Would you trust a computer to plan your next hospital surgery?
We’re entering a bold new future, where every cut and stitch may be precisely planned by artificial intelligence (AI), long before you go under the knife.
That might sound far-fetched, but today I’m speaking to someone who’s making it happen. My interview subject today is Fangde Liu. He’s a researcher working on surgical robots, and leading the development of big data systems for pharmacovigilance of Imperial College – one of the UK’s top universities. His startup, Surgical AI, is aiming for the prize of democratised, automated surgery. It’s a bold and exciting vision – and I’ll let Fangde tell you all about it.
AL: Can you explain the need for your startup, Surgical AI?
FL: My background is in robotics. Before working on surgical AI, I worked for a few years on computer graphics. I built virtual humans or characters, with each bone or muscle as accurate as possible. In the beginning, I thought it was very exciting to put my knowledge in robotics and graphics in surgical robots. For me, that will be more meaningful.
I’ve been working for many years on surgical robots doing surgery in an innovate manner. However, little has been done to improve the accessibility of surgery. When I visit hospitals, I realised the many challenges involved, just in performing daily procedures. Many procedures are high risk, with a success rate of less than 30%. Surprisingly, most doctors will still conduct these by experience, without any help from technology.
AL: I’m guessing this is a global problem…
FL: The Lancet reports five billion people have no access to safe and affordable surgery. This could prevent one third of deaths, and remove an economic burden of 12 trillion dollars globally. Many advanced surgical technologies have been developed to treat previously untreatable diseases. These include 3D printing, laser surgery and surgical robots – even radiotherapy tools, like CyberKnife. However, little has been done to improve the accessibility of surgery. This is a big concern for healthcare equality.
AL: What exactly are you doing?
FL: Our London based startup Surgical AI is starting to attack the problem, using the power of AI and cloud computing. We are building an autonomous system to do image-based preoperative surgery planning.
Such a system can tell surgeon the best surgery plan within ten seconds – and do all the necessary measurement for the surgeons. Surgical AI is team of world-leading AI experts, plus surgical robot scientist and professor Anna Shetty (whom I regard as the best surgeon in UK). We are trying to build an autonomous surgery planning system in the cloud. Doctors can upload the medical image scans, such CT or MRI. Then our system will instantly provide the optimal surgery plan, and tell the surgeon which medical device they should use for this patient.
For orthopaedics, the system will also design 3D-printed surgical devices specifically designed for each procedure (patient-specific instruments, PSI for short). These are customised surgical devices, custom designed for each procedure. This will make surgery more efficient and accurate. This ensures the surgeon can cut the bone exactly as pre-planned on the computer, making the procedure even easier and safer. The Daily Mail has even called this technology “idiot proof”. AI and 3D printing may sound very high-tech and futuristic – but the Surgical AI team really believes that we are building a technology for to democratise surgery. As long as there is internet and a 3D printer, doctors can perform surgery with high accuracy and efficiency.
AL: Can you tell me how you envisage this new future will pan out?
FL: The Lancet provides an analysis of the access barriers for surgery. These are mainly time, capacity, safety and cost. When I visit local hospitals in China, surgeons are extremely busy. They have usually performed several procedures each day, and they have just a few minutes for surgery preparation. Existing surgery robots can improve surgery, but they cost too much money, and require too much preparation time for each surgery. That’s why surgeons are not using them for common cases – instead, deploying them only for research showcasing purposes. Surgeons have told me they want to apply this kind of technology to routine surgery. However, everything must be done in a few minutes – if we want to use it in local hospital environment, rather than for research purposes. Powered by cloud computing, the Surgical AI system can provide surgeons with the best surgery plan in seconds.
AL: AI’s readiness is often overclaimed. What are your key technical challenges?
FL: There is a lot of hype about AI and deep learning. People believe that these technologies can solve all problems. Yet, for surgery, the advantages of present-day AI are not immediately obvious. Our key challenges are the limited amounts of training data available; and regulatory issues, concerning human-machine interaction. All systems will sometimes fail, and a surgical AI should know when the task is too challenging, and invite humans to take over.
AL: Tell me more about how your AI works.
FL: We have re-engineered the imaging analysis pipeline. We first turn image analysis tasks into a series of decision-making steps. Also, our system is more like AlphaGo, rather than the dominant convolutional neural network model for computer vision. Another advantage of our reinforcement architecture is our system keeps on monitor system performance. When the AI finds the task is too challenging, our system will ask human for help. In healthcare, this is an advantage for it makes our system safer.
AL: What are the practical benefits?
FL: We have tested our system clinically. In one case it saved the hospitals 40 minutes inside the operating theatre, which equates to about £1,000. The system will reduce time, and our technology will enable junior surgeons to perform surgery that’s currently the preserve of top surgeons. This approach therefore solves the capacity problem.
Our technology is also safe. Various pieces of research show that 3D-printed PSIs increase the success rate and improve the procedure quality. Unlike robotic surgery, our technology only costs a few hundred dollars per patient. That’s affordable for mass-market use. In fact, hospitals find our technology actually saves money – because the procedure time inside the operating theatre is reduced. These time savings translate into thousands of dollars saved – by reducing staff time, and having better capital equipment utilisation in the operating theatre. AI has the potential to break all the access barriers of surgery, making surgery available for all. We call this surgery democracy.
AL: Tell me about your ultimate vision.
FL: Our ambition is to build something for surgery that becomes as ubiquitous as Uber. As long as surgeons and doctors are connected to the internet, local surgeons will get support from both AI, and the world’s top surgeons. On our platform, it is possible to let top surgeons to verify the surgery plan and PSI design remotely, while local surgeons do the cut. This ensures that every operation is performed as well as it can possibly be.
AL: What will the employment impacts be?
FL: AI and robots are coming – and they’re making a lot of people worried about their job. Further, many people start worrying about equality and democracy. However, it is really exciting to know AI will help healthcare democracy.
AL: Lastly, what’s your current progress, and fundraising situation?
FL: We are currently looking for seed funding of about $1 million. We’re trying to build a full-time team, and making our system production ready.
Would you let AI operate on you? Please let us know your views in the comments below.
Category: Artificial Intelligence