Scientists are constantly struggling to develop novel and innovative methods for cancer detection and diagnosis. Hyperspectral imaging sensors have the potential to aid this struggle by detecting the subtle differences in colour as healthy tissue turns cancerous.
Hyperspectral image sensors are able to very finely sample the colour spectrum of an object. In contrast to an ordinary camera, which collects three colour images – red, green and blue – hyperspectral image sensors collect hundreds of colour images. The high resolution colour data allows us detect colour changes that would not be visible with a standard camera. We can use this colour spectra for object identification.
The first hyperspectral imaging sensors were developed in the field of remote sensing to allow plant scientists and geologists to differentiate between different types of tree or rock. These original sensors were big and bulky and relied on the relative motion between the object being imaged and the sensor to extract colour information. Within the field of remote sensing, this has been solved by using a small aircraft to fly the sensor over the area being imaged. This is fine if one wants to look at a rock or a stone, but the need for an aircraft has hindered the applications of hyperspectral imaging methods in other areas of science.
However, with the development of more compact and cost effective sensors, hyperspectral imaging is seeing a range of new applications. For example, studies have shown its effective use in quality control of food, surveillance and biomedical imaging applications. We are amongst the scientists taking advantage of this new generation of hyperspectral imaging sensors for biomedical imaging.
Within the last few years there has been a surge of hyperspectral imaging sensors based on a range of different technologies appearing on the scientific market. We have chosen to evaluate a sensor that separates light into a hundred different colour components using filters. These filters consists of tiny mirror cages; the mirrors trap the incident light, and only allows light whose wavelength fits perfectly within the mirror walls to pass through to the sensor. Different colours (wavelengths) of light therefore strike the sensor at different spatial positions, allowing us to separate the colours and reconstruct a “technicolour” image.
We are currently investigating whether this type of sensor is sensitive enough to allow us to take a picture of cancer. Whereas we hope that we will eventually be able to capture the subtle colour differences between cancerous and healthy tissue, we are currently using fluorescent contrast agents to boost our detection sensitivity.
Contrast agents are molecules that search out and bind specifically to either healthy or cancerous cells; when we shine light on fluorescent contrast agent stained tissue, each type of cell will fluoresce, or emit light with a different colour. Since hyperspectral imaging allows us to very finely distinguish between different colours, we are able to simultaneously detect several fluorescent contrast agents. The ability to probe several types of cancer cells simultaneously will give us a more detailed image of the tissue and allow us to diagnose cancer with higher accuracy.
Hyperspectral imaging may therefore prove to be a very useful tool for cancer diagnosis: Because when it comes to cancer we really need the image to say more than a thousand words.