Hyperspectral Imaging

Credit: Dale Waterhouse

Our eyes function like a camera, capturing light and sending data for processing and storage.

But as incredible as it is, the eye has its limits. It’s sensitive to three colours – red, green and blue. While this is enough to distinguish everyday objects, it is not enough to differentiate a cancerous cell from a healthy one.

Cameras used by scientists in surgery are limited in the same way.

Now, some of our researchers have developed specialised cameras that can be used to detect cancer cells that can be hard for both surgical cameras and the naked eye to spot.

We spoke to Dr Dale Waterhouse who led this exciting research, recently published in Cancer Research, about how this new technology could illuminate the future of the early detection of oesophageal cancer.

What’s the problem?

Oesophageal cancer occurs when abnormal cells in the food pipe, the tube that carries food from your mouth to your stomach, grow in an uncontrolled way.

Around 9,100 people are diagnosed with oesophageal cancer each year in the UK. And unfortunately, many people are diagnosed with later stage cancer, when the disease tends to be more advanced.

But there can be early warning signs, in the form of a condition known as Barrett’s oesophagus. Most people with Barrett’s oesophagus won’t develop cancer, but the cellular changes seen in Barrett’s oesophagus can sometimes develop into oesophageal cancer, giving doctors the opportunity to spot abnormal cells early.

The standard diagnostic tool to detect Barrett’s oesophagus, an endoscopy, involves a long, flexible tube with a tiny camera and light at the end, which passes down the oesophagus.

People with Barrett’s oesophagus will then have follow-up endoscopies every 3 to 5 years, to look for signs of the condition changing.

“An endoscopy looks around the oesophagus using standard white light imaging, which is the same kind of imaging you get in a phone or digital camera. The endoscopist will look for some suspicious signs that might indicate whether Barrett’s oesophagus is developing towards cancer,” says Waterhouse.

Waterhouse explains the challenges that arise using this kind of imaging. “It’s actually quite difficult to see those changes with the standard white light imaging we use, because as you can probably imagine, everything in the oesophagus looks a kind of pinkish-red colour.”

So, the team at the University of Cambridge decided to investigate whether it was possible to increase the sensitivity of the standard endoscopy camera, and by doing so increase the accuracy of detecting when the condition is progressing towards cancer.

A spectral ‘fingerprint’ of cancer

A spectral fingerprint of cancer

Spectral fingerprint

When light is shone onto tissue, like the inside of your oesophagus, it is absorbed by the different biological components it’s made from. Each of these absorbs different colours and proportions of light. When light leaves the tissue again, it has a distinctive colour composition, which Waterhouse describes as a ‘fingerprint’, which can tell us about the underlying biology.

This unique fingerprint is also known as a spectrum. Just like a fingerprint, the spectrum can be used to identify the tissue that created it, and the spectral fingerprint of cancer is likely to be different to that of healthy tissue.

The special cameras Waterhouse and his team designed were able to reveal this valuable additional information that a typical camera would leave out, using something called hyperspectral imaging.

What is hyperspectral imaging?

Hyperspectral imaging is a technique that analyses a wide and finely sampled spectrum of light instead of just the 3 primary colours (red, green, blue).

“You are able to capture much narrower ‘bands of light’, each collecting a much narrower range of wavelengths, and you take more of those. So rather than 3, you could collect 10s, or even 100s of different colours.”

And detecting more colours means that scientists are able to pick up a more detailed fingerprint.

Hyperspectral imaging isn’t new. This unique kind of imaging has many applications, from observing space through the latest advanced telescopes, to monitoring emissions produced by coal and oil power plants.

“The problem with hyperspectral imaging is we don’t know which of the colours we detect will actually be useful, and no one’s really collected this data in the oesophagus,” explains Waterhouse.

The team set about collecting some of that missing information from the oesophagus. “We needed that data to assess whether there are any spectral differences between the pre-cancerous tissue and the tissue with Barrett’s oesophagus. This will help us determine which of those colours are most diagnostically useful.”

Who, what and where?

The pilot study recruited 20 patients with Barrett’s oesophagus, who would be having a surveillance endoscopy as part of their standard of care. From these 20 people, the team captured 715 spectral images.

“The patients enrolled in the study came in for a standard procedure, and the endoscopist marked two regions. One which they were very sure was Barrett’s oesophagus tissue, and then another region, which they were suspicious about, because they thought it might be early cancer,” says Waterhouse.

“And then we inserted our device through the working channel of his endoscope. The endoscopist then directed that towards these two regions, and we captured spectral information from the tissue.”

Following that, and this is the really important part, the team took biopsies of those same two regions, so that they could get a diagnosis based on gold-standard histopathology.

“And that’s important, because actually, in three of the cases, the initial diagnosis from the endoscopist, was incorrect, which is the challenge we’re trying to solve,” explains Waterhouse. “To be specific, one region the endoscopist marked as healthy was actually pre-cancer, and 2 cases marked as suspicious were actually healthy. This is why matching our spectral data with gold-standard diagnosis from histopathology is so important.”

Significantly, the team found that there are some distinct spectral differences between tissue that was Barrett’s oesophagus, and the tissue that was developing into cancer.

“We did some modelling to find that there are differences in the blood distribution in the tissue between the different disease states, which the camera is able to pick up.”

What’s more, using the results from the spectral imaging, they developed an algorithm to try to tell the two apart. The algorithm had 84.8% accuracy in distinguishing between Barrett’s oesophagus tissue, and tissue that was precancerous, which Waterhouse says “is good compared to the current standard of care and promising for the future development of this tool”.

Looking beyond oesophageal cancer

While Waterhouse and the team were going technicolour, another group was working on new ways to detect oesophageal cancer earlier, using a ‘sponge on a string’. So how does this new research fit in?

“These two technologies, as far as I see them, are very much complementary,” says Waterhouse. “You can screen the general population with Cytosponge, because it’s such a cheap and easy technique that can be done in a GP surgery. And then once you’ve found patients with Barrett’s oesophagus, that’s when they get put into the surveillance pathway.”

But Waterhouse explains how this new imaging technology has huge potential beyond the diagnosis of oesophageal cancer.

“This kind of technology would be useful anywhere that imaging is used,” he says. “It could be used in the lower gastrointestinal tract to detect precancerous lesions, or polyps, in the bowel.”

Beyond detection, these cameras could also be utilised in surgery. The idea would be that the cameras could detect, real-time, whether there was any residual cancer during the operation.

Up next for the team, a number of researchers are looking to build a second, improved prototype of the camera, “one with higher resolution and potentially one that detects more colours”. Waterhouse himself has turned his attention to glioma – a type of brain tumour – and using the same camera to assist in open brain surgery at the National Hospital of Neurology and Neurosurgery and the Wellcome / EPSRC Centre for Interventional and Surgical Sciences.

While it’s still early days, there’s huge scope for this exciting technology. The sky’s the limit for Waterhouse and the teams at the University of Cambridge and UCL.

More on this topic