The imaging tool that could prevent cancer surgery complications

Cancer surgery is a tricky balancing act. Cut out the main mass of a brain tumour, and there will likely be cancer cells intermingling with healthy brain tissue around the margins. Leave these behind and the cancer is more likely to return; remove healthy tissue and risk the patient losing brain function.

Surgeons face other kinds of battles, too. On removing a bowel tumour, for example, the next step is to rejoin the bowel’s two ends. If done incorrectly, the contents of the intestine can leak out into the abdominal cavity, which can be fatal for the patient. 

On both counts, having a better understanding of tissue during live surgery could benefit surgeons. King’s College London spinout Hypervision Surgical has developed an imaging system that can see past the pinkish hues of our insides, spotting tissue properties that could help surgeons undertaking cancer operations.

The hidden spectrum of blood

Humans can perceive just a fraction of the sensory information surrounding us. We can’t see the ultraviolet markings on flower petals that insects can see, nor can we detect the infrared signatures that rattlesnakes rely on to catch prey. According to evolution, other sensory information was more important.

Unsurprisingly, the ability to reliably distinguish tumours from benign internal tissue is also beyond us (especially since as far as we know, humans have only been doing cancer surgeries for a few thousand years).

“The evolution human eyes went through is to understand and navigate the natural environment, not to understand whether tissue is cancerous or not,” explains Michael Ebner, a medical imaging scientist and CEO and founder of Hypervision Surgical.

Whether fruit, forest or fjord, our eyes evolved to interpret our surroundings through three types of light-sensitive cells in our retinas. These are specialised to capture three bands of the visual spectrum corresponding to long, medium and shorter wavelengths of light. The sensors inside a standard digital camera are based on this model, with each pixel capturing a value for red, green, and blue (slightly differing to the ‘colours’ of our retinal cells).

Hyperspectral imaging takes this familiar visual mode of capture and explodes it into a world of rich detail that’s beyond the abilities of our retinas. Rather than splitting the visual spectrum into three bands, hyperspectral imaging captures up to hundreds of bands – including part of the infrared spectrum invisible to the human eye. Each pixel comprises massively more data than the traditional three bands we’re used to.

Buried within this mass of data are signatures that reveal hidden differences between regions of tissue. These signatures reveal differences in perfusion – the amount of blood supply in tissue – which can be higher or lower in tumours than in healthy tissue. Importantly, perfusion can help detect subtle changes in organ blood supply during surgery, potentially preventing life-threatening complications.

Assisted driving for surgeons

Until now, hyperspectral imaging systems have been too large and bulky to use in theatre, taking several seconds to minutes to capture a single image. Hypervision’s system captures high-frame-rate video information, making it much more usable for surgeons. The team has developed an AI model which is learning signatures that could indicate abnormal amounts of perfusion.

As a tool for surgeons, Ebner compares the technology to assisted driving systems. “There’s a constant feed of safety information to help drivers be better drivers. In the context of surgery, this does not exist,” he says. Most of the time, the surgeon has to decide what to cut and what to spare based on what they can see with the naked eye. “This can lead to a high – and we argue preventable – variability in patient outcomes.”

In the basement of St Thomas’ hospital in London, Ebner demonstrates how it would work with a keyhole surgery system and a model abdomen. A camera inserted into the abdomen shows what would be inside the internal cavity on a large monitor. (For demonstration purposes, it is footage captured from a real surgery.) “In colour information, it’s always very difficult to understand whether tissue is healthy or not,” explains Ebner. Indeed, to the naked and untrained eye, the pinkish-red, meaty-looking insides on the screen all look much the same.

He taps a button on the floor with his foot and the screen transforms to an image of much higher contrast. Low perfusion areas are dark purple, he explains, while high perfusion areas are orange. As the technology develops, it has the potential to enable surgeons to distinguish between healthy and unhealthy tissue.

Validation with bowel cancer and brain surgery

For its initial clinical validation studies, Hypervision is focusing on bowel cancer and how hyperspectral imaging can help to prevent the dangerous leaks that can happen after the bowel ends are rejoined.

Ebner explains that surgeons can reduce the risk of this by ensuring the two ends are well connected, with a sufficient blood supply so that the tissue will heal well. “Right now, surgeons do not have the ability to reliably gauge whether this join was performed well, whether there is enough blood supply for the tissue to heal,” he says. With hyperspectral imaging, surgeons could better understand the perfusion level of tissues in the join. As a result, Ebner says, complication rates in these types of surgery could be reduced.

Along with its work on bowel cancer surgery, the team is also gathering data to assess the potential for neurosurgery, in partnership with King’s College Hospital and King’s College London. “Surgeons need to be conservative when resecting tissue to preserve patient function. This can lead to tumour left behind,” explains Ebner. “If surgeon had a much better understanding on what's healthy and what's not… this could be game changing in terms of precision.”

Surgeries to remove brain tumours and aneurysms are among the procedures that could benefit from hyperspectral imaging. But first, the team and their partners must gather further hyperspectral data during surgery, with the aim of correlating signatures with healthy and unhealthy tissue. Hypervision’s collaborators are collecting data from 81 patients including five types of tumour – all of which should help inform how hyperspectral imaging can be used in future.

Developing medical technologies for patient impact is highly challenging, says Ebner. “It means technology, product development, clinical delivery in a regulated environment – and [running] a commercially viable business.” He explains that while King’s could offer unrivalled technological and clinical support, his Enterprise Fellowship at the Royal Academy of Engineering was vital to understand “what it means to build a commercial entity”, from corporate governance, to pitching to investors, and the legal requirements of building a startup.

The support didn’t end there, as in 2022, Ebner embarked on the Academy’s EXPLORE programme, a week-long visit to life sciences hotspot Massachusetts in the US, which he says gave the company vital learnings for commercialisation in the country. Now, with an eye on scaling the company, he’s a member of the Shott Scale Up programme.

“Every stage of the company puts completely new demands on the leadership team,” he says. “The Academy provided a pathway towards making headway and getting to the right people who have done it before, and so generously gave back and helped the next generation of entrepreneurs to follow suit.”