David Snead is Professor of Pathology Practice at the University of Warwick and founding director of PathLAKE, a world-leading centre for digital pathology.
Funded under Innovate UK’s Data to Early Diagnosis and Precision Medicine strand of the Government’s Industrial Strategy Challenge Fund, PathLAKE is a centre of excellence led by University Hospitals Coventry and Warwickshire NHS Trust in collaboration with researchers from the University of Warwick, Oxford University, Nottingham University and Queen’s University Belfast.
Professor Snead talks to Health Tech World about how AI and digital pathology are set to transform outcomes for patients with cancer and chronic illnesses.
“Over the past decade or so, we’ve been researching how AI can be used to analyse digitised images of pathology slides, referred to as whole-slide images,” Prof. Snead explains.
“Histology slides are what pathologists use to diagnose a patient’s disease. They enable pathologists to see what is going on at the cellular level.
“This is the key step in many patient pathways, for example from a biopsy (small sample of tissue) taken from a breast or lung or bowel mass, the pathologist can answer a number of questions.
“Is there cancer there or not? If there is, what sort is it and how do we treat it? How is it growing? Or if it is not cancer, what is the cause of the mass and what might happen to the patient?”
“The pathologist can tell all of this from looking at the sample under the microscope, or as in the case at Coventry, one of the world’s first digital pathology centres, looking at these slides as images on computer screens.
“We digitise the slides at high resolution, making images available for diagnosis, so we can do our work as pathologists more efficiently, but also through digitisation we can access that pixel data for research, mainly focused on machine learning using convoluted neural networks.”
This digitised information is helping to rapidly advance the field, with PathLAKE very much at the centre of the revolution.
The centre has digitised and is collecting data from 21 sites and now, the rest of the West Midlands is digitising too, offering further scope for collaboration.
“Meanwhile, the West Midlands Secure Data Environment being is set up, of which PathLAKE is an important part.”
PathLAKE is about to put into clinical practice two algorithms, one for breast cancer and the other for prostate cancer.
These algorithms help pathologists to find cancer in biopsy samples.
Some benefits of this research are clear, including shorter waiting times, faster diagnoses and fewer errors.
Others are less obvious, such as the ability to predict which mutations are present in a patient’s tumour and the way a particular tumour will respond to different therapeutic options.
“Some of the most valuable tools are the ones where we can not only find a cancer but that can also inform treatment.
“Once we can identify whether or not a patient needs adjuvant chemotherapy, we’ll have another tool in our armoury.
“Even better, if it can tell us which therapy to use, we’ll be able to treat patients more effectively.
“We’ll get there, but not for a few years yet.”
Cancer diagnostics is just the tip of the iceberg, with Prof. Snead seeing promise for everything from colitis to liver disease and kidney failure.
Anything causing chronic illness due to pathology in the tissue is within its scope.
AI: Overcoming misconceptions
The AI explosion has attracted as much skepticism and fear as it has excitement.
ChatGPT is the latest innovation to inspire countless comment pieces about cheating students and lost jobs.
In the healthcare space, the regulators are asking: ‘If you don’t know what a black box is doing, how can you regulate it?’
“Not knowing what’s happening inside that black box is disconcerting, but there are many things in medicine that we don’t fully understand, like consciousness and general anaesthesia, for example,” Prof. Snead says.
“So as long as you know what happens, do you really need to know exactly how it happens, before you start to use something?”
Cost is another adoption barrier, with one report putting the figure at between $20,000 (£16,000) and $1 million (£800,000) for a complete healthcare AI solution.
However, Prof. Snead argues that the outlay is worth the potential benefits.
“Yes, it’s expensive, but it’s important to invest in pathology,” he says.
“If you look at what we spend in medicine and healthcare generally, and look at what you actually spend on the diagnostic pathway in pathology, it’s really minimal in comparison.
“For us to diagnose a lung cancer would cost less than 0.1 per cent of a single month’s cancer treatment. So it’s a mistake to quibble about investing to make sure the diagnosis is right to start with.”
Prof. Snead believes that it will take more than financial investment for health innovations like digital pathology to reach their full potential.
He argues that a national framework should be put in place to help establish university spin-outs and negotiate the use of IP generated through biomedical research to ensure as many as possible become profitable.
This would also ensure that the technologies can start having an impact sooner.
Prof. Snead says:
“As a country, we need to be innovative in how we set these companies up. It’s really important that we get this right.”
Co-founded with long-time research collaborators, Prof. Nasir Rajpoot and Dr Simon Graham, Prof. Snead’s own spin-out is now gaining momentum.
Building on the team’s work at the University of Warwick, Histofy will provide groundbreaking interpretable AI to pathologists, biomedical researchers and the pharmaceutical industry.
While it’s been a long journey to get this far, Prof. Snead’s team of medical academics is excited about entering completely new territory.
Prof. Snead says:
“We’re now recruiting and starting to look at getting external investment in. It’s all exciting stuff moving ahead now.”