AI-Driven Computer Vision in Orthopedic Surgery: Opportunities and Benefits

There’s a rule of thumb in the tech community that goes something like this: If a trained human can see it or measure it manually, then a trained artificial intelligence (AI) system—with enough data—can perform the task faster and probably more accurately. Thus, AI increasingly is infiltrating healthcare in various ways—predictive analysis, medical imaging, and diagnostic algorithms. Its ability to automatically and quickly flag important but possibly missed data and bring them to practitioners’ attention is unparalleled.

AI is the advanced art of teaching computers to solve physical human problems. Thanks to data now being made available in real-time from operating rooms, this problem-solving tool has been introduced to orthopedic surgery. Within the context of orthopedics, an important application is computer vision. In today’s cars, driver assistance can make a person feel safer behind the wheel. Likewise, AI-driven computer vision provides orthopedic surgeons with powerful visualization tools and objective measurements, enabling them to perform procedures better than they could through human senses alone. With computer vision, orthopedic surgeons have found their driver assistance.

How Computer Vision Can Help Surgeons

Computer vision uses algorithms to process images from video, photographs, and other mediums and draws insights from them in ways human beings cannot. Surgeons have always been tasked with making accurate inferences about patient anatomy from limited information. The accuracy of these inferences can be significantly improved with reconstruction technology that trains AI to generate 3D models from two-dimensional images. In short, the objective here is to bring a fuller, more detailed view of a particular type of anatomy—the knee or hip, for example—from limited or partial information. Keep in mind that surgeons are looking for reconstruction and segmentation to properly delineate specific sections of ligaments, tendons, or joints, their landmarks, and anatomical relationships. Imagine having access to that information from the 2D representation given by an X-ray or fluoroscopy. That kind of real-time reconstruction is possible with computer vision.

How are these AI-based imaging advancements impacting treatment? In addition to the accessibility they provide, they also increase procedural efficiency. Enhanced imaging can help surgeons create precise, personalized interventions that can only be accomplished with 3D information. Physicians get better tools to help them perform more accurate and quicker procedures while reducing the level of human error. Furthermore, computer vision can help save time by automating many of the routine tasks involved in determining measurements.

Since computer vision is a flexible solution, it can be integrated with many devices. Innovations have evolved to incorporate computer vision with smaller handheld systems in the past couple of years. This can, of course, give practitioners more dexterity in successfully reaching confined or smaller spaces—a task that is much more difficult for large robotic systems. A navigation system that can deliver that level of precision without any robotic component gives surgeons more options.

Preoperative and Intraoperative Intervention

Where exactly can AI-driven computer vision play a role in procedures? There are two key phases: the preoperative and intraoperative stages. In the first phase, procedure planning can be automated to not only plan a procedure, but also determine whether one is needed. For example, precise measurements can be taken of artery length and diameter to diagnose conditions such as stenosis. Being able to observe a narrowing artery could foster treatment such as an angioplasty before the condition becomes serious. Once a procedure is set, more advanced planning would be needed, such as navigation planning, stent selection, and stent positioning. Such information would be gleaned from a 2D representation, or from a computed tomography (CT) scan to obtain 3D-level information about the coronary arteries. When that level of knowledge is brought to bear before an operation begins, the chance of error—such as incorrect stent selection—is diminished. That, in turn, decreases the need for additional interventions.

More important than preoperative planning, however, is the ability to assess in real-time a procedure’s expected impact on patient functionality. That's where intraoperative assessment and navigation comes in.

During an operation, AI can use the wealth of images from a live camera to help provide real-time analysis. The objective would be to help the surgeon visualize exactly what is taking place—i.e., where to place the hardware or drill a canal. Software can also identify a procedure’s stages, giving surgeons guidance on the kinds of tools to use next and identifying the next steps. During hip replacement surgery, for example, a 3D module can help doctors assess the true location and positioning of an implant and bones by matching a preoperative CT scan to intraoperative fluoroscopy.

Aside from determining positioning, algorithms also can help analyze surgical video. Most surgical video feeds are lengthy, but AI can help extract only the relevant parts. That kind of analysis would be extremely helpful to experienced surgeons who train their younger counterparts, as irrelevant portions could be skipped. Software not only can pick out the unconnected parts, it also can program a camera to record just the relevant parts.

Analytics

Another use for computer vision software is predictive analysis, which can determine a surgery’s predicted end time. This tool can help with hospitals’ macro planning process, enabling them to accommodate the maximum number of services as is efficiently possible.

Another opportunity that AI vision presents is the access to accumulated knowledge it provides the orthopedic community. As previously mentioned, AI models are trained, and sometimes that training occurs through the work of thousands of practitioners. The resulting massive amounts of data collected are based on information gathered from tens of thousands of cases. This presents the surgeon with objective information about potential treatments and/or procedures.

Keep in mind the wealth of data being accumulated can be used for both imaging- and non-imaging-related purposes. Data can help surgeons determine the best types of plans, treatments, and procedural practices.

Benefits: Accessibility and Cost

Clearly, physicians are not the only beneficiaries of computer vision. Patients also gain from the technology in two ways: accessibility and cost.

Numerous procedures require invasive cardiac measurements to help physicians better plan. Many patients prefer to limit the number of procedures performed on them. When 2D imaging can be converted into 3D information, the need for invasive procedures is reduced. Another concern is limiting exposure to radiation—AI technology can significantly reduce the radiation doses that accompany a full CT scan.

Radiology Business reports there has been a massive downturn in reimbursement for magnetic resonance imaging and CT scans, which has hampered the businesses of many physicians. Certainly, there are very high barriers to reimbursing these types of procedures. But giving physicians more options—including those of 3D imaging—reduces the cost of care for those undergoing these scans and bolsters a practitioner's patient flow/volume.

Conclusion

The healthcare industry has historically resisted the introduction of AI for understandable reasons—among them, the reliability of data. But data credibility is being resolved in many areas, and in the field of orthopedic surgery, it is a non-issue. The data is obtained directly from visuals and live feeds as it occurs. This provides physicians with tools for better planning, more precise measurement, better outcomes for patients, and a raised quality of care standard for everyone. 

---

Moshe Safran has more than 12 years of research and business development experience in computer vision algorithm development. Before leading RSIP’s business development for the United States as CEO, he was Vice President of R&D creating new ways for the company to solve complex technological challenges through AI. The company works with its partners to power their products and services by developing AI and computer vision modules. To learn more, please visit www.rsipvision.com.