Senior Research Engineer, Medical
Surgeons would need to make 50% fewer changes to AI-based pre-operative plans compared to current ones. This is according to a research project we conducted here at Materialise with Dr. Raf De Vloo, an orthopedic surgeon at AZ Klina in Belgium, in which we applied AI-based planning to 193 cases. This technology learns an individual doctor’s preferences for surgical approaches and, based on those, provides higher-quality pre-operative plans.
We are all becoming more and more accustomed to AI, with voice recognition software on our phones, Netflix recommendations, and email spam filters. But what about AI in healthcare? Though the industry has been slower to adopt the new technology, the first products have hit the market in the last few years, and today almost every company involved in healthcare has AI on its research radar. AI has been used in products that range from finding out if a patient has suffered from a stroke based on a CT scan, to discovering irregular heart rhythm detection, and even detecting a disease called retinopathy that affects the eyes of diabetes patients.
During the annual meeting of the American Academy for Orthopedic Surgeons (AAOS) earlier this year, AI was a major topic at almost every pre-operative planning discussion. Overall, there was a consensus that this technology has great potential to improve pre-operative planning while reducing planning time and decreasing associated costs. At the conference, we were very pleased to present, for the first time, the results of our AI research for total knee arthroplasties (TKA) planning which provides strong evidence and was validated by peers.
AI applied to patient-specific solutions
AI works by learning how to solve problems without being explicitly programmed how to do so. We saw this could be useful for planning total knee arthroplasties (TKA), in which sections of the knee joint are replaced by metal implants to help patients suffering from osteoarthritis.
Since 2006, Materialise has provided patient-specific planning, implants and guides that help surgeons perform surgeries. We are able to create solutions that are unique to each patient with the help of our software platform called SurgiCase, in which a doctor can upload a patient’s medical imaging data and work with a Materialise clinical engineer.
To improve this software solution even further, our research team developed a robust AI algorithm that can automatically detect a surgeon’s pre-operative preferences based on their past cases. For this project we collaborated with Dr. De Vloo, a surgeon with more than seven years of experience with patient-specific solutions developed by Materialise. This research applied our AI-based models to 193 past cases from Dr. De Vloo the results were remarkable: we were able to reduce the number of corrections the surgeon had to make to his pre-operative plans by half. Since then we have applied this experiment modeling for multiple other cases from different surgeons, each time with at least a 50% reduction in alterations that the surgeons have to make.
Why use AI for TKAs?
There are many steps involved before having a final surgical plan in place. First the patient’s knee joint is scanned using computer tomography (CT) or magnetic resonance imaging (MRI) during an examination. Next these scans are securely transferred to Materialise, where they are used to create a virtual 3D model of the patient’s joint. Materialise then proposes a patient-specific plan that specifies how the surgeon could place the metal implant components. The pre-operative plan is sent to the surgeon for review, and if necessary they can modify it and send it back to Materialise. Once the surgeon gives their approval, patient-specific guides are 3D printed. These guides allow the surgeon to place the implant components as planned based on the patient’s anatomy prior to surgery.
The pre-operative planning process is quite complex, and due to different schools of thought, different surgeons have different approaches for how to place the two metal implant components. This is because the surgeon has 12 degrees of freedom for translation and rotation of the implant components, and there are many different strategies to determine the degrees of freedom.
The majority of surgeons will need to make manual changes in order to adapt the pre-operative plan according to their preference. This led us to look into AI, to see if this could help us create pre-operative plans which are patient- and surgeon-specific. In that way we can propose patient-specific solutions that take into account the typical surgical strategy of the surgeon performing the procedure. By doing this we will enhance efficiency of the whole process of creating patient-specific solutions and surgeons would not have to spend a considerable amount of time altering pre-operative plans to their surgical preferences.
In the future we will be working to extend the abilities of AI-based planning and validate this application in a clinical setting. TKA was the first research application for our AI-based planning; however, the methodology is independent of the end application. The goal of our further research is to be able to apply it to pre-operative planning for other surgical procedures. Keep tuned for our next AI-based planning announcements.
Do you want to learn more about implementing patient-specific knee guides in your clinical practice?