Blue brain graphic

Brain power: new imaging raises hopes of mental healthcare revolution

Image credit: Dreamstime

While mind-reading remains in the realm of science fiction, advances in neuroimaging are putting us in a position to observe detailed brain activity in real time. Could this fresh insight reshape the way mental illness is diagnosed and treated?

Psychiatry, says Nottingham University’s Professor Peter Liddle, is “in a different phase of development” from other fields of medicine. “There are differences in practice largely because of the immense complexity of the human mind and brain,” says the psychiatrist. “The heart and the liver and the kidneys are very simple organs compared to the brain.”

When a patient complains of trouble with their “simple organs”, a doctor enquires about their symptoms, and carries out physiological tests such as blood tests or X-rays to confirm their diagnosis and decide on a course of treatment. However, physiological investigations for mental complaints are rare; structural neuroimaging can reveal crude problems such as tumours, but brains have countless, far more subtle ways of causing often distressing problems.

The best way of directly observing this is with functional neuroimaging: the field of imaging dedicated to brain function. A patient completes a task (sometimes as simple as waggling his or her toes) while the activation and connectivity of regions of the brain is monitored, providing an insight into the invisible functioning – and malfunctioning – of the mind. This could involve removing a slice of the skull and placing electrodes in the brain to measure directly any electrical signals zipping between neurons, or sticking electrodes on the scalp to sense these signals through the skull. Functional magnetic resonance imaging (fMRI) allows for the indirect observation of neural activity, as active areas of brains are indicated by blood flow; these responses, though, occur over seconds, while the ‘speed of thought’ is on a timescale of milliseconds.

Magnetoencephalography (MEG), however, offers a non-invasive way to measure the extremely faint magnetic fields arising from the brain’s electricity activity with millisecond and millimetre precision.

MEG is an expensive, intricate and stressful procedure with manifold ways to fail. The half-tonne MEG machines must be entombed in rooms with thick steel-lined walls to nullify the Earth’s magnetic field, and cooled with liquid helium for their hundreds of Squid (superconducting quantum interference device) detectors to work. These strict conditions mean that patients must sit unmoving with their head in a hollow in the machine, the supercooled detectors centimetres from their scalp.

Although these conditions are generally unenviable, they are particularly inadequate for children with autistic spectrum disorder (ASD) or attention hyperactivity disorder; even if they can sit still for half an hour slotted inside the machine, the gap between the Squids and their brains causes a significant loss in signal.

“It is a fixed system, one size fits all, so adults and children go in the same machine and for smaller heads the sensors will be further away,” says Elena Boto, a PhD student at the University of Nottingham. “Because what we are measuring is the magnetic fields created by the brain, these fields decrease in amplitude with the square of distance; the further we are the less signal we get.”

MEG machines are so costly to build and run that there is virtually no chance of hospitals installing a range of them to suit different head sizes and there are only a handful of dedicated infant MEG machines in the world.

A project led by UK scientists aims to do away with these hefty supercooled machines and replace them with adjustable helmets lined with sensors, enabling freedom of movement and greater precision. Key to this project is the replacement of Squids with a detector which can operate at room temperature: the optically pumped magnetometer (OPM). While OPMs emerged half a century ago, only recently have they been refined to detect the feeble magnetic signals emanating from the brain.

“Optical magnetometers use lasers and the fundamental properties of atoms to detect magnetic fields,” Boto explains. “The sensors have a little tiny cell of atoms. When the light polarises all the atoms, the laser is fully transmitted through the cell and the transmission is [at a] maximum, but when there’s a magnetic field around the atom, their atomic spin changes and you see a drop in the light of the laser. So by measuring the change in the intensity of the laser you have a measurement of the magnetic field interacting with the atoms.”

The University of Nottingham team, working with collaborators at University College London and supported by the Wellcome Trust, have already built a prototype MEG helmet with a handful of OPM sensors monitoring activity in one area of the brain. They are working towards an adjustable helmet filled with hundreds of OPMs to give similar coverage to Squid-based MEG machines.

These helmets would allow patients to move around, opening up the possibility of using virtual reality to conjure a less stressful environment, as well as for research purposes. For instance, a volunteer with post-traumatic stress disorder (PTSD) may be exposed to stressful virtual scenarios while wearing a MEG helmet, helping identify abnormal neural activity associated with PTSD which renders patients more susceptible to panic attacks. The Nottingham researchers also suggest that eventually this flexible approach to MEG could be used to monitor everyday brain activity relating to stress, anxiety, pleasure and exertion: essentially fitness trackers for the brain.

More significant than the freedom of movement allowed by an OPM-MEG helmet is the simple fact that this room-temperature system places the detectors much closer to the brain.

“We’re talking about a system that is much more sensitive and it’s not rocket science how it gets more sensitive, it’s just closer to the head,” says Professor Matt Brookes, the Nottingham physicist leading the project. “I think it does offer us a way to get at those really severely affected patients and obviously we can get at children. I think it really stands to benefit the neuroscience and clinical community.”

Brookes and his colleagues are achieving approximately four times greater sensitivity than was previously possible for MEG, and they believe that they could soon reach ten times greater sensitivity for adults with even stronger results for babies and children.

Although these OPM-MEG helmets are not yet commercially available, they are close to application in helping manage childhood epilepsy, which is caused by abnormal electric discharges disrupting normal neural activity and causing seizures. In severe cases, neurosurgeons remove the damaged part of the child’s brain. MEG is already used to help plan this surgery, although OPM-MEG would provide far more detailed information, which could otherwise only be acquired through invasive procedures.

“If you’re trying to monitor how an epileptic child’s brain is varying over several days you have to take the top of the skull out and put electrodes on the surface of the brain, which is brutal,” says Professor Mark Fromhold, a physicist involved with the project. “That’s the kind of area where initial applications have got much less ethical clearance because the choice is: do you want to do very invasive surgery or can you find out what is going on using [non-invasive] instruments?”

In the coming years, it is likely that this technology could prove useful in observing conditions such as mild traumatic brain injury (mTBI), which often affects soldiers and athletes, disrupting their brain function and affecting their attention span and memory. Sufferers often go undiagnosed without structural damage to the head, and would likely benefit from the rollout of a more practical MEG platform.

The most challenging and intriguing potential application for OPM-MEG, however, is in psychiatry. Structural and functional neuroimaging has already been used to identify hundreds of markers of conditions including schizophrenia, Alzheimer’s disease, chronic alcoholism and mood disorders, although these markers are unhelpful on the individual level.

“If you were to ask me what brain abnormalities have been identified with schizophrenia then I could probably list 100 brain features but these are all subtle and they can only be reliably detected if I measure that in 30 or more patients and 30 or more controls,” says Liddle, who has studied schizophrenia using neuroimaging for decades. “Single-subject data is simply too noisy because of the complexity of the mind and brain; there’s a huge amount of variability in the brain that has not got much to do with mental health.

“The step forward has got to be increasing the sensitivity and robustness of the measurements so we can get useful stuff from a single person. What comes into the clinic is not a pack of 30 people with an identical problem to be compared with another 30 people. You have a single person.”

Liddle believes that measuring neural activity directly and precisely with MEG could be crucial for producing useful predictions for individuals, particularly when combined with more sophisticated computational methods to find patterns in neuroimaging data. This would benefit researchers as they observe the functional abnormalities (including as they progress over time) but could also be integrated into psychiatric care, informing diagnosis and intervention.

Professor Margot Taylor, director of functional neuroimaging at the Hospital for Sick Children in Toronto, dedicates her work to children with ASD and preterm-born children, but was approached by the Canadian Armed Forces with a request to investigate PTSD in army veterans. Using MEG, Taylor and her colleagues identified strong functional markers for PTSD consistent with hyperarousal, which strongly distinguished these veterans from their healthy combat-exposed peers and peers with mTBI.

Taylor hopes that if MEG research into PTSD is continued, a neuroimaging test will be developed for PTSD; this may be capable of differentiating between subtypes of the condition to guide psychiatrists choosing between courses of treatment. She says that MEG would complement – not replace – a psychiatrist’s clinical skills, but that it should become increasingly important in the field. In this sense, MEG could become the mental health equivalent of an X-ray: helping to guide diagnosis and monitor patient recovery. Psychiatrists working with MEG believe that it may be integrated into standard care in 10-15 years, with OPM-MEG instrumental to this progress.

“OPMs are going to revolutionise our MEG applications,” says Taylor. “About four years ago I thought they were just pie in the sky. Now we’re trying to buy some.”

Machine learning

Could AI help treat depression?

“Our diagnostic system has done a lot of good for patients, but it was never really designed to define diagnostic categories that have a strong correspondence with their underlying biology as they do in other forms of medicine,” says Professor Conor Liston, a Weill Cornell Medical College psychiatrist. “Depression is a particularly good example of how many different people with different sorts of symptoms get lumped together in the same box and treated the same way.”

Researchers hunting for diagnostic subgroups normally look for similar groupings of clinical symptoms, resulting in a lack of one-to-one correspondences between subgroups and measurable physiological markers.

“Our approach was to flip that upside down and ask whether we could identify clusters of patients that present with similar biological features and then, if so: do those features predict differences in clinical symptoms and treatment responses?” he explains. Liston and his collaborators collected fMRI data from 1,200 patients with depression, and then processed the data using two computational approaches.

First, they applied hierarchical clustering: an algorithm paired patients with the most similar patterns of altered neural connectivity, paired some pairs and continued to group patients until four large clusters were formed, each with similar abnormalities. Next, they trained a machine-learning algorithm to sort individual patients into these subgroups based on features of their fMRI data.

The researchers found that each patient subgroup was associated with distinctive clinical symptoms and responded differently to a standard brain stimulation therapy. Liston hopes that these computational methods will be applied to other psychiatric conditions, and eventually help improve clinical practice.

“There’s no reason why this should only work for depression,” says Liston. “I think at some point in the not too distant future we’ll be doing things differently.”

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them