AI, Robotics, & Additive Manufacturing Transform Healthcare | Manufacturing Industries

Artificial intelligence, data analytics, robots and additive manufacturing are transforming the healthcare industry in exciting ways such as robot-assisted surgery, rare-disease diagnosis, image interpretation and customized joint replacements. Further adoptions in seemingly mundane areas such as medical billing offer additional benefits.

“Cardiology and orthopedic practices and the associated life science companies that support them have made great progress in using AI, robotics and additive manufacturing,” said Jim Thompson, senior director of strategy for Siemens’ medical device and pharmaceutical industries. “Research and development tends to optimize itself around the most prevalent diseases involving the most money.

“Artificial intelligence is just a tool to analyze data and create correlations to make predictions based on machine-learning algorithms. But it happens to be a very useful tool for healthcare.”

Cardiovascular Diagnostics Level Up

In cardiovascular medicine, one of the first things identified in AI diagnostics was predicting blood potassium levels based on EKG wave forms, said Mark Wehde, chair of engineering at the Mayo Clinic.

“It was a novel concept no one had tried,” Wehde said. “Cardiologists knew blood potassium levels would impact how the heart functioned; if someone had a really low blood potassium level, you would see changes in the electrical activity of the heart. They wondered if there was a correlation. They fed information from patients into an algorithm training set and then looked at how well EKG waves predicted blood potassium levels—and it was really well.”

The correlation matters because monitoring EKG levels on patients outside the hospital is relatively routine compared to having the patient come in to a hospital or physician’s office for a blood draw to measure potassium levels, Wehde said. “It’s hugely impactful.”

Speed and cloud storage have been the game changers for AI, Wehde said.

“AI is one of these things that has been around for a long time,” Wehde said. “Even as a young engineer, we were talking about neural nets. The promise was always there, but it never really panned out. Only in the last few years has the technology caught up to the promise. It’s a matter of having processing power and microcontrollers that can run AI algorithms efficiently and in a reasonable amount of time. There has been an explosion in cloud storage making is possible to store massive amounts of data inexpensively. Combine that with incredible powerful computational engines and we now have the necessary pieces to run increasingly sophisticated AI/ML algorithms.”

“The Mayo Clinic is putting a big effort into AI,” Wehde said. “We’re partnering with engineers from Google to get all of our data in the cloud to be able to run algorithms on it to find ways to improve diagnostics and therapeutics.”

For example, AI now can be used to analyze diagnostic imaging from X-rays, CT scans and MRIs more quickly, accurately and consistently than humans, Thompson said. Such analysis is extremely labor-intensive and might take a highly trained nurse, engineer or technician from two hours to all day, Thompson said.

The old way was to take the data, create a virtual 3D representation on a computer, manipulate that image using a computer mouse, and then move through that image layer by layer, slice by slice looking for issues, Wehde said.

A modern, AI-informed system does the laborious scanning to identify images with issues that are not usual and flag for more attention from a radiologist, he said.

AI does a good job with pattern recognition in radiology, Wehde said. “That was one of the first big applications in healthcare, helping radiologists, reducing the burden of work on them. Now we are able to quickly process thousands of MRI and CT images.”

Such a system is “capable of spotting things a trained radiologist will miss,” Wehde said. “It increases quality, reduces the amount of time involved. It’s a real win for everyone.”

“An AI program that is properly trained and optimized can do that analysis in less than a minute,” Thompson said. “It’s a huge productivity gain.”

An AI-informed analysis is both more accurate and more consistent. “Because it’s a computer program looking at the results, it’s as accurate as it possibly can be compared to humans who get tired or bored of looking at pixels,” Thompson said. “That doesn’t mean AI is perfectly accurate because the pictures aren’t perfectly accurate.” As for consistency, three different people might get three slightly different results while AI provides much more consistent results, he said.

AI can help with deciding when is an optimal time to perform surgery, said Ed Cuoco, vice president of strategy and solutions at PTC. Consider a patient who will need a knee replacement and considering when to get the joint replaced “AI can augment the doctor’s decision making by showing more possible outcomes,” Cuoco said.

One limitation of AI is that its algorithms depend on the data provided to apply machine learning, he said. “If you present AI with a case radically different from what has been seen before, the result could be highly inaccurate because the algorithm hasn’t been trained for that combination of data and attributes,” Thompson warned. “The algorithm is only as good as the machine learning data that has been pre-processed.”

Cost-effective Data Analytics

At the Mayo Clinic, clinicians are loading data on a variety of common and uncommon diseases and conditions from around the world to help close data gaps. For example, Mayo Clinic physicians last year treated eight instances of a rare brain tumor, Wehde said.

“We see the very sickest patients from all over the world,” Wehde said. “We see things other providers don’t. Physicians at a regional or community hospital are not seeing these conditions. AI is going to close the gap, provide those answers that aren’t necessarily front of mind for you.”

“Even with the internet, it’s hard for a doctor in Boston to know the latest going on with a hospital in Los Angeles, let alone a hospital in Tokyo,” Cuoco said. “There’s not sudden new sources of data but it’s much easier to bring a lot of sources of data together.”

Data analytics can help uncover patterns of behavior, patterns of physical degradation, and unexpected drug interactions, Cuoco said.

While smaller, rural hospitals may not be able to invest in $100 million machines, those hospitals could gain important benefits by investing much less in good data analytics, Cuoco said.

“Data and analytics are relatively cheap compared to machines. The bang for the buck of good data analytics that facilitate good diagnostics and care is more cost effective for a rural hospital,” Cuoco said. “Not everyone can live near a top-tier hospital. Data analytics can’t replace that, but it can help narrow the gap.”

Surgeon/Robot Partnerships

Robots are promising in assisting and performing surgery, especially in orthopedics, cardiac care and minor surgeries, Thompson said.

A surgeon partnering with a robot, for example, allows the surgeon to guide the robot while looking at a zoomed-in image of the patient’s anatomy, with more precise vision than the surgeon working alone, Thompson said.

Robots make more feasible to perform endoscopic surgeries through small ports, as small as one inch in diameter, in the patient, Wehde said.

“Robots are really good at doing the same thing over and over,” Wehde said. “Robotic systems are able to move in ways that a human hand can’t. They can rotate in a circle, 360 degrees.”

Surgeons can guide robots carefully and precisely to target tumors deep into a patient’s brain, Wehde said.

Robots also can extend a surgeon’s career, Wehde said. “There comes a point where a surgeon’s manual dexterity isn’t there; the hands aren’t steady enough,” he said. “The computer and robotic system can compensate for that. As long as the mind is still sharp, a surgeon can still do surgery.”

Based on predicted shortages of physicians—the American Association of Medical Colleges projects a shortage of 46,000 to 90,000 doctors by 2025—extending the working life of surgeons is significant, Wehde said.

Augmented Reality, 3D Printing Aid Training

Augmented reality also plays a role, in connecting experts to remote hospitals and battlefields and in training, Cuoco said.

“AR is allowing physicians and even medical students to have access to realistic training, not just in surgery but in drug processing and sample processing, rather than more expensive training involving actual humans,” Cuoco said.

An emerging trend is using augmented reality to project a patient’s diagnosis and medications onto the patient sitting in the physician’s office or the patient on the screen in a telemedicine appointment, Cuoco said.

Additive manufacturing is delivering value in training physicians because instead of using a cadaver, the same portion of anatomy can be 3D printed over and over for physician training, Thompson said. In unusual cases, a surgeon can use additive manufacturing at the hospital to show how a disease is manifesting itself in a patient’s body to give better results than with two-dimensional imaging, he said.

Orthopedic surgeons and orthopedic medical device companies are leading the way in leveraging additive manufacturing for personalized surgical instruments and, in some cases, the actual implants including knee, hip, ankle and shoulder replacements, Thompson said.

As for prosthetics, additive manufacturing is enabling the company Unlimited Tomorrow to create lightweight, artificial arms that are the mirror image of the opposing limb much more affordably than alternatives, the company said.

“This is especially important for children because they are growing so fast,” Thompson said.

Faster, More Accurate Billing with AI

While the above examples grab more headlines, AI is providing big benefits in medical billing in catching errors and waste, Cuoco said. Hypothetically, if an insurance provider knew that a test coded 123 often accompanied another related test coded ABC, then a test coded ABD instead would alert an AI system that a human should check for a potential error, Cuoco said.

“Medical code billing in insurance is a big manual effort, and extremely prone to error” Cuoco said. “AI can look for unusual patterns and red-flag codes. Improved accuracy up front can save potentially million so dollars in insurance waste. It’s an underappreciated piece of the medical complex where AI has profound impact. It’s just not as exciting as robots doing surgery or additive joint design.

But there’s a shortage of medical billing coders. The possibility for AI to make medical billing happen more efficiently and deal with that shortage is as profound as more cutting edge directly applicable instances of AI.”

Overall, regulatory approval remains a challenge, as does training medical providers on the new technology, Thompson said.

“The biggest challenge is in evolving healthcare practices,” Thompson said. “Having a technology that works in the lab and can be cleared by the FDA or other regulatory agencies around the world is a big step forward. Then everyone in the healthcare system and in individual practices has to learn about the technologies, adopt them, and start using them. How do you train them, educate them? The adoption is slowed down by delivery practices and older personnel.”

In the past decade, more digitally native people have entered healthcare and are starting to ask, “Why are we not doing this digitally?” Thompson said.

Affordability is emerging as another challenge, especially in developing countries where healthcare systems, such as those in Africa, don’t have deep pockets to pay for emerging technology, he said.

Looking ahead, new devices will help with a variety of neurological problems, Thompson said. “More and more investments are going to neurology,” Thompson said. “That’s the next frontier.”