Advanced technology in shoulder arthroplasty surgery: Artificial intelligence, extended reality, and robotics

Abstract

The purpose of this review is to provide an overview of the integration of technological advancements in orthopedic shoulder surgery. Recent technological advancements in orthopedic shoulder surgery include predictive analytics, computer-navigated instrumentation for operative planning, extended reality, and robotics. Separately, these advancements provide distinct methodological attempts to improve surgical experiences and outcomes. Together, these technologies can provide orthopedic surgeons with the tools and capabilities to improve patient care and communication in shoulder arthroplasty. From artificial intelligence-generated predictive analytics to extended reality and robotics, technical innovations may lead to improvements in patient education, surgical accuracy, interdisciplinary communication, and outcomes. A comprehensive narrative review was conducted to explore the technological advancements of orthopedic shoulder arthroplasty. Our findings emphasized the impact of these advancements, exemplified by early enhancements in efficacy and safety. However, certain challenges remain, such as a lack of reproducibly improved outcomes and cost considerations. While the reviewed studies indicate hope for improving shoulder arthroplasty, the true cost-effectiveness and applicability remains to be determined, indicating the need for further research.

Introduction

Reports that the large language model, ChatGPT, performing at or near the passing threshold for the United States Medical Licensing Exam sparked discussion regarding the role that artificial intelligence (AI) can play in medical education and clinical decision-making. ¹ In the evolving environment of orthopedics, technological advancements have the potential to redefine the gold standard of total and reverse shoulder arthroplasty (TSA/RTSA), offering unique approaches for preoperative planning, intraoperative execution, and postoperative outcomes. This article explores the integration of technological advancements in orthopedic shoulder surgery, including predictive analytics, computer-navigated instrumentation, extended reality, and robotics. Separately, these advancements provide distinct methods for improving surgical results. Combined, these technologies can provide orthopedic surgeons with the capabilities to improve patient care by enhancing surgical precision, efficiency, and accuracy.

Predictive analytics

AI is a term that encompasses computer systems that simulate human intelligence, with machine learning as a subset of the term. ² While AI does not yet have the capabilities to replace human surgical decision making, it can serve as a valuable tool to aid surgeons in predicting outcomes and crafting preoperative plans for shoulder arthroplasty. Predict+ is a clinical decision tool that uses AI to predict patient-specific outcomes following shoulder arthroplasty. ² Predict+ uses machine learning techniques to extend beyond traditional statistical methods, evolving from simple risk factor identification to a prediction model that optimizes clinical outcomes, allowing providers and patients to accurately establish outcome expectations. ³

The American Shoulder and Elbow Surgeon (ASES) score has been categorized as a standard method of prediction for potential patient outcomes. Machine learning provides opportunities to improve the ASES’ measurement capabilities. ⁴ Kumar et al. ⁵ used machine learning techniques to create predictive models for the Shoulder Arthroplasty Smart (SAS) score, ASES score, and Constant score to compare the accuracy of each algorithm. The three machine learning techniques more accurately predicted outcomes of TSA and RTSA using the SAS score, followed by the Constant score, and then finally the ASES score. While the ASES score remains the gold standard in measuring outcomes, the findings from Kumar et al.'s study suggests that machine learning can enhance current models by incorporating additional variables to provide more accurate predictions.

Using additional patient-specific information, machine learning can expand its predictive abilities beyond patient outcomes. Karnuta et al. ⁶ created an artificial neural network (ANN) model that “learns” through experience rather than statistical regression, and demonstrated accuracy for predicting costs, length of stay, and disposition following shoulder arthroplasty. This model further demonstrates the trajectory of machine learning by extending its application to predict additional statistics.

Predictive models can become decision support tools, with the potential to enhance the consent process by creating personalized patient predictions. Integral to the preoperative discussion are the expectations of both the patient and the surgeon regarding the outcomes of shoulder arthroplasty. Higher preoperative expectations have been linked to both improved and at times worsened postoperative results based on the type of expectation. ⁴ By using predictive analytics, patients will have well-defined and more realistic expectations regarding the outcomes of their procedures, with a higher likelihood for improved postoperative results and perception of the results.

Preoperative planning

There has been a recent shift from traditional two-dimensional (2D) imaging platforms to three-dimensional (3D) reconstruction models when formulating preoperative plans for shoulder arthroplasty. Notably, computerized 3D computed tomography (CT) has witnessed a growing adoption in preoperative planning. Raiss et al. ⁷ reported high concordance between planned and implanted glenoids for both RTSA and anatomic shoulder arthroplasty when using 3D planning programs. In patients undergoing TSA, complete concordance was 85%, while patients undergoing RTSA had 90% complete concordance, revealing that the 3D-based planning resulted in minimal deviation from the preoperative plan. ⁷ Parsons et al. discovered significant inter and intra-surgeon variability in glenoid augment sizing when 3D planning was not used, with considerable differences in what surgeons consider to be an optimal reconstruction of the glenoid. ⁸ The integration of computerized 3D CT has the potential to standardize procedures across surgeons, thereby enhancing procedural efficiency, and improving both quality and safety. Segmentation of CT scans is typically one of the most tedious processes, but Zhao et al. ⁹ demonstrated reproducible glenoid segmentation using a U-Net bone segmentation mask model with near-perfect accuracy and volumetric error less than 10%.

Intraoperative planning

Computer-navigated instrumentation has proven to be beneficial in the execution of preoperative plans. ¹⁰ In the context of shoulder arthroplasty, the adoption of computer navigation has notably enhanced the placement of the glenoid component. This is particularly significant, considering that malposition of the glenoid component is one of the primary causes of revision surgery following shoulder arthroplasty. ¹¹

Nashikkar et al. ¹² examined the role of computer navigation in improving the accuracy of glenoid base plate positioning and bony fixation, discovering that CT–based computer navigation displayed significant improvement in median screw purchase length for both anterior and posterior screws. In another study conducted by Nashikkar et al., ¹³ a computer-assisted surgery group noted a significant reduction in between-patient variability in the postoperative version. This resulted in a greater number of glenoid components in a “neutral” position for both inclination and version when compared to the manual technique. In a controlled cadaveric study, Jones et al. ¹⁴ demonstrated that using a navigation system was more accurate and precise in achieving the desired version and inclination of glenoid base plates than traditional instrumentation.

Although there have been observed improvements in glenoid component implantation, few studies have reported clinical patient outcomes. In a retrospective study, Holzgrefe et al. ¹⁰ reported that there was no significant difference between the range of motion and functional outcome scores when comparing computer-navigated and non-computer-navigated RTSAs. Although notching and revision surgeries were more common in the non-navigated procedures, it was not statistically significant. With limited reports of clinical patient outcomes, more research is needed to determine the true clinical benefit of using computer-based navigation.

Extended reality

Extended reality (XR) is a term that encompasses virtual reality, augmented reality, and mixed reality. ¹⁵ Virtual reality (VR) replicates scenarios by immersing a user in a 3D, computer-generated environment. While VR is primarily being utilized for patient education and training of surgical techniques, it has not yet been implemented intra-operatively, as it needs further development. ¹⁶ Augmented reality (AR) can superimpose preoperative CT scans and MRIs over the surgical field, enhancing surgeons’ perception of a patients’ anatomy for improved precision. The mixed reality (MR) environment includes digital overlays (like the AR environment) but allows for interaction with the projected holograms through verbal commands and hand gestures. ¹⁶

Head-mounted displays (HMD), such as the Microsoft HoloLens, allow the implementation of MR in the surgical space. Surgeons can overlay preoperative parameters on the surgical field and decide in real-time how to approach procedures. Using an HMD, Rojas et al. ¹⁷ showed low deviation between the planned and postoperative values of RTSA, as well as low deviation between intra- and postoperative values in terms of glenoid inclination, retroversion, entry point, depth, and rotation. In another study of MR surgical navigation using glenoid axis pins, pins were inserted within 2 mm from the planned entry point and within 2 degrees of version and inclination. ¹⁸ MR navigation systems can aid surgeons in accurately placing glenoid, and eventually humeral, components. While promising, the literature on extended reality remains premature, and requires further exploration before becoming the gold standard.

Robotics

Robotic-assisted technology systems are currently being used in hip and knee procedures, but there is a notable absence of this technology in shoulder arthroplasty. Although the current use of robotics in total shoulder arthroplasty is limited due to lack of commercialization, there is an increasing interest in robotic-assisted arthroplasty.

Orthopedic robotic-assisted platforms are subdivided into three separate categories. Active systems perform tasks independently (following initial programming by the surgeon), while passive systems perform under the direct control of the surgeon.^19,20 Semi-active systems follow a mixed approach, where the robot is controlled by the surgeon following an initial programming but has an automatic control to ensure that predetermined surgical goals are met. ¹⁹

The MAKO system (Stryker, Mahwah, New Jersey, USA) is a semi-active robotic system currently used for total hip and knee arthroplasties since 2015. The MAKO system has a firm monopoly in the robotics market, embodying over 70% of all robotic arthroplasty systems. ²¹ When compared to manual total hip arthroplasty, a retrospective cohort study found no statistically or clinically significant differences in post-operative scores, but the MAKO group had a shorter hospital stay. ²² Batailler et al. ²³ discovered that total knee arthroplasty using a MAKO system reduced post-operative pain and improved implant positioning with the improvement of functional outcomes when compared to a conventional method. ²³ A prospective randomized controlled trial evaluating the clinical and cost-effectiveness of the MAKO is underway, known as the RACER trial. ²⁴ The MAKO shoulder arthroplasty system is forthcoming. ²⁵ The ROSA system (Zimmer Biomet, Warsaw, Indiana, USA) represents another robotic platform that has been used for total hip and knee arthroplasty since 2019. However, there have been major concerns with regard to this platform's accuracy in the sagittal plane for total knee arthroplasty with 26% of patient alignments deviating from the sagittal plan by more than 3°. ²⁶ The ROSA system recently obtained U.S. Food and Drug Administration (FDA) 510(k) clearance for robotic-assisted shoulder replacement surgery in February of 2024. ²⁷ Some of the hoped-for benefits of the Rosa Shoulder system include the ability to measure bone and tissue tension while continuously monitoring them throughout the procedure to ensure improved dexterity and accuracy—decreasing the chance of surgical errors. ²⁸

New robotic-assisted surgical platforms for TSA have gained traction, aiming to improve precision and integration beyond commercial robots. Smith et al. ²⁹ created a compact, hand-portable device with the ability to consistently match or outperform a fellowship-trained surgeon; however, the true value likely lies in the less experienced orthopedic surgeons. Darwood et al. ³⁰ introduced an intraoperative robotics platform for guidewire placement, with average angular accuracies of 1.9° version, 1.2° inclination, and positional accuracy of 1.1 mm compared to preoperative plans. The system is considered low-cost when compared to conventional intraoperative robots, potentially reducing overall operative costs. As new technologies aim to address the limitations of current robotic systems, further research is essential to validate their efficacy in clinical settings.

Future directions

The potential of AI is promising, but limitations remain regarding integration in orthopedic surgery. Gupta et al. ² described how some of the greatest limitations of the clinical transition of AI are the lack of external validity and limited generalizability, as none of their 16 reviewed studies on AI-shoulder-based models externally validated their findings. Current studies are limited by a lack of external validity, hindering their generalizability, as they cannot analyze how the models perform in current healthcare climates. For future orthopedic AI studies, external validation is crucial to provide evidence of generalizability.

As we navigate the changing technological landscape in shoulder surgery, it is imperative to further explore the value of these technological advancements. In healthcare, value is defined as the ratio of outcomes relative to costs. ³¹ The introduction of new medical technology is a primary driving force of increased healthcare costs. ³² Moschetti et al. ³³ discovered that robot-assisted knee arthroplasty was only cost-effective in high-volume arthroplasty centers, with no cost-effective benefit in low and medium-volume centers, further exemplifying how technological improvements may not pose beneficial to all healthcare centers. Physicians and allied healthcare providers must weigh the clinical benefits against the burgeoning costs of introducing technological enhancements. The cost-effectiveness of introducing these technologies has yet to be determined, but with a deeper analysis of costs and clinical outcomes, we can soon discern whether there is true value in the introduction of advanced technology in the orthopedic surgical space.

Conclusion

Shoulder arthroplasty is reaching a new horizon of technological innovation. This wave of technological integration is being rapidly adopted by surgeons and industry alike, with tools such as customizable patient outcome calculators, streamlined preoperative planning procedures, and improved surgical accuracy and precision. Although early results have proven fruitful, there are still many areas that need additional exploration, such as clinical outcomes. Within the existing literature, studies use cadavers and 3D models to integrate technological advancements, but there have been limited reports of long-term patient outcomes in clinical settings. As the integration of technology in the orthopedic space keeps evolving, further research and clinical trials are necessary to ensure that the results discussed in this paper are generalizable to real-world patient outcomes. In addition, it is essential to determine the true cost-effective value that these advancements are bringing to the healthcare space. While we are optimistic about the potential of this new technology, it is evident that further research is necessary to determine its true benefits and address any remaining challenges.