Just as radiology is set to be transformed by AI-powered medical image interpretation, surgery stands on the brink of a historic change as robotics becomes precise enough to automate some of the routine tasks in this high-risk field.

AIMed Magazine interviewed Tamás Haidegger (read his blog here), the technical lead of medical robotics research at the Antal Bejczy Center for Intelligent Robotics,

Tamás is an active member of the IEEE Robotics an Automation Society and is a national delegate to an ISO/IEC standardization committee focusing on the safety and performance of medical robots.

We asked him how surgery will be impacted by the fourth industrial revolution.

AI MED: What are the challenges for surgical AI?

Tamás Haidegger: Surgery is probably the most complex medical domain, featuring decision making on the spot, intense time pressure and the possibility of dire consequences. The execution has to be perfect, you have to have very delicate motions, you have to have everything under control.

As a human surgeon, you have to have a very good understanding of where you are, what you are doing, and really a tenth of a millimetre might be a deal maker or a deal breaker. All that combined, it is a great challenge for robotics.

AI MED: How does AI feature in surgery today?

TH: We are living in the age of robot assisted surgery. What that means is that robots are actively or passively helping the human surgeon to better execute their task.

robotic technology artificial intelligence virtual assistants robotic surgery artificial intelligence medicine ai healthcare doctors patients surgeons hospital data

This can happen in various forms. The most common platform is the Da Vinci surgical system from Intuitive Surgical that was approved by the FDA in 2000, and executes the exact motions of a surgeon. It is a calibrated system and decision-making stays with the surgeon. A human surgeon is directing the motions through a master manipulator and everything is executed at a smaller scale inside the patient – we are talking here minimally invasive surgery of course.

But there are many other forms of robotic surgery. A particularly interesting one is image-guided surgery. And that means that, especially in orthopedics or in neurosurgery, the robot is able to perform event the finest motion based on a pre-operative plan prescribed by the surgeon.

If you need a hip replacement surgery there are certain bone chopping procedures involved, and that can be done very accurately based on the CT plan given to the robot. There is a 3D mesh volume that the robot can understand, and once it’s registered or translated down to the human patient, the robot will execute that prescribed drilling, milling, and cutting very accurately.

There are certain records, especially in more targeted procedures like needle insertion and catheter-based procedures, where there are on an experimental level even more complex surgeries being done autonomously by robots. This means there is a limited decision-making capability built in the robot which ensures that, without the surgeon’s active intervention, the robot can still make certain basic decisions such as navigating through vessels and down the airways to the lung, taking biopsy autonomously or stopping for inspection if it sees something interesting.

Or even further we can think of radiosurgery, which is a little different than traditional surgery with a scalpel but happens fully autonomously with some systems which can get the exact amount of required dosage to a tumour and save all the rest of the organs. This is happening, and it is coming, but there is still a long way to go.

AI MED: What do you see as the most impressive examples of surgical AI?

TH: In the past two or three years, there have been three or four new surgical systems that have appeared on the market and went through successful human clinical trials focusing on retinal vessel surgery. 

robotic technology artificial intelligence virtual assistants robotic surgery artificial intelligence medicine ai healthcare doctors patients surgeons hospital data

In the human eye, we can now do such delicate surgeries: peeling the membrane, reconnecting arteries, or just putting needles and applying some local medication in case of vein occlusion – only one eye surgeon out of 100 has delicate enough fine motion control to do that.

With robotics we are getting able to really go beyond the physical human capabilities. There is no hand tremor or shake, no limitation in terms of fatigue, you don’t really need to be there for five hours you can do things remotely, so there are various aspects that heighten the current surgery.

Decision making in most of these systems still stays with surgeon, which means they can still cognitively decide what they are willing to do. But already there are numerous attempts to bring in cognitive surgery, or surgery 4.0 as it has been called recently, which means that during the procedure learning systems, AI algorithms, are continuously monitoring what is happening and already being able to provide you with some navigation. In the future we’ll also be getting decision support for the most critical parts of the procedures.

AI MED: What are new skills surgeons need to develop to use this tech?

TH: For now, I think most of these systems are able to perform at the level of an average surgeon, but once we are able to move beyond that it’s going to be critical for surgeons to keep relevant by better understanding what’s happening.

The takeaway for surgeons is: embrace technology! In the next two or three decades, leading surgeons will be those who are employing and exploiting technology capabilities to the full extent to really push their own limits further.

AI MED: Would you be happy for a robotic system to perform major surgery on you?

TH: I saw a survey recently, and it was surprisingly positive. People have changed their opinion in the past five years dramatically towards allowing AI algorithms or robots to operate on themselves or their beloved ones. We tend to trust technology more and more. This is good and bad. Good because if people are not willing to accept these forms of treatment, then why are we as engineers developing it? There is no point in that.

On the other hand, it’s kind of sad because many people do not understand the technology and so either over trust it or just reject it. I think that we really have a deep responsibility to prepare society, so they will have the right amount of knowledge to understand where to use technology and where to stick with the traditional methods.

AI MED: What are common concerns surgeons have about robotics?

TH: The number one concern is job security, in terms of are they getting replaced – and the answer is no they’re not getting replaced any time soon. Although, limited functions, repetitive and boring tasks might get overtaken by automation. But it’s not any better or worse than applying staplers instead of manual sutures, because it’s faster, quicker, saves time and energy for everyone.

But it’s also true that there is another significant concern. Traditionally we trust the surgeon to make decisions, and if they make a bad decision we have certain mechanisms we can conduct at a public level to make sure that we as a society accept or deny that they made the right decision. And if they are guilty then they get punished or we withdraw their license. However, when an AI takes a decision we don’t agree with, then who do we blame?

If an algorithm says it is statistically a better decision to remove a limb and then it turns out it was wrong in removing that limb and making someone an amputee for the rest of their life, and that was a programming error that someone made years before at the manufacturer, who do we blame?

A human surgeon, sure he or she may make errors – that’s why we have disclaimers and that’s why you have to give your consent to everything. But people in general expect technology and robotics to be perfect. And even if you told them that statistically a robot is better than 99.5% of all human surgeons in the world, if it fails they will still blame the robot manufacturer even though statistically it was a better decision to use the technology. This is a really big issue and we don’t have a good answer for it. Society itself is not ready for it.

This article first appeared in AIMed Magazine issue 05, a Deep Dive on Robotic Technology & Virtual Assistants, which you can access here.