To scan or not to scan? The image debate in robotic joint arthroplasty

­

Robotic technology has been shown to improve precision in component positioning in joint arthroplasty procedures.^1-3 Although the long-term benefits and cost-effectiveness are still being investigated,⁴ the use of robotic techniques in joint arthroplasty is on the rise.⁵ Kim et al⁶ reported that technology-assisted total hip arthroplasty (THA) has doubled over the past decade in the USA alone. Moreover, recent projections indicate that by 2030, robotic arm-assisted total knee arthroplasty (RO TKA) is anticipated to make up over 70% of all TKAs performed in the USA.⁷ The surge in robotic arthroplasty adoption stems not only from evidence of improved accuracy in component placement and reduced variability in surgical outcomes, but also from a rapidly growing, patient-driven appetite for technological innovation. Brinkman et al⁸ captured this phenomenon, demonstrating an exponential rise in public interest in robotic arthroplasty through analysis of online search trends over a ten-year period from 2011 to 2021. Amid this robotic momentum, a fundamental debate persists: image-based or imageless? Both approaches promise precision, yet differ in workflow, cost, and radiation burden, with the evidence distinguishing them being far from settled.

Image-based robotic systems utilize preoperative CT imaging to generate a patient-specific 3D bone model upon which the surgical plan is constructed, prior to entering the operating theatre.⁹ Intraoperative registration of bony landmarks is used to align the patient’s anatomy with the preoperative CT-derived 3D model. The principal argument in favor of image-based systems lies in the fidelity between preoperative planning and intraoperative execution. Image-based robotic arthroplasty was developed with the purpose of enhancing accuracy in determining component sizes preoperatively and aid surgeons in minimizing errors. In THA, this technique has been shown to significantly improve the accuracy in restoring horizontal and vertical centre of hip rotation compared to conventional THA, with fewer outliers in acetabular component positioning.^1,10 In TKA, CT-based preoperative planning has demonstrated accurate femoral and tibial component size prediction within one size in 98% of cases, reducing implant inventory and enhancing theatre efficiency.¹¹ Furthermore, image-based techniques have allowed for another layer of precision through haptic guidance, which has been associated with reduced iatrogenic soft-tissue and bony trauma and earlier functional recovery compared to conventional TKA.^12,13

A particular advantage of image-based systems is their utility in patients with complex or deformed anatomy. Hayashi et al¹⁴ demonstrated that an image-based robotic system was able to achieve comparable cup positioning in dysplastic versus non-dysplastic patients, with mean errors of 1.7° for inclination and 2.5° for anteversion. Overall, 11 of the 30 patients in their dysplastic cohort had severely dysplastic hips (Crowe III/IV), suggesting that preoperative CT-based planning confers particular advantages in cases where distorted anatomy would otherwise compromise the reliability of intraoperative landmark registration. This is further supported in TKA, where a prospective study of 155 robotic-assisted cases demonstrated hip-knee-ankle (HKA) axis restoration to within 0.76° of the intraoperative plan, highlighting that CT-based robotic planning achieves reliable deformity correction even in the context of significant coronal malalignment.¹⁵ The benefits of image-based robotic arthroplasty appear to extend beyond radiological parameters. In a study of 393 robotic TKAs with preoperative CT-based planning, Marchand et al¹⁶ reported that implant sizes were within one size of the planned sizes in 100% of cases and exactly matched in 80% of cases, including cases with retained hardware. Moreover, Pietrzak et al¹⁷ showed that CT-based preoperative imaging was significantly more accurate at predicting final implant sizes compared to 2D templating in TKA (p < 0.001). This finding has also been echoed in unicompartmental knee arthroplasty (UKA), with an image-based robotic system achieving more accurate component sizing compared to both imageless robotics and conventional techniques.¹⁸ These findings suggest that image-based robotic planning may enhance operative efficiency by streamlining instrument and implant preparation, thereby reducing the number of trays opened and contributing to overall carbon savings.^19,20 A further advantage, particularly relevant in complex or prolonged cases and in patients with osteoporotic bone, is intraoperative resilience to tracker array loosening or displacement. In image-based systems, the surgical plan exists independently of intraoperative registration and the patient’s anatomy can be re-matched to the preoperative CT model at any point. Imageless systems afford no such fallback, as the plan is generated entirely from intraoperative landmark acquisition and cannot be reliably re-established if the reference frame is lost mid-case.

Despite this encouraging body of evidence, image-based systems are not without limitation. The requirement for preoperative CT scans introduces additional costs, logistical burdens with potential time delays to surgery, and radiation exposure for patients. These considerations have fueled interest in imageless alternatives. Imageless systems require no preoperative imaging. The surgeon registers bony landmarks intraoperatively using a handheld probe – on the tibial plateau, femoral condyles, and malleoli – and the hip centre of rotation is captured through circumduction. The software uses these data to morph a statistical shape model from a pre-existing anatomical database to the patient’s specific bone geometry, upon which the surgical plan is constructed and adjusted based on intraoperative gap assessment and soft-tissue balance. There is no preoperative plan; everything is generated in theatre. The most immediate advantage is that imageless technology removes the need for preoperative imaging, which decreases preoperative healthcare resource use, reduces the level of radiation patients are exposed to, and simplifies the overall surgical preparation process. This streamlined workflow has been proposed as a key facilitator of wider robotic adoption beyond high-volume institutions.²¹

Studies have shown that the absence of preoperative imaging does not appear to compromise accuracy. Bosco et al²² found that imageless robotic TKA achieved accurate functional alignment and gap balancing across a range of knee deformities, with a mean alignment error of 1.18° and a 95% implant alignment accuracy rate. Mancino et al²³ compared imageless robotic TKA with navigated TKA in 172 patients and found that robotic TKA with imageless technology achieved significantly greater accuracy in achieving planned implant positioning. Tran et al²⁴ also found that imageless RO TKA achieved more accurate implant alignment and significantly better early functional recovery than navigated TKA, findings that are encouraging but confined to short-term follow-up. While the evidence base for imageless TKA is growing, data in THA remain limited. The application of imageless robotic technology to THA is an emerging area, and robust evidence evaluating its accuracy and clinical outcomes against both conventional techniques and image-based robotic systems is currently lacking. This is a gap that future research must aim to address.

Despite the individual promise demonstrated by both methods, direct head-to-head comparative data is sparse, heterogenous, and largely confined to unicompartmental knee arthroplasty (UKA).²⁵ Gaggiotti et al²⁶ retrospectively compared 292 medial UKAs and found that the when using image-based robotic assistance, superior accuracy for posterior tibial slope, tibial coronal alignment, and joint line restoration was demonstrated. Favroul et al²⁷ conducted the first study to directly compare both methods in the lateral compartment. This study reported that image-based robotic surgery achieved significantly better hip-knee-ankle (HKA) alignment and joint line restoration, along with with notably shorter operating times (61.6 vs 81.9 minutes; p < 0.001). A systematic review corroborated these findings, confirming that image-based robotic systems produce significantly fewer HKA outliers; however, this radiological advantage did not translate into a difference in revision rates at any follow-up period examined.²⁸ In the only randomized controlled trial (RCT) to date addressing this question in TKA, Narkbunnam et al²⁹ compared both methods within the same robotic platform, demonstrating equivalent radiological accuracy across all alignment parameters and comparable functional outcomes at 12 weeks. However, when comparing the robotic-validated intraoperative measurements to postoperative scanograms, fewer than 60% of cases fell within the 2° accuracy threshold for femoral sagittal alignment, raising the question of whether the demonstrated equivalence reflects true robotic precision or simply the limitations of the measurement tool itself. Furthermore, a 12-week follow-up horizon is insufficient to draw meaningful conclusions regarding implant survival or long-term functional outcomes. Moreover, a cadaveric study performed provides the mechanistic basis for the observed sagittal discrepancy in Narkbunnam et al’s²⁹ trial. Using CT-based reference as ground truth across 18 specimens, they demonstrated that while imageless technology could achieve comparable accuracy in defining the tibial sagittal plane, it lacked the precision and repeatability of image-based systems, with SDs approximately 30% larger across key rotation axes.³⁰ The clinical implication is significant, as errors in defining the tibial sagittal plane propagate into coronal alignment, potentially introducing unwanted tibial varus that may compromise implant fixation and long-term survival. With regard to THA, no study has yet directly compared image-based and imageless robotic systems, rendering this question entirely unanswered in that context. Across both hip and knee arthroplasty, the question of whether image-based systems confer a clinically meaningful advantage over imageless platforms therefore remains far from resolved, and robust long-term comparative data are conspicuously absent.

Emerging imaging methods may yet reshape this debate further. AI-driven 2D to 3D reconstruction algorithms, which generate bone models from standard radiographs as an alternative to CT, have existed for some time.^31,32 Clinical adoption has remained slow, however, with validation largely confined to small cohorts and concerns persisting in terms of accuracy in real-world pathological anatomy. Ultrasound-based registration represents an additional avenue under active investigation.^33-35 This method offers a radiation-free, low-cost, non-invasive alternative for intraoperative bone registration, although this technology remains at an early stage and has not yet been integrated into a clinically validated robotic arthroplasty platform. Phlippen et al³⁵ demonstrated an in vivo ultrasound-based femoral reconstruction workflow achieving sub-millimetre accuracy, and further reported an early-stage autonomous robotic scanning approach. However, the bone coverage in this autonomous technique remained incomplete, highlighting the gap that remains before clinical transition. MRI provides a further radiation-free alternative. Its established role in patient-specific instrumentation (PSI) in TKA has provided the most substantial body of comparative evidence to date. A recent systematic review and meta-analysis comparing CT- and MRI-based PSI found comparable alignment accuracy and clinical outcomes between methods, with MRI demonstrating marginally shorter operating times and slightly lower coronal alignment outliers.³⁶ Its translation into robotic workflows remain limited by cost, reduced access relative to CT, and the inherent difficulty of MRI in resolving cortical bone with sufficient clarity for surgical planning. Whether any of these approaches go on to mature into viable pathways, distinct from the conventional CT-based and imageless methods, remain to be determined.

The enthusiasm surrounding robotic arthroplasty, irrespective of modality, demands a degree of scrutiny. A recurring limitation in the existing literature is the conflation of radiological precision with clinical superiority. Improved component positioning is an appealing surrogate outcome; however, to date, neither image-based nor imageless robotic systems have shown long-term superiority in implant survival or patient-reported outcomes measures (PROMs) compared to conventional techniques.^1,37 A further complexity is the evolving debate in TKA surrounding alignment philosophy.³⁸ There has been an increase in popularity of alignment strategies that deviate from the classical mechanical alignment philosophy, with robotic technology permitting more contemporary philosophies such as restricted kinematic and functional alignment.^39,40 Image-based systems hold a theoretical advantage in this context: preoperative CT-derived modelling allows for personalized alignment planning that intraoperative landmark registration alone cannot fully replicate. As alignment philosophy continues to evolve, the comparative performance of each robotic modality may shift accordingly, which is a consideration that future trial design must account for. Perhaps the most consequential unanswered question, however, is not which system achieves superior radiological accuracy, but whether the additional cost burden of image-based workflows, encompassing preoperative CT acquisition, image segmentation, and associated consumables, is justified by a commensurate improvement in patient outcomes. In resource-constrained healthcare systems such as the NHS, where adoption decisions are increasingly governed by cost-effectiveness thresholds, this question may ultimately determine real-world implementation at scale.

A dimension that has received insufficient attention in this debate is the role of implant design as an independent variable in both system selection and outcomes interpretation. Current robotic platforms operate within closed, proprietary ecosystems in which robotic system selection implicitly dictates the implant family used. As a result, when image-based and imageless systems are compared, it is always across different implant families, rendering it impossible to observed differences to image modality alone, rather than the implant design itself.^25,41 Until there are open-platform robotic systems that permit the use of implants from different companies, implant design will remain an acknowledged confounder in this debate.

In conclusion, both image-based and imageless robotic arthroplasty systems represent meaningful advances in surgical precision, each offering distinct and legitimate value proposition. Image-based systems confer the advantages of preoperative patient-specific planning and superior implant size prediction, and it is particularly useful for cases with complex or deformed anatomy. Imageless systems offer comparable accuracy in the majority of cases, a streamlined workflow, reduced radiation burden, and greater accessibility across a broader range of institutions. Yet, despite the compelling individual evidence supporting each modality, no study has yet demonstrated that either system is clinically superior to the other. The field lacks the high-quality, adequately powered, prospective randomized data necessary to make evidence-based recommendations regarding system selection. Until such data exist, the choice between image-based and imageless platforms will continue to be driven by institutional infrastructure, surgeon preference, case complexity and commercial availability of the respective robotic systems. Looking ahead, the trajectory of robotic joint arthroplasty points toward a new generation of smaller, lower-cost systems with simplified workflows, driven in part by increasing market competition among device manufacturers. These developments may broaden access beyond high-volume centres, including into ambulatory surgical settings, and accelerate adoption across a wider range of healthcare settings. What is clear is that the next chapter of robotic arthroplasty must be written not in millimetres and degrees, but in years of implant survivorship and meaningful patient outcomes.