use was during World War II by the U.S. Army Air Corps to screen out pilots in danger of having seizures;^[90] long-term EEG recordings in epilepsy patients are still used today for seizure prediction. Neurofeedback remains an important extension, and in its most advanced form is also attempted as the basis of brain computer interfaces. The EEG is also used quite extensively in the field of neuromarketing.

The EEG is altered by drugs that affect brain functions, the chemicals that are the basis for psychopharmacology. Berger's early experiments recorded the effects of drugs on EEG. The science of pharmaco-electroencephalography has developed methods to identify substances that systematically alter brain functions for therapeutic and recreational use.

Honda is attempting to develop a system to enable an operator to control its Asimo robot using EEG, a technology it eventually hopes to incorporate into its automobiles.^[91]

EEGs have been used as evidence in criminal trials in the Indian state of Maharashtra.^[92][93]

A lot of research is currently being carried out in order to make EEG devices smaller, more portable and easier to use. So called "Wearable EEG" is based upon creating low power wireless collection electronics and ‘dry’ electrodes which do not require a conductive gel to be used.^[94] Wearable EEG aims to provide small EEG devices which are present only on the head and which can record EEG for days, weeks, or months at a time, as ear-EEG. Such prolonged and easy-to-use monitoring could make a step change in the diagnosis of chronic conditions such as epilepsy, and greatly improve the end-user acceptance of BCI systems.^[95] Research is also being carried out on identifying specific solutions to increase the battery lifetime of Wearable EEG devices through the use of the data reduction approach. For example, in the context of epilepsy diagnosis, data reduction has been used to extend the battery lifetime of Wearable EEG devices by intelligently selecting, and only transmitting, diagnostically relevant EEG data.^[96]

EEG signals from musical performers were used to create instant compositions and one CD by the Brainwave Music Project, run at the Computer Music Center at Columbia University by Brad Garton and Dave Soldier.

Event-related functional magnetic resonance imaging

From Wikipedia, the free encyclopedia

Event-related functional magnetic resonance imaging (efMRI) is a technique in magnetic resonance imaging that can be used to detect changes in the BOLD (Blood Oxygen Level Dependent) hemodynamic response to neural activity in response to certain events. Within fMRI methodology, there are two different ways that are typically employed to present stimuli. One method is a block related design, in which two or more different conditions are alternated in order to determine the differences between the two conditions, or a control may be included in the presentation occurring between the two conditions. By contrast, event related designs are not presented in a set sequence; the presentation is randomized and the time in between stimuli can vary. efMRI attempts to model the change in fMRI signal in response to neural events associated with behavioral trials. According to D'Esposito, "event-related fMRI has the potential to address a number of cognitive psychology questions with a degree of inferential and statistical power not previously available." Each trial can be composed of one experimentally controlled (such as the presentation of a word or picture) or a participant mediated "event" (such as a motor response). Within each trial, there are a number of events such as the presentation of a stimulus, delay period, and response. If the experiment is properly set up and the different events are timed correctly, efMRI allows a person to observe the differences in neural activity associated with each event.

History

Positron Emission Tomography (PET), was the most frequently used brain mapping technique before the development of fMRI. There are a number of advantages that are presented in comparison to PET. According to D’Esposito, they include that fMRI “does not require an injection of radioisotope into participants and is otherwise noninvasive, has better spatial resolution, and has better temporal resolution."^[2] The first MRI studies employed the use of “exogenous paramagnetic tracers to map changes in cerebral blood volume”,^[3][4] which allowed for the assessment of brain activity over several minutes. This changed with two advancements to MRI, the rapidness of MRI techniques were increased to 1.5 Tesla by the end of the 1980s, which provided a 2-d image. Next, endogenous contrast mechanisms were discovered by Detre, Koretsky, and colleagues was based on the net longitudinal magnetization within an organ, and a “second based on changes in the magnetic susceptibility induced by changing net tissue deoxyhemoglobin content”,^[3] which has been labeled BOLD contrast by Siege Ogawa. These discoveries served as inspiration for future brain mapping advancements. This allowed researchers to develop more complex types of experiments, going beyond observing the effects of single types of trials. When fMRI was developed one of its major limitations was the inability to randomize trials, but the event related fMRI fixed this problem.^[2] Cognitive subtraction was also an issue, which tried to correlate cognitive-behavioral differences between tasks with brain activity by pairing two tasks that are assumed to be matched perfectly for every sensory, motor, and cognitive process except the one of interest.^[2] Next, a push for the improvement of temporal resolution of fMRI studies led to the development of event-related designs, which according to Peterson, was inherited from ERP research in electrophysiology, but it was discovered that this averaging did not apply very well to the hemodynamic response because the response from trials could overlap. As a result, random jittering of the events was applied, which meant that the time repetition was varied and randomized for the trials in order to ensure that the activation signals did not overlap.

Hemodynamic response

In order to function, neurons require energy which is supplied by blood flow. Although it is not completely understood, the hemodynamic response has been correlated with neuronal activity, that is, as the activity level increases, the amount of blood used by neurons increases. This response takes several seconds to completely develop. Accordingly, fMRI has limited temporal resolution. The hemodynamic response is the basis for the BOLD (Blood Oxygen Level Dependent) contrast in fMRI.^[5] The hemodynamic response occurs within seconds of the presented stimuli, but it is essential to space out the events in order to ensure that the response being measured is from the event that was presented and not from a prior event. Presenting stimuli in a more rapid sequence allows experimenters to run more trials and gather more data, but this is limited by the slow course of hemodynamic response, which generally must be allowed to return baseline before the presentation of another stimulus. According to Burock “as the presentation rate increases in the random event related design, the variance in the signal increases thereby increasing the transient information and ability to estimate the underlying hemodynamic response”.^[3]

Rapid event-related efMRI

In a typical efMRI, after every trial the hemodynamic response is allowed to return to baseline. In rapid event-related fMRI, trials are randomized and the HRF is deconvolved afterwards. In order for this to be possible, every possible combination of trial sequences must be used and the inter-trial intervals jittered so that the time in between trials is not always the same.

Advantages of efMRI

1. Ability to randomize and mix different types of events, which ensures that one event isn’t influenced by others and not affected by the cognitive state of an individual, doesn’t allow for predictability of events.
2. Events can be organized into categories after the experiment based on the subjects behavior
3. The occurrence of events can be defined by the subject
4. Sometimes the blocked event design cannot be applied to an event.
5. Treating stimuli, even when blocked, as separate events can potentially result in a more accurate model.
6. Rare events can be measured.^[1]

Chee argues that event related designs provide a number of advantages in language-related tasks, including the ability to separate correct and incorrect responses, and show task dependent variations in temporal response profiles.^[6]

Disadvantages of efMRI

1. More complex design and analysis.
2. Need to increase the number of trials because the MR signal is small.
3. Some events are better blocked.
4. Timing issues: sampling (fix: random jitter, varying the timing of the presentation of the stimuli, allows for a mean hemodynamic response to be calculated at the end).
5. Blocked designs have higher statistical power.^[6]
6. Easier to identify artifacts arising from non-physiologic signal fluctuations.,.^[1][6]

Statistical analysis

In fMRI data, it is assumed that there is a linear relationship between neural stimulation and the BOLD response. The use of GLMs allows for the development of a mean to represent the mean hemodynamic response within the participants. Statistical Parametric Mapping is used to produce a design matrix, which includes all of the different response shapes produced during the event. For more information on this, see Friston (1997).^[7]

Applications

• Visual Priming and Object Recognition
• Examining differences between parts of a task
• Changes over time
• Memory Research - Working Memory using cognitive subtraction
• Deception - Truth from Lies
• Face Perception
• Imitation Learning
• Inhibition
• Stimulus Specific Responses

Teleoperating robots with virtual reality: getting inside a robot’s head

Jobless video-gamer alert

October 6, 2017
Original link:  http://www.kurzweilai.net/teleoperating-robots-with-virtual-reality-getting-inside-a-robots-head

A new VR system from MIT’s Computer Science and Artificial Intelligence Laboratory could make it easy for factory workers to telecommute. (credit: Jason Dorfman, MIT CSAIL)

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a virtual-reality (VR) system that lets you teleoperate a robot using an Oculus Rift or HTC Vive VR headset.

CSAIL’s “Homunculus Model” system (the classic notion of a small human sitting inside the brain and controlling the actions of the body) embeds you in a VR control room with multiple sensor displays, making it feel like you’re inside the robot’s head. By using gestures, you can control the robot’s matching movements to perform various tasks.

The system can be connected either via a wired local network or via a wireless network connection over the Internet. (The team demonstrated that the system could pilot a robot from hundreds of miles away, testing it on a hotel’s wireless network in Washington, DC to control Baxter at MIT.)

According to CSAIL postdoctoral associate Jeffrey Lipton, lead author on an open-access arXiv paper about the system (presented this week at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) in Vancouver), “By teleoperating robots from home, blue-collar workers would be able to telecommute and benefit from the IT revolution just as white-collars workers do now.”

Jobs for video-gamers too


The researchers imagine that such a system could even help employ jobless video-gamers by “game-ifying” manufacturing positions. (Users with gaming experience had the most ease with the system, the researchers found in tests.)

Homunculus Model system. A Baxter robot (left) is outfitted with a stereo camera rig and various end-effector devices. A virtual control room (user’s view, center), generated on an Oculus Rift CV1 headset (right), allows the user to feel like they are inside Baxter’s head while operating it. Using VR device controllers, including Razer Hydra hand trackers used for inputs (right), users can interact with controls that appear in the virtual space — opening and closing the hand grippers to pick up, move, and retrieve items. A user can plan movements based on the distance between the arm’s location marker and their hand while looking at the live display of the arm. (credit: Jeffrey I. Lipton et al./arXiv).

To make these movements possible, the human’s space is mapped into the virtual space, and the virtual space is then mapped into the robot space to provide a sense of co-location.

The team demonstrated the Homunculus Model system using the Baxter humanoid robot from Rethink Robotics, but the approach could work on other robot platforms, the researchers said.

In tests involving pick and place, assembly, and manufacturing tasks (such as “pick an item and stack it for assembly”) comparing the Homunculus Model system with existing state-of-the-art automated remote-control, CSAIL’s Homunculus Model system had a 100% success rate compared with a 66% success rate for state-of-the-art automated systems. The CSAIL system was also better at grasping objects 95 percent of the time and 57 percent faster at doing tasks.*

“This contribution represents a major milestone in the effort to connect the user with the robot’s space in an intuitive, natural, and effective manner.” says Oussama Khatib, a computer science professor at Stanford University who was not involved in the paper.

The team plans to eventually focus on making the system more scalable, with many users and different types of robots that are compatible with current automation technologies.

* The Homunculus Model system solves a delay problem with existing systems, which use a GPU or CPU, introducing delay. 3D reconstruction from the stereo HD cameras is instead done by the human’s visual cortex, so the user constantly receives visual feedback from the virtual world with minimal latency (delay). This also avoids user fatigue and nausea caused by motion sickness (known as simulator sickness) generated by “unexpected incongruities, such as delays or relative motions, between proprioception and vision [that] can lead to the nausea,” the researchers explain in the paper.

MITCSAIL | Operating Robots with Virtual Reality

---

Abstract of Baxter’s Homunculus: Virtual Reality Spaces for Teleoperation in Manufacturing

Expensive specialized systems have hampered development of telerobotic systems for manufacturing systems. In this paper we demonstrate a telerobotic system which can reduce the cost of such system by leveraging commercial virtual reality(VR) technology and integrating it with existing robotics control software. The system runs on a commercial gaming engine using off the shelf VR hardware. This system can be deployed on multiple network architectures from a wired local network to a wireless network connection over the Internet. The system is based on the homunculus model of mind wherein we embed the user in a virtual reality control room. The control room allows for multiple sensor display, dynamic mapping between the user and robot, does not require the production of duals for the robot, or its environment. The control room is mapped to a space inside the robot to provide a sense of co-location within the robot. We compared our system with state of the art automation algorithms for assembly tasks, showing a 100% success rate for our system compared with a 66% success rate for automated systems. We demonstrate that our system can be used for pick and place, assembly, and manufacturing tasks.

References:

• Jeffrey I Lipton, Aidan J Fay, and Daniela Rus. Baxter’s Homunculus: Virtual Reality Spaces for Teleoperation in Manufacturing. arXiv. Mar. 3, 2017 (open-access)

Misinformation effect

From Wikipedia, the free encyclopedia

The misinformation effect happens when a person's recall of episodic memories becomes less accurate because of post-event information. For example, in a study published in 1994, subjects were initially shown one of two different series of slides that depicted a college student at the university bookstore, with different objects of the same type changed in some slides. One version of the slides would, for example, show a screwdriver while the other would show a wrench, and the audio narrative accompanying the slides would only refer to the object as a "tool". In the second phase, subjects would read a narrative description of the events in the slides, except this time a specific tool was named, which would be the incorrect tool half the time. Finally, in the third phase, subjects had to list five examples of specific types of objects, such as tools, but were told to only list examples which they had not seen in the slides. Subjects who had read an incorrect narrative were far less likely to list the written object (which they hadn't actually seen) than the control subjects (28% vs. 43%), and were far more likely to incorrectly list the item which they had actually seen (33% vs. 26%).

The misinformation effect is a prime example of retroactive interference, which occurs when information presented later interferes with the ability to retain previously encoded information. Essentially, the new information that a person receives works backward in time to distort memory of the original event.^[3] The misinformation effect has been studied since the mid-1970s. Elizabeth Loftus is one of the most influential researchers in the field. It reflects two of the cardinal sins of memory: suggestibility, the influence of others' expectations on our memory; and misattribution, information attributed to an incorrect source. Research on the misinformation effect has uncovered concerns about the permanence and reliability of memory.^[4]

Visual display of retroactive memory interference

Basic methods

A recreation of the type of image used by Loftus et. al in their 1978 work. Two versions of the same image, one showing a "stop" sign and the other a "yield" sign.

Loftus, Miller, and Burns (1978) conducted the original misinformation effect study. Participants were shown a series of slides, one of which featured a car stopping in front of a stop sign. After viewing the slides, participants read a description of what they saw. Some of the participants were given descriptions that contained misinformation, which stated that the car stopped at a yield sign. Following the slides and the reading of the description, participants were tested on what they saw. The results revealed that participants who were exposed to such misinformation were more likely to report seeing a yield sign than participants who were not misinformed.^[5]

Similar methods continue to be used in misinformation effect studies. Today, standard methods involve showing subjects an event, usually in the form of a slideshow or video. The event is followed by a time delay and introduction of post-event information. Finally, participants are retested on their memory of the original event.^[6] This original study by Loftus et al. paved the way for multiple replications of the effect in order to test things like what specific processes cause the effect to occur in the first place and how individual differences influence susceptibility to the effect.

Neurological causes

Functional magnetic resonance imaging (fMRI) from 2010 pointed to certain brain areas which were especially active when false memories were retrieved. participants studied photos during an fMRI. Later, they viewed sentences describing the photographs, some of which contained information conflicting with the photographs, i.e. misinformation. One day later, participants returned for a surprise item memory recognition test on the content of the photographs. Results showed that some participants created false memories, reporting the verbal misinformation conflicting with the photographs.^[7] During the original event phase, increased activity in left fusiform gyrus and right temporal/occipital cortex was found which may have reflected the attention to visual detail,associated with later accurate memory for the critical item(s) and thus resulted in resistance to the effects of later misinformation.^[7] Retrieval of true memories was associated with greater reactivation of sensory-specific cortices, for example, the occipital cortex for vision.^[7]. Electroencephalography research on this issue also suggests that the retrieval of false memories is associated with reduced attention and recollection related processing relative to true memories.^[8]

Susceptibility

It is important to note that not everyone is equally susceptible to the misinformation effect. Individual traits and qualities can either increase or decrease one's susceptibility to recalling misinformation.^[5] Such traits and qualities include: age, working memory capacity, personality traits and imagery abilities.

Age

Several studies have focused on the influence of the misinformation effect on various age groups.^[9] Young children are more susceptible than older children and adults to the misinformation effect.^[9] Additionally, elderly adults are more susceptible than younger adults.^[9][10]

Working memory capacity

Individuals with greater working memory capacity are better able to establish a more coherent image of an original event. Participants performed a dual task: simultaneously remembering a word list and judging the accuracy of arithmetic statements. Participants who were more accurate on the dual task were less susceptible to the misinformation effect. This, in turn, allowed them to reject the misinformation.^[5][11]

Personality traits

The Myers Briggs Type Indicator is one type of test used to assess participant personalities. Individuals were presented with the same misinformation procedure as that used in the original Loftus et al. study in 1978 (see above). The results were evaluated in regards to their personality type. Introvert-intuitive participants were more likely to accept both accurate and inaccurate postevent information than extrovert-sensate participants. Therefore, it was speculated that introverts are more likely to have lower confidence in their memory and are more likely to accept misinformation.^[5][12] Individual personality characteristics, including empathy, absorption and self-monitoring, have also been linked to greater susceptibility.^[9]

Imagery abilities

The misinformation effect has been examined in individuals with varying imagery abilities. Participants viewed a filmed event followed by descriptive statements of the events in a traditional three-stage misinformation paradigm. Participants with higher imagery abilities were more susceptible to the misinformation effect than those with lower abilities. The psychologists argued that participants with higher imagery abilities were more likely to form vivid images of the misleading information at encoding or at retrieval, therefore increasing susceptibility.^[5][13]

Influential factors

Time

Individuals may not be actively rehearsing the details of a given event after encoding. The longer the delay between the presentation of the original event and post-event information, the more likely it is that individuals will incorporate misinformation into their final reports.^[6] Furthermore, more time to study the original event leads to lower susceptibility to the misinformation effect, due to increased rehearsal time.^[6] Elizabeth Loftus coined the term discrepancy detection principle for her observation that a person´s recollections are more likely to change, if they do not immediately detect the discrepancies between misinformation and the original event.^[9][14] At times people recognize a discrepancy between their memory and what they are being told.^[15] People might recollect, "I thought I saw a stop sign, but the new information mentions a yield sign, I guess I must be wrong, it was a yield sign."^[15] Although the individual recognizes the information as conflicting with their own memories they still adopt it as true.^[9] If these discrepancies are not immediately detected they are more likely to be incorporated into memory.^[9]

Source reliability

The more reliable the source of the post-event information, the more likely it is that participants will adopt the information into their memory.^[6] For example, Dodd and Bradshaw (1980) used slides of a car accident for their original event. They then had misinformation delivered to half of the participants by an unreliable source: a lawyer representing the driver. The remaining participants were presented with misinformation, but given no indication of the source. The misinformation was rejected by those who received information from the unreliable source and adopted by the other group of subjects.^[6]

Discussion and rehearsal

The question of whether discussion is detrimental to memories also exists when considering what factors influence the misinformation effect. One particular study examined the effects of discussion in groups on recognition. The experimentors used three different conditions: discussion in groups with a confederate providing misinformation, discussion in groups with no confederate, and a no-discussion condition. They found that participants in the confederate condition adopted the misinformation provided by the confederate. However, there was no difference between the no-confederate and no-discussion conditions, proving that discussion (without misinformation) is neither harmful nor beneficial to memory accuracy.^[16] In an additional study, Karns et al. (2009) found that collaborative pairs showed a smaller misinformation effect than individuals. It appeared as though collaborative recall allowed witnesses to dismiss misinformation generated by an inaccurate narrative.^[17] In a 2011 study, Paterson et al. studied "memory conformity", showing students two different videos of a burglary. It was found that if witnesses who had watched the two different videos talked with one another, they would then claim to remember details shown in the video of the other witness and not their own. They continued to claim the veracity of this memory, despite warnings of misinformation.^[18]

State of mind

Various inhibited states of mind such as drunkenness and hypnosis can increase misinformation effects.^[9] Assefi and Garry (2002) found that participants who believed they had consumed alcohol showed results of the misinformation effect on recall tasks.^[19] The same was true of participants under the influence of hypnosis.^[20]

Other

Most obviously, leading questions and narrative accounts can change episodic memories and thereby affect witness' responses to questions about the original event. Additionally, witnesses are more likely to be swayed by misinformation when they are suffering from alcohol withdrawal^[17][21] or sleep deprivation,^[17][22] when interviewers are firm as opposed to friendly,^[17][23] and when participants experience repeated questioning about the event.^[17][24]

Arousal after learning

Arousal induced after learning reduces source confusion, allowing participants to better retrieve accurate details and reject misinformation. In a study of how to reduce the misinformation effect, participants viewed four short film clips, each followed by a retention test, which for some participants included misinformation. Afterward, participants viewed another film clip that was either arousing or neutral. One week later, the arousal group recognized significantly more details and endorsed significantly fewer misinformation items than the neutral group.^[25]

Anticipation

Educating participants about the misinformation effect can enable them to resist its influence. However, if warnings are given after the presentation of misinformation, they do not aid participants in discriminating between original and post-event information.^[9]

Psychotropic placebos

Research published 2008 showed that placebos enhanced memory performance. participants were given a phoney "cognitive enhancing drug" called R273. When they participated in a misinformation effect experiment, people who took R273 were more resistant to the effects of misleading postevent information.^[26] As a result of taking R273, people used stricter source monitoring and attributed their behavior to the placebo and not to themselves.^[26]

Implications

Implications of this effect on long-term memories are as follows:

Variability

Some reject the notion that misinformation always causes impairment of original memories.^[9] Modified tests can be used to examine the issue of long-term memory impairment.^[9] In one example of such a test,(1985) participants were shown a burglar with a hammer.^[27] Standard post-event information claimed the weapon was a screwdriver and participants were likely to choose the screwdriver rather than the hammer as correct. In the modified test condition, postevent information was not limited to one item,instead participants had the option of the hammer and another tool (a wrench, for example). In this condition, participants generally chose the hammer, showing that there was no memory impairment.^[27]

Rich false memories

Rich false memories are researchers' attempts to plant entire memories of events which never happened in participants' memories. Examples of such memories include fabricated stories about participants getting lost in the supermarket or shopping mall as children. Researchers often rely on suggestive interviews and the power of suggestion from family members, known as “familial informant false narrative procedure.”^[9] Around 30% of subjects have gone on to produce either partial or complete false memories in these studies.^[9] There is a concern that real memories and experiences may be surfacing as a result of prodding and interviews. To deal with this concern, many researchers switched to implausible memory scenarios.^[9]

Daily applications

The misinformation effect can be observed in many suituations. For example, after witnessing a crime or accident there may be opportunities for witnesses to interact and share information. Late-arriving bystanders or members of the media may ask witnesses to recall the event before law enforcement or legal representatives have the opportunity to interview them.^[17] Collaborative recall may lead to a more accurate account of what happened, as opposed to individual responses that may contain more untruths after the fact.^[17]

In addition, while remembering small details may not seem important, they can matter tremendously in certain situations. A jury's perception of a defendant's guilt or innocence could depend on such a detail. If a witness remembers a moustache or a weapon when there was none, the wrong person may be wrongly convicted.

IBM scientists say radical new ‘in-memory’ computing architecture will speed up computers by 200 times

New architecture to enable ultra-dense, low-power, massively-parallel computing systems optimized for AI

October 25, 2017
Original link:  http://www.kurzweilai.net/ibm-scientists-say-radical-new-in-memory-computing-architecture-will-speed-up-computers-by-200-times

(Left) Schematic of conventional von Neumann computer architecture, where the memory and computing units are physically separated. To perform a computational operation and to store the result in the same memory location, data is shuttled back and forth between the memory and the processing unit. (Right) An alternative architecture where the computational operation is performed in the same memory location. (credit: IBM Research)

IBM Research announced Tuesday (Oct. 24, 2017) that its scientists have developed the first “in-memory computing” or “computational memory” computer system architecture, which is expected to yield 200x improvements in computer speed and energy efficiency — enabling ultra-dense, low-power, massively parallel computing systems.

Their concept is to use one device (such as phase change memory or PCM*) for both storing and processing information. That design would replace the conventional “von Neumann” computer architecture, used in standard desktop computers, laptops, and cellphones, which splits computation and memory into two different devices. That requires moving data back and forth between memory and the computing unit, making them slower and less energy-efficient.

The researchers used PCM devices made from a germanium antimony telluride alloy, which is stacked and sandwiched between two electrodes. When the scientists apply a tiny electric current to the material, they heat it, which alters its state from amorphous (with a disordered atomic arrangement) to crystalline (with an ordered atomic configuration). The IBM researchers have used the crystallization dynamics to perform computation in memory. (credit: IBM Research)

Especially useful in AI applications

The researchers believe this new prototype technology will enable ultra-dense, low-power, and massively parallel computing systems that are especially useful for AI applications. The researchers tested the new architecture using an unsupervised machine-learning algorithm running on one million phase change memory (PCM) devices, successfully finding temporal correlations in unknown data streams.

“This is an important step forward in our research of the physics of AI, which explores new hardware materials, devices and architectures,” says Evangelos Eleftheriou, PhD, an IBM Fellow and co-author of an open-access paper in the peer-reviewed journal Nature Communications. “As the CMOS scaling laws break down because of technological limits, a radical departure from the processor-memory dichotomy is needed to circumvent the limitations of today’s computers.”

“Memory has so far been viewed as a place where we merely store information, said Abu Sebastian, PhD. exploratory memory and cognitive technologies scientist, IBM Research and lead author of the paper. But in this work, we conclusively show how we can exploit the physics of these memory devices to also perform a rather high-level computational primitive. The result of the computation is also stored in the memory devices, and in this sense the concept is loosely inspired by how the brain computes.” Sebastian also leads a European Research Council funded project on this topic.

* To demonstrate the technology, the authors chose two time-based examples and compared their results with traditional machine-learning methods such as k-means clustering:

• Simulated Data: one million binary (0 or 1) random processes organized on a 2D grid based on a 1000 x 1000 pixel, black and white, profile drawing of famed British mathematician Alan Turing. The IBM scientists then made the pixels blink on and off with the same rate, but the black pixels turned on and off in a weakly correlated manner. This means that when a black pixel blinks, there is a slightly higher probability that another black pixel will also blink. The random processes were assigned to a million PCM devices, and a simple learning algorithm was implemented. With each blink, the PCM array learned, and the PCM devices corresponding to the correlated processes went to a high conductance state. In this way, the conductance map of the PCM devices recreates the drawing of Alan Turing.
  
• Real-World Data: actual rainfall data, collected over a period of six months from 270 weather stations across the USA in one hour intervals. If rained within the hour, it was labelled “1” and if it didn’t “0”. Classical k-means clustering and the in-memory computing approach agreed on the classification of 245 out of the 270 weather stations. In-memory computing classified 12 stations as uncorrelated that had been marked correlated by the k-means clustering approach. Similarly, the in-memory computing approach classified 13 stations as correlated that had been marked uncorrelated by k-means clustering. 

---

Abstract of Temporal correlation detection using computational phase-change memory


Conventional computers based on the von Neumann architecture perform computation by repeatedly transferring data between their physically separated processing and memory units. As computation becomes increasingly data centric and the scalability limits in terms of performance and power are being reached, alternative computing paradigms with collocated computation and storage are actively being sought. A fascinating such approach is that of computational memory where the physics of nanoscale memory devices are used to perform certain computational tasks within the memory unit in a non-von Neumann manner. We present an experimental demonstration using one million phase change memory devices organized to perform a high-level computational primitive by exploiting the crystallization dynamics. Its result is imprinted in the conductance states of the memory devices. The results of using such a computational memory for processing real-world data sets show that this co-existence of computation and storage at the nanometer scale could enable ultra-dense, low-power, and massively-parallel computing systems.

References:

• Abu Sebastian et al. Temporal correlation detection using computational phase-change memory. Nature Communications 8, Article number: 1115 (2017) doi:10.1038/s41467-017-01481-9 (open access)

Researchers Detect a Global Drop in Fires

Original link:  https://earthobservatory.nasa.gov/images/90493

(DJS):  Climate change is causing wildfires to increase, right? Wrong again. 

2003 - 2015

Across the grasslands of Asia, the tropical forests of South America, and the savannas of Africa, shifting livelihoods are leading to a significant decline in burned area. Using NASA satellites to detect fires and burn scars from space, researchers have found that an ongoing transition from nomadic cultures to settled lifestyles and intensifying agriculture has led to a steep drop in the use of fire for land clearing and an overall drop in natural and human-caused fires worldwide.

Globally, the total acreage burned by fires declined 24 percent between 1998 and 2015, according to a new paper published in Science. Scientists determined that the decline in burned area was greatest in savannas and grasslands, where fires are essential for maintaining healthy ecosystems and habitat conservation.

The map above, based on data from the international research team, shows annual trends in burned area over the study period. Blues represent areas where the trend was toward less burning, whether natural or human-caused, while red areas had more burning. The line plot shows the annual fluctuations in global burned area and the overall downward trend. The research team, led by Niels Andela of NASA’s Goddard Space Flight Center, analyzed fire data derived from the Moderate Resolution Imaging Spectrometer (MODIS) instruments on NASA’s Terra and Aqua satellites. They then compared these data sets with regional and global trends in agriculture and socio-economic development.

Across Africa, fires collectively burned an area about half the size of the continental United States every year. In traditional savanna cultures, people often set fires to keep grazing lands productive and free of shrubs and trees. But as many of these communities have shifted to cultivating permanent fields and building more houses, roads, and villages, the use of fire has declined. As this economic development continues, the landscape becomes more fragmented and communities then enact legislation to control fires. This leads the burned area to decline even more.

By 2015, savanna fires in Africa had declined by 700,000 square kilometers (270,000 square miles)—an area the size of Texas. “When land use intensifies on savannas, fire is used less and less as a tool,” Andela said. “As soon as people invest in houses, crops, and livestock, they don’t want these fires close by anymore. The way of doing agriculture changes, the practices change, and fire disappears from the grassland landscape.”

A slightly different pattern occurs in tropical forests and other humid regions near the equator. Fire rarely occurs naturally in these forests; but as humans settle an area, they often use fire to clear land for cropland and pastures. As more people move into these areas and increase the investments in agriculture, they set fewer fires and the burned area declines again.

The changes in savanna, grassland, and tropical forest fire patterns are so large that they have so far offset some of the increased risk of fire caused by global warming, said Doug Morton, a forest scientist at NASA Goddard and a co-author of the study. The impact of a warming and drying climate is more obvious at higher latitudes, where fire has increased in Canada and the American West. Regions of China, India, Brazil, and southern Africa also showed increases in burned area.

“Climate change has increased fire risk in many regions, but satellite burned area data show that human activity has effectively counterbalanced that climate risk, especially across the global tropics,” Morton said. “We’ve seen a substantial global decline over the satellite record, and the loss of fire has some really important implications for the Earth system.”

Fewer and smaller fires on the savanna mean that there are more trees and shrubs instead of open grasslands. This is a significant change in habitat for the region’s iconic mammals like elephants, rhinoceroses, and lions. “Humans are interrupting the ancient, natural cycle of burning and regrowth in these areas,” said senior author Jim Randerson of the University of California, Irvine. “Fire had been instrumental for millennia in maintaining healthy savannas, keeping shrubs and trees at bay and eliminating dead vegetation.”

There are benefits to fewer fires as well. Regions with less fire saw a decrease in carbon monoxide emissions and an improvement in air quality during fire season. With less fire, savanna vegetation is increasing—taking up more carbon dioxide from the atmosphere.

But the decline in burned area from human activity raises some tricky questions. “For fire-dependent ecosystems like savannas,” Morton said, “the challenge is to balance the need for frequent burning to maintain habitat for large mammals and to maintain biodiversity while protecting people’s property, air quality, and agriculture.”

Click here to explore the data.
NASA Earth Observatory images by Joshua Stevens, using GFED4s data courtesy of Doug Morton/NASA GSFC. Story by Kate Ramsayer, NASA’s Goddard Space Flight Center.