In the final months of 2023, the Addenbrooke Hospital in Cambridge, UK, were reported as the first UK National Health Service (NHS) facility to deploy artificial intelligence (AI) during thermal ablation of liver tumours. Speaking to Interventional News, consultant radiologist Nadeem Shaida shares how the technology has enabled his team to bypass the “Achilles’ heel of liver tumour ablation” and treat tumours with better precision to ultimately reduce recurrence.
IN: The Addenbrooke Charitable Trust supporters donated the £200,000 thermal ablation machine to your hospital—how did this come about?
Liver ablation as a treatment option has grown in popularity following its incorporation into all major national and international guidelines for treatment of liver cancers such as cellular carcinomas, or hepatocellular carcinoma (HCC), but also for tumours that have spread to the liver from elsewhere, typically the colon.
We’ve been doing liver ablation here for many years—coming on 20 years in fact. Over this period, we found that demand for our services continued to rise, and we were being increasingly asked to do cases that we couldn’t do with our conventional means either via ultrasound or with a free-hand computed-tomography (CT) scan to guide the needle. This was due to poor image quality and/or obscuration of the tumour because of its location.
We had to find a solution because liver ablation works very well, and in looking for one we came across navigation. Essentially, navigation segments the tumour via markers on the skin which guides the needle. With this it became easy to do the cases we previously couldn’t such as high lesions near the dome of the liver.
IN: How has the addition of AI changed your practice?
It has helped us get around the Achilles’ heel of liver ablation. If you follow these patients up long enough, quite a number of them will recur and this is mainly because you haven’t ablated all of the tumour in the first sitting. To prevent this, you have to ablate the tumour, but also a safety margin around the tumour, which is where AI comes in.
AI has increased the accuracy of tumour segmentation, meaning patients are much more likely to be rid of the entire tumour in the first instance and have no chance of recurrence. Evidence for this is still emerging however, but there are promising data already which suggest AI-driven segmentation leads to better outcomes and lower recurrence rates.
IN: Are there any drawbacks to the use of AI technology to ablate liver tumours?
Yes. Quite often with AI segmentation it will tell you that you probably haven’t burned enough, and this can be quite a leap of faith for an operator, particularly if they’ve come from an ultrasound background where they put the needle in, burn, fire and forget. We need to see clinical data validating how accurate post-ablation measurements are when using AI and whether following these leads to improvements in long-term survival.
IN: In your opinion, what will the trajectory of AI look like within interventional radiology of the near future?
Ours is a technology-driven industry and we’re seeing a symbiosis between AI and systems we currently use through innovations in stereotactic navigation and robotics. So, I expect to see that develop and come to fruition—which we have already seen in some regards.
Expect to start seeing technology suggest to you, rather than you picking, the trajectory at the beginning of a case. AI won’t however eclipse the human element of healthcare and radiology, which is to understand the clinical processes and patient factors. Although when you look at HCC, with the variety of treatment options—embolization, radioembolization, chemoembolization, surgery, drugs, ablation—I do think in the future we may see patient selection become more driven by AI.
