Currently, there are two main ways of performing brain scans on small babies. Magnetic resonance imaging (MRI) can provide data on brain function, but MRI scanners are large and static, and the baby may need to be sedated and wheeled to the scanner for the procedure to be carried out. The other alternative, ultrasound, can be performed at the cot side and is effective at revealing brain anatomy, but cannot show how the brain is actually functioning, e.g. in terms of oxygen supply and blood flow. Combining the advantages of MRI and ultrasound but avoiding their disadvantages, the first scanner of its kind in the world, uses an innovative technique called optical tomography to generate images showing how the brain is working. In optical tomography, light passes through body tissue and is then analysed by computer to provide information about the tissue. A helmet incorporating 32 light detectors and 32 sources of completely safe, low-intensity laser light is placed on the baby’s head. The sources produce short flashes and the detectors measure the amount of light that reaches them through the brain and the time the light takes to travel. A software package also developed uses this information to build up a 3D image. This can show which parts of the brain are receiving oxygen, where blood is situated, evidence of brain damage.
In the wider context of artificial intelligence, it may be that further progress in emulating some of the complexity of the human brain will come about based on a more detailed mapping of neural tissue. At the present time, current technology cannot achieve both the resolution and sample size needed for this task. It is estimated that a minimum resolution in the order of 1-10nm is required and although 'Electron Microscopy (EM)' can provide this resolution, the samples are required to be both small and very thin. In contrast, 'Magnetic Resonance Imaging (MRI)' can achieve roughly 1 mm resolution of an intact human brain.
After scanning the brain, another major requirement will be the automated processing of images associated with brain tissue, e.g. mitochondria, nucleus and nucleolus, vesicles, synapses etc. This will require establishing 3-D structures from 2-D information that today is still manually derived. Advances in computing would suggest that this process could be automated through a combination of image recognition and signal processing algorithms. Such automation could be vital if the huge amount of neural tissue is to be scanned in any reasonable amount of time.
Messages from the senses travel so swiftly through the brain that imaging machines, such as MRI, cannot keep up with them. To track these messages in real time, scientists now use faster methods, such as 'Magneto-Encephalography (MEG)' or 'Electro-Encephalography (EEG)'. These techniques rely on large arrays of sensors or electrodes that are placed on the scalp to record the firing of brain cells. This data may then be combined with anatomical information obtained by structural MRI scans.
The next generation of scanner technology could use 'functional
MRI (fMRI)' in various combinations with MEG and EEG.
Functional MRI will allow activity deep in the brain to be shown
with high spatial resolution. However, it is relatively slow, since
it is based on blood-flow responses, which takes about 450
milliseconds. In contrast, EEG's spatial resolution is relatively
poor, but because of its speed it may help to better reveal the
sequence of events. It is hoped that continued
development in this area of medical science will help researchers determine
how the various parts of the brain exchange information and, most intriguing,
how sensory information leads to thought.