Computers Shrink Orthopedic Surgeons' Roles

Laird Harrison

Disclosures

September 13, 2017

Computers already fly airplanes and direct trains. They can beat humans at the games of Go, chess, and Jeopardy! How long before they can make diagnoses and perform surgery?

The technical barriers are falling fast, experts say. Soon, it may only be a matter of preference that keeps human surgeons in charge. And orthopedics could among the first specialties to feel the change.

Orthopedic surgery lends itself to robotic surgery because the bones and soft tissue that orthopedic surgeons treat are relatively easy to image accurately.[1] Already, the specialty makes use of robots for multiple procedures.

Human surgeons have so far made all of the decisions about what treatments patients need and how to go about them, using robots only to execute these decisions in circumstances requiring special dexterity, precision, or stamina.

But this year, researchers in Sweden reported using artificial intelligence to identify wrist, hand, and ankle fractures in radiographs.[2] How long before a computer equipped with such information can control its own mechanical arm to, for example, locate, align, and fixate a broken bone? And how long after that before human surgeons become superfluous?

"I believe that's still quite a number of years off, and part of it is because the technology requires further refinement," says Alan H. Daniels, MD, an assistant professor of orthopedic surgery at Brown University. "But in addition, physicians are not eager to hand over control of clinical practice to machines."

One of the most impressive examples of robotic surgery comes from gastroenterology. Last year, researchers at Children's National Health System in Washington, DC, and Johns Hopkins University in Baltimore announced that an autonomous robot had sewn up tears in the intestines of living pigs. Its sutures were tighter and more evenly spaced than those of experienced surgeons performing the same procedure.

And already, some urologists are controlling robots without scrubbing into the operating room, Dr Daniels says.

Computer-Assisted Navigation Currently Used in Orthopedics

The potential for a similar innovation exists in orthopedics. Orthopedic surgeons have used robots since the 1990s, but recent advances are making the machines more useful. In their simplest forms, robots can act as guides to help surgeons position screws, prostheses, or tunnels.

For example, the Optotrak 3020 (NDI Measurement Sciences) uses infrared light to locate bones intraoperatively.[3] With the Acrobot (REO), on the other hand, surgeons control surgical tools held by a robotic arm that is constrained to a predefined volume.[4] Similarly, the Robodoc (Curexo Technology) mills the canal for a prosthesis using prepared surgical plans based on CT scans.

In hip replacement, computer assistance in placing the cup of the prosthesis has proved more accurate than conventional methods. Other robots have used CT to provide a real-time virtual model of the surgical field. Some use haptic technology to provide the surgeon controlling a robotic arm with a virtual sense of touch.[1]

In the knee, robots also assist with alignment of prostheses and with minimally invasive surgery. Robots are able to position tunnels in the femur or tibia more accurately than surgeons using conventional means, potentially reducing the risk for revision of anterior cruciate ligament repair.[1]

Computer-aided navigation is proving useful in shoulder arthroplasty or where normal anatomy is distorted, such as with fractures, revisions, and glenoid wear or dysplasia.[1]

In the spine, computer-assisted navigation helps surgeons avoid neurovascular structures, and it is well established in the placement of lumbar pedicle screws.[5] The incidence of misplaced screws is as high as 42% with traditional surgical techniques, according to some studies, but it is only 10% with computer assistance.

Robotic surgery does still have its downsides. Notably, it takes longer to complete the procedures, raising the cost and the risks from anesthesia. But in the spine, at least, the consensus is emerging that the improved accuracy merits the extra cost.[1]

Using Artificial Intelligence for Analyzing Radiographs

On the diagnosis side, little progress was made until a team of researchers at Karolinska Institutet, Danderyd Hospital, and the Royal Institute of Technology in Stockholm, Sweden, teamed up to develop an artificial intelligence system for reading radiographs.[2]

To teach computer networks to read radiographs, they fed them 256,458 hand, wrist, and ankle radiographs with labels stating whether or not the images contained fractures, the body part, laterality (left or right), exam view (anteroposterior/frontal, lateral, or oblique [two different types]), and four scaphoid-specific views (proximal, distal, ulnar, and radial). Fifty-six of the images contained fractures.

Rather than programming the computer networks to look for such features as lines that might indicate these characteristics, they instructed the computer to find whatever patterns were statistically correlated with the labels.

The researchers then created a gold-standard set of 400 images by reviewing them in full resolution, together with alternative views and the radiologist's report. For fractures, the best computer network had an accuracy of 83% in identifying fractures in these images. By comparison, the reviewers were accurate 82% of the time. On a subset of images, the computer networks reached an accuracy of 99% on body part, 95% on exam view, and 90% on laterality.

"We were excited that we got human-level performance when both are presented with the same radiograph," says Max Gordon, MD, PhD, a fellow at Karolinska Institutet.

More than 80% of orthopedic surgeons' decisions are made on the basis of radiographs, said Dr Gordon. The next step, he predicts, will be for artificial intelligence systems to use images to quantify risk, perhaps telling one patient that a knee replacement has an 80% chance of eliminating the pain and telling another that the procedure has a 50% chance. But patients will most likely want human experts to talk through these statistics with them.

"There will be possibly be a day when we will have robots examining patients and everything will be automated, but that's a future that is so far off," he says.

Will Surgeons Be Removed From the Cutting Table?

Patients have so far shown enthusiasm for surgery assisted by robots, dazzled by the high-tech sophistication, says Dr Daniels. But he thinks patients may balk at the idea of removing human surgeons from the equation.

Already, orthopedics has the reputation of a "high-tech, low-touch" profession that needs to work on bedside manner. "A large component of the physician/patient relationship, even for surgeons, is the communication and the relationship, and that absolutely would be lost if artificial intelligence starts to makes all the clinical decisions along with, or for, the patient and then also performs the surgery," he says.

The legal ramifications remain to be worked out as well, he points out. Even the best robot guided by the best artificial intelligence will make mistakes. Then who will be liable? A surgeon who is nominally supervising the procedure? The manufacturer of the robot?

Some of the biggest barriers may come from surgeons themselves, who still have the option in many settings to refuse technology that could replace them. The American Society of Anesthesiologists lobbied against an anesthesiology robot developed by Johnson & Johnson and approved by the US Food and Drug Administration, and few hospitals were willing to sign on. Finally, in March of last year, Johnson & Johnson quietly stopped selling the device.

Still, technological innovations have steadily overcome such objections in other fields. Predictions about self-driving cars seem to be crystallizing steadily, with cruise-control features becoming ever more sophisticated and self-parking features already available on high-end models.

The near-term future of surgery may more closely resemble that of flight, Dr Daniels says. "The airline pilot is there to assist in terms of landing and flying a plane, but the instruments are able to do a very large amount of the work."

Get ready for surgery on autopilot.

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as:

processing....