SUPPOSE doctors could use light to watch blood flowing around your heart and through your veins and arteries, or to see whether a tumour was malignant or benign without having to cut you open. And that they could do all this with a cheap, portable system in the comfort of your local health centre. If they could, routine screening for conditions such as breast, skin and other cancers might become a reality.
Doctors already have an impressive armoury of techniques for imaging the body (see The inside story), but they still cannot do everything they would like. For example, tumours are extremely difficult to detect in their early stages using X-rays or other established techniques. What is needed is a cheap, safe way for monitoring the body鈥檚 chemistry in real time, without the need for surgery or other invasive procedures, and without any harmful effects. Researchers are now wondering whether lasers 鈥� instruments more familiar as tools for cutting and burning tissue 鈥� can do the trick.
First there is the question of the wavelength or colour of the light that is used to scan the body. The shorter the wavelength, the higher the resolution that is available. But high energy (short-wavelength) X-rays can ionise tissue, and so cause tumours if they are used too often or at intensities that are too high. The light must also have exactly the right energy range to detect the key chemicals that can tell doctors how the patient鈥檚 body is functioning. To produce sharp images, it should also be of a wavelength that is not strongly scattered. Unfortunately, the most useful wavelengths from the point of view of monitoring body chemistry tend to be just those that are most strongly scattered. So the best compromise that researchers have come up with is to use light that meets the first three of these criteria, and then to try and overcome the problem of scattering with some clever technical tricks.
Advertisement
Burning issue
Many researchers are using near infrared light, with wavelengths in the range 600 to 1000 nanometres. Light in this range is not strongly absorbed by tissue, and so will be less likely to cause burns. This region of the spectrum is also convenient because there are now versatile near-infrared lasers that can be tuned to several useful wavelengths, opening up the possibility that doctors could look for different chemical species simply by tuning to different frequencies. Unlike X-rays, infrared is safe enough to be used for regular routine screening, and recent developments in low-cost infrared laser technology could soon make this a practical option.
But first it will be necessary to get to deal with the scattering problem. One approach is to investigate biological systems that do not scatter light very much. David Delpy and colleagues at University College London have developed an instrument that exploits the fact that the brains of newborn babies scatter light much less than adult brain tissue does. They place fibre-optic probes one on either side of a baby鈥檚 head and send infrared laser light through the head from one probe to the other. The idea is that measuring the absorption of light by the baby鈥檚 head at different wavelengths, they can estimate the ratio of haemoglobin to oxyhaemoglobin, and hence the degree of oxygenation of the blood in the brain.
But light scattering remains a problem where a precise image of a particular organ or lesion is required. After light has passed through more than a few hundreds of micrometres of tissue, 鈥渂allistic鈥� light which has not been scattered 鈥� and carries the most direct information about the area being imaged 鈥� is swamped by light scattered at random from the intervening tissue (see Diagram) The task is to pick out the useful, image-bearing light from the background scattered light. One approach is to capitalise on the fact that scattered light emerges from the tissue in all directions, while ballistic light emerges in the same direction it went in. So to collect ballistic light, you need to cut out everything that isn鈥檛 travelling in the right direction.
A technique for doing this, called 鈥渃onfocal imaging鈥�, is being tested by research groups in the US, Europe and Japan for clinical applications such as mammography. It shines light from a single optical fibre onto the object, and collects the light that passes through using another optical fibre (see Diagram). It builds up a two-dimensional picture point by point, by scanning the optical fibres across the object while keeping them very precisely aligned with one another. Because the collecting fibre is very thin, it acts as a 鈥渟patial鈥� filter, cutting out all the light that has not passed straight through the sample. But this technique does not cut out all the scattered light. In particular it cannot filter out 鈥渙n-line鈥� scattered light, which has been scattered out of the line of sight and then back into it. This prevents the system detecting features smaller than about 1 centimetre across when viewed through several centimetres of a scattering medium such as breast tissue.
To try to cut out this on-line scattered light, researchers have tried a technique known as time-gating, which filters the light over time rather tan through space. It relies on the fact that although some scattered light may be travelling in the same direction as the ballistic light, it will emerge rather later, since the scattering process has sent it on a longer path. The theory is that a camera with a very fast shutter could be triggered to take a picture the instant the ballistic light emerges, and close again in time to prevent any of the delayed, scattered light being recorded. But in practice, to take a picture using only the ballistic light requires an exposure time much less than a picosecond (10鈭�12 seconds) and unfortunately there are no cameras that are both this fast and sensitive enough to capture two-dimensional images in a single shot.
Coherent patterns
Researchers led by James Fujimoto and colleagues at the Massachusetts Institute of Technology and Tufts University School of Medicine in Boston have devised a method that uses optical interference to pick out ballistic light. They use the confocal technique to get rid of some of the scattered light, and then use a form of time-gating that exploits optical interference to separate the image-bearing ballistic light from the rest.
For two light beams of laser light to form an interference pattern they must be coherent 鈥� in other words they must have the same wavelength and their relative phase must be fixed. But when light is scattered it loses its original phase, and so will not interfere. This means that if an image-bearing beam is made to interfere with a reference beam, any resulting fringe pattern will come from the ballistic light, not the scattered light.
The MIT group uses this approach to build up a two-dimensional image of the object, scanning it pixel by pixel and collecting light at each spot. This boosts the weak ballistic signal and gives very high sensitivity. Fujimoto鈥檚 researchers can also resolve some features in three dimensions. They do this using an approach known as 鈥渢ime of flight鈥�. If both the object and reference beams consist of trains of very short pulses of laser light, then a fringe pattern will be recorded only if the light pulses from both beams arrive at the recording medium at exactly the same time. Changing the arrival time of the reference beam makes it possible to choose which of the image-bearing pulses interferes with the reference beam and forms the recorded pattern. The light that arrives earliest will convey an image of the front of the object, while the light arriving later will carry images from further back. In this way it is possible to build up a three dimensional image from a series of two-dimensional slices.
In 1993 Fujimoto reported that his researchers had managed to detect a ballistic signal 10诲鈭�13 times weaker than the incident light. They needed high sensitivity because the amount of ballistic light emerging from the sample falls exponentially as the sample thickness increases. But unfortunately, even the impressive sensitivity they managed to achieve is only enough to form images through a tissue depth of about three millimetres.
This is not as bad as it sounds, as there are some important clinical applications for three-dimensional imaging through a few millimetres of tissue. For example, being able to watch the flow of blood and particular chemicals, such as oxyhaemoglobin or glucose, in arteries close to the surface of the skin could provide a noninvasive way for diabetics to monitor their blood glucose levels. It might also be possible to use this approach to detect skin cancers such a malignant melanoma.
A more important drawback of this technique is that it builds up the image pixel by pixel, and so can take 10 minutes or more to scan even a relatively small sample. This makes it too slow for clinical use. My research group at Imperial College is working with the Institute of Cancer Research in London to develop a three-dimensional imaging system capable of recording depth-resolved images in a single shot, rather than scanning pixel by pixel. For this we are using an approach using time-gated holography. This is similar to Fujimoto鈥檚 approach, but instead of recording images pixel by pixel we shine a broad laser beam onto the sample and record the whole of the resulting two-dimensional interference pattern all in one go, as if we were using an array of thousands of fibres in parallel. We record the interference fringe pattern 鈥� called a hologram 鈥� on a device known as a photorefractive crystal, which does not need to be developed in the way that photographic film does, and is reuseable. An additional laser beam, or else the original reference beam, can be used to reconstruct the object beam from the holographic interference pattern. This reconstructed image may be viewed in real time by a TV camera and then stored on a computer. Alternatively, a doctor could view the image on a monitor and vary the delay of the reference beam 鈥� perhaps by operating a joystick 鈥� to inspect the patient鈥檚 body at different depths. In addition, by tuning the wavelength of the laser, it should be possible to detect different chemical features of the tissue being imaged. This might eventually help to distinguish between, say, a melanoma and harmless tissue.
One of the most promising aspects of this technique is that the whole apparatus can be made relatively compact and portable. At Imperial College, Roy Taylor and I have developed compact ultrafast lasers which require only a few watts of electrical power (see New 杏吧原创, 15 January 1994).
To image through thicker layers of tissue, it is possible to pick up useful information from the light that has been scattered only a few times on its way through the intervening tissue, and so arrives at the detection system before the rest of the scattered light. Because the path taken by this light resembles a wriggling snake it has been described as 鈥渟nake light鈥� by Robert Alfano of City College, New York (see diagram). Near-infrared light tends to be scattered forwards in biological tissue, so snake light can make it through samples ten times thicker than ballistic light.
Alfano鈥檚 group and others have used combinations of very high speed detectors and confocal imaging techniques to record images of objects in scattering media equivalent to a tissue thickness of several centimetres. The detectors need a fast response time so they can record the snake light and then shut off quickly before the rest of the scattered light arrives. This approach could eventually be used clinically, but at present the detectors are rather expensive, and the maximum tissue thickness through which snake light is detectable is still too thin for mammography.
A completely different approach will be needed for tissue that is more than a few centimetres thick. There is no point in trying to filter out the scattered light, as this is the only light that will make it through. Delpy鈥檚 group and others are developing techniques to measure the light transmitted through a scattering medium as comprehensively as possible, using multiple light sources and detectors, and making what they can of the information they glean. The random scattering that takes place in biological materials can be described by mathematical models, so computers can be used to determine how a particular distribution of scatterers and absorbers inside the medium could affect the incident light. The idea is to collect and measure the scattered light and work back to an image of the tissue that produced it. However, the computing and modelling required is very difficult, and success so far has been limited.
The first practical clinical applications of near-infrared imaging are likely to come only when a specific medical problem happens to fit in with the available technology, and the cost of the techniques that experimenters have tried remains a major factor. As compact, portable, versatile and cheap lasers become available, the chances of such fortuitous 鈥渇its鈥� will improve, and the use of light for imaging biological tissue will come closer.
The inside story
FINDING out about the body鈥檚 internal workings is one of the oldest priorities of medical research. Doctors and scientists began by dissecting corpses, but peering inside living patients is obviously not so easy to arrange. Realising that human tissue is not completely opaque to light, doctors in the Victorian age conducted the first experiments in mammography. Unfortunately, the only light source they had at their disposal was candles. But they ran into problems, not just with finding volunteers. The main problem was that so much of the light was scattered by the breast tissue that the images they were hoping for became hopelessly scrambled.
Since then, several different 鈥� and less painful 鈥� techniques have emerged for looking inside the body. X-ray imaging has been very successful in diagnosing many conditions 鈥� from bone fractures to tooth decay. It works so well because the X-ray radiation, with its very short wavelength, is not severely scattered in tissue.
Light scattering is caused by the random microscopic structure of biological tissue, and becomes more severe as the wavelength approaches the size of the elements that are scattering the light. For biological tissue, the scattering elements are of the order of nanometres to micrometres in size, so visible wavelengths, which are in the range 0.4 to 0.7 micrometres, are strongly scattered by less than a millimetre of tissue. But X-rays, with a wavelength of around 0.01 nanometres, can pass through several centimetres without scattering. Unfortunately, with short wavelengths come high-energies, so X-rays can ionise tissue and cause tumours if used too frequently or at high intensities.
One thing that X-rays cannot do is provide any information about the chemical nature of the tissue being imaged. This is because the relevant chemical information is associated mainly with the electrons in atoms and molecules. Electromagnetic radiation whose photons have the right energy to knock some of these electrons into a higher energy level can be used to probe the chemistry of these tissues. But X-rays cannot gather any of this information 鈥� their energy is much too high.
One way to get round the safety problem is to use ultrasound instead of X-rays, though this still does not reveal anything about the tissue鈥檚 chemistry. Ultrasound waves have relatively long wavelengths, of the order of millimetres. This means that they are not strongly scattered by biological tissue. Unfortunately, these long wavelengths also limit the resolution in any images to a few millimetres, which is not good enough for many applications. For screening for breast cancer to be useful, for example, the tumours must be detected before they are larger than about 1 millimetre across.
Perhaps the most powerful existing technique for looking inside the body is magnetic resonance imaging (MRI). It reveals the distribution of atoms of certain elements, and therefore the molecules that contain them, by detecting the magnetic signature of their nuclei as they react to a powerful magnetic field. Not only is MRI capable of sub-millimetre spatial resolution, it can also detect the presence of specific chemicals, and give information about functions such as blood flow and heart-valve pumping. But there are some important elements, including oxygen, that cannot be detected by MRI. And the superconducting magnets needed to generate intense magnetic fields make it expensive for routine diagnostic use.