ABSTRACT. Modern surgical practice increasingly relies on data-driven technologies within the operating room (OR) to enhance precision, safety, and clinical outcomes. Advancing these technologies requires research infrastructures capable of supporting large-scale intraoperative data acquisition, prototyping, and validation. Current surgical settings face two main obstacles: limitations due to safety regulations, proprietary systems, and compliance requirements, and technical challenges in managing large datasets, achieving temporal and spatial synchronization, and developing robust interfaces for structured data integration. These constraints hinder the scalability and quality of intraoperative data acquisition and the application of artificial intelligence (AI), augmented reality (AR), robotics, and other advanced technologies in surgical research.

The Operating Room-X (OR-X) addresses these challenges by providing a fully equipped ex-vivo operating room designed to support the research, development and implementation of advanced technologies such as AR, AI and Robotics within a robust digital infrastructure. OR-X incorporates a multimodal data acquisition framework centered on custom-built data hubs that integrate various medical and research devices via flexible ports, and middleware that supports synchronized data acquisition and high-throughput streaming to a GPU cluster. This infrastructure enables real-time acquisition of high-resolution data from different sources. OR-X is developing an Open Research Data initiative that builds on this foundation, automating data fusion and structuring, while ensuring compliance with FAIR principles. This initiative includes a Surgical Data Cloud Platform, which facilitates the secure sharing of standardized datasets for collaborative research and development of interoperable surgical models.

OR-X also performed a first proof-of-concept for Surgical Digital Twins (SDTs), high-fidelity virtual models of surgical environments that integrate real-time data from multiple sources as showcased in Figure 1. These SDTs facilitate immersive training, workflow optimization and pre-operative planning. OR-X also aims to advance basic surgical models that use large-scale, standardized datasets for machine learning applications, including surgical phase recognition, AI-driven decision support, and robotic surgery.

OR-X demonstrates the potential of connected ORs, where devices communicate seamlessly as illustrated in Figure 2. The digitization of surgery enhances training, education and research, enabling immersive simulations and remote collaboration through SDTs. Automated data collection and classification are needed to generate large data sets for machine learning applications, while real-time feedback loops are needed for intra-operative decision making. By bridging experimental research and clinical practice, OR-X accelerates the translation of innovations into routine surgical workflows.

ABSTRACT. Developmental dysplasia of the hip (DDH) affects up to 3\% of newborns and if left undiagnosed can lead to early-onset of arthritis. It is typically diagnosed in neonates using 2D ultrasound (US) imaging, which can have high variability in the DDH metrics that are extracted from the scans. While 3D US can drastically reduce the variability in these metrics, it is costly and tends to be less accessible. Tracked 2D US systems can offer a more accessible alternative, but these systems conventionally rely on outside-in tracking, which can be cumbersome to set up due to typically large external tracking hardware. To address this limitation, we propose an inside-out tracking system for freehand tracked 2D US of the neonatal hip. This system uses a camera mounted on the US probe that tracks a non-planar arrangement of ArUco markers. We evaluate our system's performance by simultaneously tracking the US probe with both our proposed tracker and a conventional outside-in optical tracker (NDI Polaris Vega ST). Root mean squared (RMS) errors for positions and rotations were computed over multiple trajectories, including translational, rotational, and combined motions. Our results demonstrate that our inside-out tracking system achieves sub-millimetric and sub-degree accuracy, comparable to conventional outside-in trackers. This suggests that inside-out tracking is a viable alternative for freehand tracked 2D US scanning of the neonatal hip. Future work will assess clinical feasibility, the impact of patient movement, and the accuracy of dysplasia metrics extracted from reconstructed volumes. If validated, this approach could enable clinicians to perform reliable 3D reconstructions using affordable 2D US probes, thereby increasing accessibility to high-quality DDH diagnostics.

ABSTRACT. External fixation is a therapy option for operative treatment of open fractures, especially in low-resource settings. Pins for the fixator need to penetrate cortical bone on both sides of the target bone to ensure mechanical stability, but without extensive protrusion. Evaluation of pin placement is challenging, since medical imaging like radiography may not be available in low-resource settings. To address this, we propose a device for ultrasound-assisted pin placement using a portable, robust and low-cost ultrasound probe. The device comprises a guiding sleeve and an arm that centers the probe on the opposite side in the pin axis. Penetration of the opposite cortical bone can thus be easily detected in the ultrasound image in real-time, without the need for manual probe placement. We evaluated our concept in small scale experimental feasibility study on porcine lower legs, where we used a prototype to place 4 pins. Afterwards, we assessed pin protrusion in the ultrasound images as well as manually on the dissected bone. Mean pin protrusion measured in the US images was 1.6 mm, compared to 1.4 mm for manual measurement, with a mean deviation between measurements of 0.5 mm. Pin penetration of the opposite cortical bone was easily detectable in the ultrasound images, and the device facilitated central pin placement. Moreover, our device allowed for use by a single operator. We have thus demonstrated the feasibility of our concept. Future studies will focus on further optimization of the device and evaluation in a cadaver study with medical experts.

ABSTRACT. The use of fused colour and depth imaging (RGB-D) in conjunction with Deep-Learning algorithms is an emerging methodology in the field of surgical navigation and tracking. Training such systems to be robust across variations in patient anatomy requires large multi-site studies to collect sufficient data. Consequently, before initiating any such data collection, the efficacy of candidate RGB-D camera systems must first be trialled, to ensure suitability for capturing a specific target anatomy. We aim to investigate the fidelity of the point cloud captures, taken of a spine phantom model, from 3 different commercial camera systems. By assessing the success of classical registration methods of each camera's output point cloud against a 3D ground-truth, we evaluate the relative suitability of each system for collecting clinical orthopaedic data. In addition, for clinical data-collection, the need to distribute specialist hardware to each site as well as researcher time at each location can greatly limit the volume of data collected. Thus, we are highly interested in leveraging technology already present in hospital systems across the world, and with which clinicians are already highly familiar, to potentially make such studies vastly more scalable. Accordingly, as part of our comparative study,, we show how the TrueDepth camera, the inbuilt structured-light sensor used within iPhones for face-detection, can be used to capture spine anatomy with a fidelity comparable to standalone consumer grade depth cameras.

ABSTRACT. Navigated and robotic spine surgery systems improve pedicle screw placement accuracy but remain costly and inaccessible in many facilities. To address this, we propose self-localizing drill guides incorporating ultrasound (US) transducers to estimate position relative to pre-scanned anatomy. This study evaluates the localization accuracy of four potential tool designs—flat, curved radial, linear radial, and annular stack—by emulating anticipated US image acquisition with conventional US probes. Testing was performed on three spine models immersed in a water bath, including a single vertebra, an upscaled pedicle surface, and a full lumbar spine. Localization performance was assessed using translational and rotational metrics, combining these into a summary metric.

The flat and curved radial designs demonstrated the highest reliability, achieving localization errors of 1–3 mm in most scenarios. Calibration accuracy was generally acceptable, with reprojection errors under 2.5 mm, although deeper depths reduced accuracy for curved radial and annular stack designs. Localization errors increased on the full spine model except for the curved radial design, which delivered consistent performance.

These findings suggest US-integrated drill guides can achieve clinically acceptable accuracy, comparable to existing navigation systems, without requiring costly optical navigation. Strengths of the study include realistic spinal models and real US imaging, though limitations include testing in water baths and the use of conventional probes rather than custom transducers.

This approach represents a promising, cost-effective, and radiation-free alternative for spine navigation, justifying further development toward safer and more accessible surgical tools.

ABSTRACT. Patient-specific bone models are required for surgical planning of computer-assisted percutaneous scaphoid fracture fixation. 3D sonography may present an alternative to computed tomography (CT) or magnetic resonance imaging (MRI) for the acquisition of those bone models, but it requires a completion process to derive full models from partial sonographic surfaces. To date, methods based on statistical shape models (SSMs) represent the state-of-the-art for this completion process. However, we have shown feasibility of a deep learning (DL)-based method for scaphoid bone model completion in a previous study. In this study, we compare our DL-based approach against three SSM-based completion methods: Active shape models (ASM), least squares optimization (LSO) and general-purpose optimization (GPO). 85 scaphoid bone models were used for training the DL-based AdaPoinTr as well as for building the SSM. All completion methods were evaluated on 20 additional test models, with partial input point clouds generated using a subsampling algorithm that mimics 3D sonography. Evaluation in terms of symmetric surface distance between completed mesh and corresponding ground truth mesh showed 1.1 mm for ASM, 0.7 mm for GPO, 0.5 mm for LSO and 0.3 mm for AdaPoinTr. The assessment of suitability for screw planning showed 12 protruding screws for ASM, 4 protruding screws for GPO and LSO each, and no protrusion for AdaPoinTr. Also, AdaPoinTr was found to be at least one order of magnitude faster than all other methods. Nevertheless, SSM-based completion methods may be better suited for smaller datasets and if the generation of plausible shapes is to be ensured.

ABSTRACT. Objective Instability such as mid-flexion instability (MFI) is a leading cause of early failure following total knee arthroplasty (TKA). Advanced enabling technologies for TKA allow planning based on soft-tissue balance consideration, potentially reducing the risk of MFI. This study aimed to evaluate the incidence of MFI during TKA enhanced with force-controlled gap-balancing by analyzing joint laxity curves obtained during the trial reduction.

Material & Methods This study reviewed technical logs from 2864 TKA cases performed by 120 surgeons without any exclusions. At the trial reduction stage, the final joint laxities were acquired by placing an intra-articular force-controlled tensioner between the proximal tibial cut and the trial femoral component. Each individual joint laxity curve was processed to detect the possibility of MFI defined as the maximal joint gap measured in mid-flexion being 2mm larger than the maximal gaps measured in both extension and 90° flexion.

Results The incidence of potential MFI was 1.05% for the medial compartment and 1.12% for the lateral compartment. The incidence of potential MFI in both compartments in the same patient was 0.2%. Among cases with potential MFI, the mean MFI was 2.67 mm (range 2.03 to 5.03 mm) in the medial compartment and 2.77 mm (range 2.00 to 5.52mm) in the lateral compartment.

Conclusion Despite the adoption of enabling technologies, MFI remains a challenge in TKA. Even considering a conservative MFI threshold, the presented technique translated into a reduced risk of MFI, which may explain its previously reported improved clinical outcomes.

ABSTRACT. Objective Alignment techniques in total knee arthroplasty (TKA) are constantly evolving, with modern approaches providing clear recommendations for bone-cutting parameters. However, ligament laxities remain unclear. This study evaluates the laxity signatures established by surgeons when defining femoral cut planning based on different tibia cut scenarios.

Material & Methods A retrospective review analyzed 1762 TKA cases performed by 20 surgeons, stratified by bearing type: posterior-stabilized (PS), cruciate-retaining constrained (CRC), and cruciate-retaining (CR). The surgical technique allowed femoral planning concerning alignment, size, and ligament balance. Planned laxities were compared under actual tibia cuts during surgery (Group A) and simulated perpendicular cuts to the mechanical axis (Group B). While the simulated cut alters lateral gaps due to tibia cut angles, medial gaps remain consistent across both groups. Relative planned laxities were calculated for full flexion arc (10 to 120).

Results ANOVA showed significant differences (p < 0.05) in relative laxities among surgeons in both groups, regardless of bearing type or compartment side. Tukey comparisons revealed that significant pairs in Group B lateral laxity (PS: 71.4%, CRC: 80%, CR: 66.7%) were equal to or greater than those in group A (PS: 72.4%, CRC: 46.7%, CR: 66.7%), suggesting laxity definition is surgeon specific, regardless of tibia cut choice.

Conclusion Surgeons exhibit wide variability in defining laxities, preferring rectangular to trapezoidal gaps or greater flexion gaps than extension. Laxity definition for femoral cutting planning tends to be surgeon-specific, regardless of tibial cut reference. As knee arthroplasty customization progresses, there is an opportunity to define patient-specific laxities.

ABSTRACT. Posterior instability during deep flexion is a commonly identified risk factor in THA. This study investigated the risk of posterior instability due to adverse spinopelvic mobility before and after THA. 555 patients undergoing THA had pre-operative and 1-year post-operative standing and flex-seated radiographs measured for spinopelvic mobility parameters. Flexion impingement risk was determined by ∆PT≥20°. Patients were stratified into groups for pre- vs post-operative flexion risk: Group-1: Pre-No/Post-No, Group-2: Pre-No/Post-Yes, Group-3: Pre-Yes/Post-No, Group-4: Pre-Yes/Post-Yes. A classifier model was evaluated to predict post-operative flexion risk from pre-operative parameters. Pre-operatively, 13% of patients exhibited flexion risk (Group-3+Group-4), increasing to 29% post-operatively (Group-2+Group-4) (p<0.001). Group-2 patients (23%), had less ∆LL (37° vs 42°), and higher ∆PT (27° vs 6°), ∆PFA (110° vs 88°) and HUI (76% vs 68%) post- vs pre-operatively (all p<0.001). Compared with Group-1 patients, Group-2 had higher changes in ∆PT pre- to post-operative (20° vs 7°), were older (69 vs 64), were more likely female (71 vs 48%) and had lower ∆LL both pre- (42° vs 48°) and post-operatively (37° vs 45°) (all p<0.001). Sensitivity and specificity of the classifier was 76% and 70% respectively, with AUC accuracy of 80%. Gender, age, low lumbar flexion (∆LL≤20°) and Sagittal Deformity (PI-LL≥20°) were strong predictors. Nearly one in four patients exhibited a flexion risk (∆PT≥20°) between seated and standing positions post-operatively not identified pre-operatively, elevating their risk of posterior instability. At risk patients tended to be older, female and have stiffer spines pre-operatively than those not at risk both pre- and post-op.

ABSTRACT. The aim of this study is to investigate the progression of pain catastrophizing in end-stage osteoarthritis patients, to clarify the relationship among pain catastrophizing scale, patient-reported outcome measures (PROMs) and daily activity using a wearable physical activity monitor (WPM). This study evaluated 50 knees in patients with end-stage knee osteoarthritis before undergoing Total knee arthroplasty (TKA). Daily activities were assessed using a WPM one month before TKA. PROMs were assessed using the 2011 Knee Society Score before TKA. In patients with end-stage knee osteoarthritis, a reduction in high-impact physical activity time leads the progression of catastrophic thinking. However, no correlation was found between the patient satisfaction score and daily activities.

ABSTRACT. Objectives

Physiotherapy is an established part of the post-operative protocol for total knee replacement (TKR). As length of hospital stay has decreased, rehabilitation has moved to the home setting with little direct supervision. The Slider, a smart exercise device utilises gamification to optimize patient engagement during self-directed physiotherapy. This pilot study aimed to evaluate whether the Slider device in addition to standard physiotherapy could improve outcomes following TKR.

Methods

18 patients undergoing robotic primary TKR surgery at a single institution. Nine patients were allocated to the Slider group (device & standard physiotherapy) and nine patients to the standard physiotherapy group. Outcome measures included range of motion (ROM), Oxford Knee Score (OKS), EQ-5D-3L, and patient-reported outcome measures (PROMs) at six weeks postoperatively.

Results

Intraoperative and discharge ROM values were similar between groups (124 vs. 123, p=0.430; 76 vs. 78, p=0.624). Six-week postoperative ROM was superior in the Slider group (104 vs. 89, p=0.121). The Slider group had a shorter hospital stay and a similar number of inpatient physiotherapy sessions (2 vs. 3, p=0.332; and 4 vs. 4, p=0.999). The Slider group reported higher Likert scale satisfaction scores for post-op care (p=0.017). Both OKS and EQ-5D-3L were better in the Slider group, with OKS reaching statistical significance (39 vs. 33, p=0.045; 85 vs. 79, p=0.778).

Conclusion The Slider device aids early patient rehabilitation after total knee replacement, improving OKS and satisfaction scores. This device shows promise in supporting home-based rehabilitation, and will allow clinicians to supervise the process and identify struggling patients early.

ABSTRACT. Lateral locking-plate fixation is commonly used for distal femur fractures, and pre-contoured plates attempt to match the bony anatomy of the target patient population. However, plate fit is highly variable due to inter-subject morphological differences. If plate misfit is not recognized and addressed, this can result in axial malalignment. This study evaluates the effects of age, sex, height, ethnicity, and femoral bowing on the anatomical fit of a distal femur plate. Unilateral 3D models of 80 (40 male, 40 female) Caucasian and 79 (34 male, 45 female) Vietnamese femora were utilized. Both cohorts consisted of young (< 65 years) and old (≥ 65 years) subjects, with 40 young and 40 old Caucasian, and 36 young and 43 old Vietnamese. The plate undersurfaces of 9-, 11- and 13-hole LCP Distal Femur plate 3D models were positioned on the bone models and anatomical fitting was assessed by applying developed clinical criteria. Satisfactory plate conformity was achieved from plate head up to hole 6, with most measurement locations fitting 52–100% of bones for both ethnicities. There was proximal plate misfit from hole 8 onwards of 0–41% fit in that region, and a mean distance of 11.6 mm at the proximal plate tip for Caucasians, with a larger misfit (16.3 mm, p≤0.017) for Vietnamese. Patient sex, height, ethnicity, and femoral bowing all had significant impacts on fit, while the effect of age was limited. Due to sex and ethnicity related height differences, this suggests that patient height and femoral bowing are the main variables affecting fit. The observed proximal plate misfit for both Caucasians and Vietnamese suggests they might not fit as anatomically close as generally assumed. If the plate is used as a reduction tool, pre-operative templating of the intact contralateral femur can potentially identify plate misfit.

ABSTRACT. The Garden classification, based on X-ray interpretation and established over 50 years ago, remains the standard clinical classification system for femoral neck fractures (FNFs). Yet, this classification has a high interobserver variability of 70%. We sought to develop a deep-learning algorithm capable of accurately predicting FNF types, using only X-ray images, with performance comparable to that of computed tomography (CT). We retrospectively collected data from 1,588 patients who underwent X-ray and 3D-CT scans and were diagnosed with femoral neck fractures at Asan Medical Center. The input X-ray dataset consisted of paired X-ray images of the hip, with anteroposterior (AP) and lateral views. Using 3D-CT as the reference standard, patients were labeled as Garden types I (n = 378, 23.8%), II (n = 68, 4.3%), III (n = 477, 30.0%), and IV (n = 665, 41.9%). Our algorithm consisted of hip-joint detection followed by Garden classification, for which 12 different deep-learning architectures were evaluated. Algorithm performance was externally validated in 100 patients. Our algorithms showed a 91.3% overall accuracy, 90.5% sensitivity, 88.6% precision, and 88.9% specificity, indicating excellent FNF type discernment. Our algorithms could serve as a valuable tool for diagnosing FNF based on X-ray data only, with accuracy comparable to that of CT.

ABSTRACT. The aim of this study is to investigate the relationship between patient-reported outcome measures (PROMs) and patients’ daily movements, as evaluated by a wearable physical monitor, and to identify factors associated with improvements in daily life activities following total knee arthroplasty (TKA). In this study, varus knees underwent TKA was evaluated. Daily life activities were assessed using a WPM one month before TKA and from 4 to 6 weeks postoperatively. PROMs were assessed using the 2011 Knee Society Score before and 6 weeks after TKA. Additionally, preoperative limb muscle strength was evaluated using the Cybex system. Preoperative PROMs did not directly predict daily life movements after TKA. After 6 weeks after TKA, total activity recovered to approximately 74% of preoperative levels in this elderly Japanese population. Adequate preoperative flexion muscle strength significantly influences moderate physical activity levels in the early postoperative period.

ABSTRACT. Objective Traditional knee alignment assessment methods in total knee arthroplasty (TKA) rely on static radiographs, which do not account for dynamic changes in hip-knee-ankle (HKA) angles during movement. This can limit the accuracy of personalized alignment techniques, leading to suboptimal outcomes. The motivation of this study is to improve intraoperative dynamic HKA (dHKA) measurements using a novel method that applies a constant distraction force through the knee’s flexion arc, ensuring more precise and stable alignment for personalized surgical planning.

Material & Methods This retrospective review included 1790 tibia-first TKA cases performed by 87 surgeons using a proprietary cloud-based computer-assisted orthopedic system (CAOS). Conventional dHKA was acquired by manually manipulating the leg through flexion (0° to 120°), while the proposed method used an intra-articular device to apply a distraction force. Precut and postcut dHKA values from both methods were compared.

Results dHKA differences between the conventional and proposed methods were calculated for both precut and postcut phases. Precut Newton acquisition showed a more neutral alignment compared to conventional precut kinematics, with flatter differential curves observed across the full flexion arc. Postcut measurements showed minor differences, with the proposed method showing slightly less valgus.

Conclusion The novel intra-articular tensioner improves stability during dHKA acquisition, providing more reliable and neutral alignment data. The findings suggest that proposed acquisitions based on intra-articular device can serve as accurate surrogates for conventional kinematic methods. Incorporating dynamic knee phenotypes into surgical planning could further personalize alignment strategies in TKA, leading to better functional outcomes.

ABSTRACT. Objective: 3D bone shapes play a critical role in preclinical and clinical orthopedic applications. This study aimed to identify and evaluate 10 most relevant existing online CT databases to see if they meet requirements of biomedical experts. Method: We performed a systematic search to identify relevant online CT databases for lower extremities. Additionally, a workshop with n=40 biomedical experts was held to gather insights on the benefits, challenges, and users for an online 3D bone shape database. This information was used to establish criteria to evaluate the identified databases. Results: We found that currently available online databases inadequately address experts’ needs, particularly regarding inclusion of different shape formats, such as 3D meshes and CAD models, and inclusion of mechanical properties of bones. Conclusion: These findings highlight a significant gap between databases’ offerings and users’ needs, underscoring the need for more comprehensive, accessible resources and advanced tools to support the field's progression

ABSTRACT. Malunion of distal radius fractures alters joint alignment, mechanical environments, and subchondral bone density, contributing to osteoarthritis. Computer-simulated corrective osteotomy enables precise alignment restoration, but its effects on subchondral bone density remain unclear. This study evaluated subchondral bone density distribution changes in eight women (mean age: 63.3 ± 7.2 years) who underwent computer-simulated corrective osteotomy for malunited distal radius fractures. Pre- and postoperative CT data were used to create 3D bone surface models. Subchondral bone density within 3 mm of the articular surface was analyzed in four regions: volar and dorsal scaphoid fossa, and volar and dorsal lunate fossa. High-density volume (HDV) was defined as areas above the median bone density, and %HDV was compared to pre- and post-correction. Changes in the centroid coordinates of HDV distribution were also evaluated. Post-correction, %HDV increased significantly on the volar side (p < 0.05) and decreased on the dorsal side (p < 0.05). When comparing normal sides and the corrected, no significant differences in %HDV were observed in any area. Additionally, the centroid of HDV distribution shifted 0.69 mm toward the volar side in the volar-dorsal direction after correction. Corrective osteotomy normalized the subchondral bone density distribution, suggesting improved stress distribution. This may inhibit osteoarthritic progression, highlighting the biomechanical benefits of alignment correction. These findings indicate that computer-simulated corrective osteotomy normalizes subchondral bone density distribution, potentially mitigating osteoarthritis progression and highlighting its biomechanical benefits in malunited distal radius fractures.

ABSTRACT. Background: Elbow osteoarthritis often results in restricted joint range of motion (ROM) and endpoint pain due to osteophyte formation and joint contracture. Arthroscopic osteophyte debridement, although minimally invasive, presents challenges in accurately identifying lesions during surgery. This study aimed to evaluate the clinical utility of a surgical system combining 3D computer simulation and a CT-based navigation system for real-time visualization during arthroscopic debridement. Methods: Three patients (two men, one woman; mean age 53.3 years) with elbow osteoarthritis underwent arthroscopic osteophyte debridement using the proposed navigation system. Preoperative CT images were used to create 3D bone models for surgical planning. During surgery, a tracker system synchronized the arthroscopic and navigation monitors, enabling real-time visualization of the osteophyte removal sites. Surgical accuracy and ROM were evaluated postoperatively, with a mean follow-up of 9.6 months. Results: The overall osteophyte removal rate was 86%, with a total excessive removal rate of 12%. The posterior compartment achieved an 89% osteophyte removal rate with no excessive removal, while the anterior compartment achieved an 82% removal rate with a higher excessive removal rate of 27%. Postoperative ROM improved from -32 ± 10 degrees to -15 ± 13 degrees in extension and from 115 ± 10 degrees to 133 ± 8 degrees in flexion. Conclusion: The CT-based navigation system demonstrated high surgical accuracy and favorable short-term clinical outcomes in arthroscopic osteophyte debridement for elbow osteoarthritis. Further studies with larger cohorts and extended follow-ups are necessary to optimize techniques and validate long-term efficacy.

ABSTRACT. Context : The segmentation of forearm bones (radius and ulna) in medical images is essential for applications such as fracture diagnosis, surgical planning, 3D printing, and monitoring musculoskeletal pathologies. Objectives : This project aims to develop an automatic method for segmenting forearm bone structures using various deep learning techniques, enabling precise analysis of CT images.

Methods : 79 CT scans were manually annotated, constituting the reference annotations. Among them, 64 volumes were used for training, and 15 for testing. Models were evaluated, including the pre-trained TotalSegmentator model and nnU-Net, which was trained on our dataset.The nnU-Net method demonstrated the highest performance. The obtained segmentations were compared to the manual segmentations using the Dice metric and the Average Symmetric Surface Distance (ASSD).

Results : Preliminary results show that the nnU-Net model achieved an average Dice score of 0.9721, with values ranging from 0.9606 to 0.9811. The mean ASSD was 0.3304, with values varying between 0.2188 and 0.7141. Figure Figure 1 presents visualizations of the segmentation results, while Table Table 1 summarizes the obtained performances.

Conclusion : These results demonstrate that the nnU-Net deep learning method enables precise and automatic segmentation of forearm bones, thereby facilitating the creation of a model for diagnosis, decision support, and planning in emergency settings where the time between patient admission and surgical intervention is limited.

ABSTRACT. This comparative study evaluates functional outcomes in total knee arthroplasty (TKA) performed with a navigation system using femur-first measured resection (MR) or tibia-first gap balancing (GB) surgical workflows. A single surgeon at one center conducted all procedures using the same implant and navigation system. Data from 123 patients, including demographic information and Oxford Knee Scores (OKS) at preoperative, six weeks, one year, and two-year follow-ups, were analyzed. Results showed no significant differences in early postoperative outcomes between workflows. However, from one year onward, the GB group demonstrated superior functional results, with an OKS improvement of 5.8 points at two years, exceeding the MCID. The GB approach, which integrates joint laxity data into femoral planning, may offer better joint balance and antero-posterior stability over time.

ABSTRACT. While diffuse idiopathic hyperostosis (DISH) has a predilection for the thoracic spine and is not associated with limb joint disorders, few reports have investigated the disease in detail. In this study, we investigated the association between ossification of spinal ligaments and ossification around the hip joint in patients with DISH by CT. Among patients who had CT from the neck to the pelvis from 2009 to 2018, 113 had DISH localized to the thoracic spine (T-DISH) and 43 had DISH extending to the lumbar spine (T/L-DISH), 56 patients did not meet Resnick's criteria but had a certain amount of ossification between three vertebrae. （This was defined as Pre-DISH.) In these patients, the presence of periprosthetic ossification, narrowing of the joint space, lateral center edge angle (CE), anterior CE angle, and posterior acetabular sector angle (PASA) were statistically analyzed. Ossification increased significantly in the superior and posterior hip joints with progression from Pre-DISH, T-DISH, and T/L-DISH. The joint fissure was predominantly narrowed bilaterally in Pre-DISH and T/L-DISH, and T-DISH and T/L-DISH. With the development of DISH, especially when DISH extends to the thoracolumbar spine, there is increased ossification around the hip joint, with parameter changes and narrowing of the joint space.

ABSTRACT. Robotic platforms enable TKA planning to reproduce the native joint line. Coronal plane alignment of the knee (CPAK) classifies knee phenotypes using joint line obliquity (JLO) and arithmetic hip-knee-ankle (aHKA). This study evaluates the impact of automated pre-resection gap acquisition on CPAK in robotic TKA. A single surgeon’s 500 consecutive primary TKAs in which a robotic ligament tensioner was used to assess medial and lateral joint gaps were retrospectively analyzed. The first 250 (mid-resection) assessed gaps after the proximal tibial resection, before femoral resections. The next 250 cases (pre-resection) assessed gaps before any bone resections using a novel version of the system. Medial proximal tibial angle (MPTA) and lateral distal femoral angle (LDFA) were measured using the landmark data collected from the navigation system, adjusting for cartilage wear. Native MPTA was filtered between 85°-91° for both groups to ensure comparable populations. Native, post-operative, and ∆ values were analyzed, along with CPAK phenotype distributions and changes. Preoperative MPTA and LDFA were within 1.5° between groups. Postoperative measurements showed significant differences between pre- and mid-resection for MPTA (p<0.001), and LDFA (p<0.001). The pre-resection group had ∆MPTA further from 0 (p<0.001), ∆LDFA closer to 0 (p<0.001), and more patients whose CPAK phenotype didn’t change postop (35% vs 17%, p<0.001). Differences in laxity were within 0.4mm between groups, which are not likely clinically significant. Planning a TKA using robotic pre-resection ligament data better restored native CPAK compared to using robotic soft tissue assessment after tibial resection and achieved similar final laxity and balance.

ABSTRACT. Advances in robotic assisted TKA enable pre-resection soft tissue assessment, supporting both femur- and tibia-first workflows. This study compares predictive planning accuracy for both femur- and tibia-first workflows. The first 734 cases from seven surgeons were retrospectively reviewed. Four used a pre-resection femur-first workflow while three used a traditional tibia-first workflow with post-resection ligament assessment. Laxity accuracy was compared between femur-first and tibia-first workflows, along with robotic surgical time from registration to trialing. The robotic tensioner applied 70-90N per side, recording joint laxity from flexion to extension. A planning algorithm predicted final laxity based on the laxity data and planned component placement. Planned laxity (predicted femoral-tibial gap) and final laxity (measured during trialing) were compared using mean absolute error (MAE) at 10°, 45°, and 90°. Tibia-first medial laxity MAE was greater at 10° (1.3±1 vs 1.1±0.9 mm, p<0.05), and 90° (1.3±1.1 vs 1±1 mm, p<0.01), while lateral laxity MAE was greater at 10° (1.4±1.2 vs 1.1±1 mm, p<0.001). Femur-first had shorter mean robotic surgical time (19±6 vs 23±7 min, p<0.0001). Assessing soft tissues with a robotic tensioner before bone resection resulted in accurate final gaps with shorter robotic surgical times. This study supports the integration of predictive planning using a novel pre-bone resection robotic soft tissue assessment tool and supports femur-first surgical protocols as a safe and effective means of optimizing soft tissue laxity.

ABSTRACT. Purpose: This study aimed to predict the pull-out strength of robot-assisted pedicle screws using bone mineral density (BMD) of preset 3D trajectory, to provide data for planning the screw trajectory with optimal strength. Methods: Before and after robot-assisted pedicle screw insertion in cadaveric lumbar spine specimens, scans were conducted using the same high-resolution quantitative Computed Tomography (QCT) parameters. Actual bone parameters along the preset trajectory were obtained by merging the two scan results. BMD calculations for the cylindrical 3D screw channel (3DSC) and its surrounding cylindrical 3D peripheral channel (3DPC) were performed using Python. The 3DPC thickness was gradient-divided by 2 times the QCT resolution. The 3D channel was segmented into pedicle and vertebral body parts for correlation analysis. Pull-out strength from biomechanical experiments was correlated with different 3D channel parts to construct prediction models. Results: A total of 81 pedicle screws (47 of 5.0mm*40mm and 34 of 6.0mm*40mm) were inserted using robot. Pull-out strength exhibited stronger correlation with 3DPC bone parameters than 3DSC parameters for both screw types. The correlation was stronger with pedicle part bone parameters than vertebral body part parameters. The best prediction model for 5.0mm*40mm screws was 1.95 * 0.3mm 3DPCp_BMD - 51.81 (R2=0.7414), and for 6.0mm*40mm screws, the model was 3.98 * 0.3mm 3DPC_BMD - 124.92 (R2=0.7502). Conclusion: The pull-out strength has the strongest correlation with the bone parameters of 0.3mm 3DPC, which is stronger than that of 3DSC.

ABSTRACT. Augmented reality (AR) provides opportunities to further enhance surgical navigation, offering immersive visualisation during complex orthopaedic procedures like periacetabular osteotomy (PAO). This study introduces an AR-enhanced navigation platform utilising the HoloLens 2 headset to address challenges of traditional fluoroscopy-guided PAO, including poor visibility of cutting-planes and the cognitive challenges of interpreting 2D images for 3D corrections of acetabular malorientation. By extending the functionality of an existing PAO planning-software, our system integrates tracking capabilities with immersive, in-situ virtual guidance projected onto the surgical site, to enable the execution of patient-specific plans. The platform employs an OptiTrack motion-capture system to precisely track patient and tool motion. A Unity-based HoloLens application tracks infrared markers, enabling co-registration with OptiTrack and real-time projection of virtual holograms of cutting planes and surgical plans onto a mobile pelvis. Preliminary evaluations of the system’s pose-tracking and patient-registration accuracy indicate that while our HoloLens software provides sufficient precision for visualisation, it falls short of the sub-millimeter, sub-degree accuracy required to be the sole tracking device in our setup. The complementary OptiTrack system achieves the necessary precision for navigation. The patient registration process of our study, using iterative closest-point optimisation, yielded a maximum mean absolute point-to-surface error of 1.383 mm, highlighting areas for refinement. This work demonstrates the successful integration of AR into PAO navigation, combining immersive visualisation with high-precision tracking. Future efforts will focus on improving registration workflows and evaluating clinical efficacy to reduce procedural complexity and improve surgical outcomes.

ABSTRACT. Cervical spine surgery, particularly for dorsal decompression and fusion in cervical myelopathy treatment, requires precise usage of surgical devices and instruments to achieve an optimal outcome. The surgery aims to decompress the spinal cord by removing any structures compressing the nerves. While open networked central workstations have the potential to increase efficiency and safety, they can face workflow-driven conflicts, such as limited input resources, insufficient screen space for device or patient information, and inadequate control methods. Moreover, current standard operating procedures (SOP) lack detailed information about the inter-device communication requirements. Preventing workflow-driven errors is already addressed in high-risk applications, such as airspace and nuclear power plant control rooms. This paper proposes a method to mitigate workflow-driven conflicts for open networked ISO IEEE 11073 SDC service-oriented device connectivity workstations. By extending clinical SOPs by specific device and instrument usage specifications (eSOP) especially related to human-machine-interaction (HMI) requirements, we could identify potential conflicts in a proposed central OR workstation solution before bringing devices into service. The eSOP has been discussed with spine surgery specialists from the University Hospital RWTH Aachen.

ABSTRACT. Computer-Assisted Surgery (CAS) systems enhance joint replacement accuracy. Total ankle arthroplasty (TAA) is a viable alternative to ankle fusion, but achieving precise implant alignment remains challenging due to limited surgical exposure and reliance on fluoroscopy. This study aimed to compare the accuracy of a conventional TAA technique using fluoroscopy to a previously developed CAS system.

Twelve artificial ankle joints were used for the study. TAA was performed using conventional instrumentation and fluoroscopy. Bone resections were performed, and the resections were compared to the planned resections using 3D scanning and analysis software. Tibial and talar resections showed greater accuracy with the CAS system compared to the conventional technique. The conventional technique had larger deviations in tibial closed slope, tibial internal rotation, and talar slope. The conventional technique demonstrated acceptable accuracy, and was aligned with previous literature. However, the CAS system improved accuracy and precision, particularly by reducing outliers. Limitations of this study include the absence of soft tissues and limited surgeon variability with only one operator. Future studies should investigate surgeon variability, cadaveric models, and comparisons with patient-specific instrumentation (PSI) techniques. In conclusion, the conventional technique provides acceptable results, while the CAS system offers potential for enhanced accuracy and precision in TAA.

ABSTRACT. Pedicle screw placement is a critical procedure in spinal surgery, traditionally guided by bone-anchored tracking systems. These systems, while effective, are invasive and prone to complications such as soft tissue damage and bleeding. To minimize invasiveness, surface navigation systems using external tracking markers have been developed. However, their accuracy can be compromised by dynamic factors such as respiration and spinal movement. This study introduces a novel non-invasive tracking device designed to enhance the stability and precision of surface navigation in dynamic surgical environments. The device consists of five reflective markers integrated into a silicone base, which allows automatic marker recognition and real-time tracking. We tested this device in a simulated dynamic environment using a robotic arm and lumbar model to replicate respiratory motion. Results demonstrated that the device achieved high accuracy in guiding pedicle screw placement, with registration errors of 0.524 ± 0.229 mm and minimal angular deviations. The tracking device performed reliably under dynamic conditions, offering potential benefits in reducing surgical trauma and improving patient outcomes. This study provides a foundation for further developing and optimizing non-invasive navigation technologies in complex spinal surgeries.