Artificially Intelligent Design for Orthopedic Devices

Increasingly, orthopedic device manufacturers are gearing their prototyping and design efforts toward a single goal—time to market. Time to market is a critical factor for getting new products into the marketplace and beating out the competition. As a result, design and development are not just about how new or advanced a function or technology is—they are also about using powerful software tools such as generative design, finite element analysis (FEA), virtual prototyping, and digital twins to shorten production schedules, error-proof designs, and speed up time to market. The abundance of creative design opportunities these software tools provide is exciting but also challenging for orthopedic device design and prototyping. An increasing number of manufacturers are embracing digital tools such as 3D anatomical printing, image-based patient-specific planning, and virtual reality/augmented reality; however, they are also aware these technologies must operate within established processes that still have limitations, such as scalability.

“For example, one of the biggest bottlenecks for any patient-specific workflow remains segmentation and landmarking, which limit the scalability—and therefore the uptake—within the orthopedic industry,” said Kerim Genc, product manager for Synopsys, a Mountain View, Calif.-based provider of software and services for clinical 3D image processing and model generation.

MDMs that became comfortable with digital technologies during COVID-19 now often prefer to connect with their contract manufacturers (CMs) for a digital review, rather than building a physical prototype right away. The added benefit of building a digital prototype first is revisions can be made and exchanged in real time, shortening the development phase and ensuring the first physical prototype will be very close to what the design team was hoping for. Rapid prototyping is also supported by the steady shift toward additive manufacturing (AM) over the last few years. “AM has become a large part of our prototyping process and allows for quick production and speedy responses from our clients,” said Zachary Bradley, marketing director for BoneModels, a Kingsford Heights, Ind.-based medical model manufacturer specializing in surgical bone models.

Built for Speed

Speed, of course, is in high demand, as are those companies that have invested wisely in technology, innovation, and best practices. “MDMs want partners that can contribute to and review the design for manufacturing [DFM] process and utilize the tools and capabilities they need to deliver their products quickly,” said David Cabral, president/CEO of Five Star Companies, a New Bedford, Mass.-based medical device contract manufacturer of orthopedic and spine devices and instrumentation.

DFM is becoming more popular within the orthopedic development space for streamlining product design and prototyping, with the goal of creating devices faster, with fewer steps and components, and at lower cost—without sacrificing accuracy, quality, and regulatory compliance. AM is fast-becoming an essential DFM tool that gives device manufacturers the ability to design and build prototypes and devices that cannot not be machined or built otherwise in a cost-effective manner—a capability that greatly expands engineering and design options, as well as speed to market.

For patient-specific devices, there are plenty of AM-enabled opportunities for improving the design and prototyping process to help patients receive the best custom fits that will reduce risks like revision surgeries. “Machine learning-based artificial intelligence [AI], for example, is proving to be the best way to speed up the image segmentation process and make these workflows scalable enough for production-level output,” said Genc. “However, as orthopedic OEMs explore new technologies like AI, they must also build the appropriate in-house expertise and partner with the right companies to fill their knowledge gaps.”

“OEMs are also turning to industry experts for rapid access to designs, to which they then add their personal touches and brand products to stand out in the operating room,” added Isidro Landra, leader of Intech Medical’s Prototype Garage in Kenosha, Wis., which houses one of the prototyping centers for this manufacturer of complex orthopedic medical devices. “We are also seeing increased demand for 3D-printed instruments that allow for iterative design improvements in a fast-paced environment.”

Manufacturers are eager to build workflows that are repeatable, reliable, and lower cost, while also leveraging the knowledge and experience of their manufacturing teams as effectively as possible. This can be achieved through the deployment of Industry 4.0/IoT technologies, especially the automation of manual work. Automation makes it easier to scale up medical device production to meet increasing demands from surgeons and other healthcare professionals. This also relieves some of the time demands on their engineers and eases the often-challenging logistics associated with getting a patient-specific surgical design plan from the concept stage to the operating room.

“Many OEMs face the build-or-buy dilemma when it comes to automating and scaling up their patient-specific processes,” said Genc. “They can either invest in the expertise and software infrastructure needed to build and train their own AI solutions, or they can go to vendors that already have the necessary expertise and can deploy their solutions quickly and efficiently.”

Software Breakthroughs

Much of the innovation happening in the orthopedic design space is driven by computer and programming advances, including IoT technologies that rely on sensor deployment, machine learning, and data analytics. AI-based solutions in particular are becoming more prevalent in prototyping and product design; for example, machine learning-based AI algorithms are used to automate large parts of image segmentation and landmarking for customized implants and other devices, making these tasks go much faster.

“Manual methods slow down product development for OEMs, due to the time and expertise needed to complete routine tasks,” said Genc. “We are finding that AI can get engineers where they need to be with a 3D model much faster than doing it manually. A task that previously took hours now can only take minutes, so it is a major shift in terms of workflow efficiency. The almost complete elimination of segmentation/landmarking time enables both scaling up without the additional head count and captures the expertise within the AI algorithm.”

These same benefits can be applied in areas such as 3D anatomical printing and in-silico clinical trials, where speeding up the segmentation workflow is important for maintaining an efficient turnaround for high-volume model generation. The COVID-19 pandemic brought about a keen interest regarding in-silico trials, where design engineers use advanced simulation and modeling technologies to create virtual patients that have the same physiological traits as human patients—so essentially creating human digital twins. These can vary in complexity, depending on the medical application and level of testing required. As a first step, the human digital twin incorporates data captured from the monitoring of real-life patients, such as basic vital signs. Other biochemical, circulatory, anatomical, and physiological data can be added at any time, including data from laboratory tests and diagnostic imaging studies. The behavior of individual organs can also be studied as virtual models, within or outside the “body” of the digital twin. For example, robust testing with a virtual heart model can lead to the development of better pacemakers.

In-silico trials have already been proven to be as effective as in-person clinical trials. For example, a recent study showed that in-silico trials replicated the results of three in-person clinical trials that evaluated the effectiveness of a flow diverter, a medical device used to treat brain aneurisms. “Our findings demonstrate that in-silico trials of endovascular medical devices can 1) replicate findings of conventional clinical trials and 2) perform virtual experiments and sub-group analyses that are difficult or impossible in conventional trials to discover new insights on treatment failure—for example, in the presence of side-branches or hypertension,” wrote the researchers in Nature Communications.¹

Ultimately, in-silico testing may become so effective that it replaces the need for clinical trials in the future, thereby reducing the cost and time it takes to run clinical trials and eliminating the need for human and animal testing.

As devices become increasingly complex, with more parts and added functionality, there is greater need for DFM to identify the best and most cost-effective manufacturing process. Using DFM during the prototyping stage will identify and troubleshoot any design flaws before the production process gets locked in. Finite element analysis is a crucial DFM tool for predicting how a proposed medical device will respond to real-world forces such as vibration, load, heat, fluid flow, electrostatics, and other physical effects. FEA also reveals a product’s most probable points of failure, which can then be fixed through redesign if needed before prototyping.

“FEA should be considered as part of the design process,” said Landra. “It ensures that the product will perform as intended in the operation room.”

Finite element analysis applications are expanding—for example, branching out from R&D and product design to in-silico clinical trials, which complement benchtop testing and speed up the clinical trial process, allowing for faster iteration and faster time to market. “This area is also being recognized by FDA and other regulatory bodies and is very flexible in terms of design and cost,” said Genc. “We see this as the future for orthopedic design and think that, eventually, clinical trials will not be completed without an in-silico simulation component.”

Other powerful software tools for orthopedic device designers and engineers include Solidworks, Visualize, nTopology, and Amphyon for 3D printing for analysis and simulation tools. Many software programs can now tackle AM and its complex geometries, functionalities, and performance designs, which are becoming more important to orthopedic device manufacturers.

However, traditional computer-aided design (CAD) solutions cannot always account for the levels of complexity that OEMs want when they create these custom designs, especially in the growing market of  software for patient-specific data.

“Software can design, but machines cannot always meet the criteria shown on a computer screen,” said Cabral. “This is where DFM and expert engineering assistance become especially important.”

“Software can only get you so far,” agreed Landra. “Close proximity and involvement with machinists and programmers are essential for an effective DFM program.” With regulatory
requirements becoming more stringent, she added it is essential to prototype devices on production-equivalent machines, which helps streamline manufacturing and regulatory approvals.

Fewer Parts, Simpler Assembly

As product designs continue to become more complex, with advanced materials, miniaturized components, and challenging geometries and assemblies, there is more interest by MDMs in combining multiple parts into a single part. This not only speeds up production and simplifies assembly, it reduces the number of potential failure points, thereby improving product quality and longevity. AM is often the enabling technology for creating a single combined part (often with a unique geometry) to replace multiple parts, which is a faster and less costly process than machining the individual parts and trying to fasten or weld them together and make them fit in limited or challenging spaces.

According to the American Society of Engineers (ASME), “design for additive manufacturing [DfAM], which utilizes powerful generative software and other analytical tools such as FEA, is usually the go-to technology for assembly consolidation, often creating a single final part that can only be manufactured with AM.”²
ASME lists four main steps for part consolidation:

1. Select the assembly for consolidation.
2. Use generative design and FEA to test hundreds of possible consolidation designs to find the best solution for the assembly.
3. Use the data from Step 2 to redesign the parts and possibly the assembly to accommodate the geometry of the new single part (or several, depending on the complexity of the parts and their location in the assembly) and any other modifications needed to improve functionality and performance.
4. Retest and validate the new design and parts using FEA to be sure their specifications match the original requirements of the design. Then 3D-print the consolidated design as a prototype or production-ready component.

“When done in the correct circumstances,” noted ASME, “combining parts makes for a better design and saves time and money.”

Virtual Prototyping

Virtual prototyping (VP) is a design process that relies on powerful software to model a system in remarkable detail, which is then tested by adjusting a wide variety of variables that replicate real-world operating conditions. The design is then refined by a rapid iterative process to a near-perfect state, where it will only require one or a few solid prototypes before being approved for production.

VP is increasingly used as a substitute for rapid prototyping. VP draws on technical elements from digital twins, AI, FEA, generative design, and other powerful software programs, allowing designers to create and test highly detailed and complicated virtual designs very quickly. In fact, it is possible to create a virtual prototype that is accurate enough that it can actually be taken to production, without the need for a physical prototype for medical device engineers to hold, examine, test, and revise.

This, of course, depends on the complexity of the product. “Generally speaking, the more simplistic the product design is, with a small number of components, the easier it will be to produce from a virtual prototype,” said Jackson Hedden, president of Jackson Hedden Inc., a Birmingham, Ala.-based industrial design firm that often works with medical device companies. “In contrast, a product with multiple components—in the range of dozens or hundreds—would be incredibly risky to produce without making a physical prototype first.”

The next advancements in VP will most likely be AI-driven, moving toward improving the ability to articulate an object or product in virtual space, thereby greatly reducing the number of prototypes needed to take it to production. “Some companies create hundreds of prototypes before making the product a reality,” said Hedden. “We have found that virtual prototyping can reduce those high numbers to a total of one to three prototypes on average. VP allows designers to make changes on the fly and truly visualize an item in its entirety, so that any shortcomings are noticed almost immediately.”

Moving Forward

There are still many micro- and macro-scale challenges that hold back “full-on” innovation in the orthopedic market—for example, the testing and validation of advanced materials or the scalability of AM technologies. Regulatory agencies are still finding their way regarding how to regulate new technologies such as 3D pre-surgical planning and the potential of in-silico clinical trials, especially in the context of their obvious benefits to surgeons and patients alike. In the right hands, AI has the best potential to disrupt the industry, but it is still mysterious to many, especially the incredible number of potential applications.

The use of AI to automate the segmentation process continues to show big benefits in time savings. For example, Synopsys has a long-term relationship with Corin Group, which plans patient-specific total hip arthroplasties. “Corin uses our Simpleware software to segment and add landmarks to image data prior to designing surgical guides printing implant templates and generating a surgical plan,” said Genc. “We have developed solutions using AI-based methods to automate a large part of Corin’s segmentation and landmarking, to the point where their image processing time has been reduced by 94% per case.”

Bradley believes the main slowdown to innovation is the initial cost of necessary equipment—especially now that AM and 3D printing have become an essential part of the development process. “Industrial-grade machines come at a high price tag and until newer, better systems start to roll out, I do not expect to see the entry price into this level of machinery drop,” said Bradley. “However, I feel this drop is coming. In the last year alone, we’ve added 15 new printers to our arsenal and plan to continue that growth as new technologies are announced and last year’s model drops in pricing.”

Many times, MDMs will come to their CMs with an idea they have been told is too difficult to manufacture. “From our perspective, the old saying of ‘if there is a will, there is a way’ is still very true for the development process,” said Bradley. “We have come to realize that a lot of companies will lead MDMs to believe that what they are asking for is impossible, due to the fact that the project is not within their capabilities. We try our best to detail all the possibilities, especially the potential failures of a request, to our clients so that they understand better that it’s not that it is impossible, it just may fall into a very costly R&D process for anyone willing to take on the task.”

Ah, yes—the potential for failure.

Speed is everything in the world of orthopedic design—and not just for positive outcomes or time to market. Achieving failure equally fast can also be considered an advantage because valuable knowledge has been gained and a minimum amount of time has been lost.

Failing fast is as critical to innovation as achievement. However, it is also important to fail fast safely, which is perhaps the biggest benefit of simulation, virtual prototyping, and digital twins.
Simulation and testing technologies allow design teams to test hundreds or thousands of ideas, many of which will fail—but without harming human or animal subjects in long-term and costly trials.

“Getting it wrong isn’t an option when you’re testing on humans,”  stated SK-Pharma (in referring to pharmaceuticals at the time, but the point holds true). “However, if all your testing is carried out in a simulation environment, you can get it wrong as many times as you like.”³ 

References

1. go.nature.com/3J49JlD
2. bit.ly/odt230332
3. bit.ly/odt230333

---

Mark Crawford is a full-time freelance business and marketing/communications writer based in Corrales, N.M. His clients range from startups to global manufacturing leaders. He has written for MPO and ODT magazines for more than 15 years and is the author of five books.