CAOS 2022:Papers with Abstracts

Abstract. Recent developments have focused on the intra-operative management of soft-tissue balancing in total knee arthroplasty (TKA) using a computer-assisted orthopaedic surgery (CAOS) system. The aim of this study was to determine and compare the reliability of acquiring the knee joint laxities during navigated TKA with a conventional method versus a newly developed instrumented technique that uses an intra-articular quasi- constant force distractor integrated with a CAOS system. A total of 96 laxity acquisitions throughout the arc of motion were performed for the conventional and instrumented procedures. For the instrumented technique, the inter- and intraobserver reliabilities were significantly higher than the conventional manual varus/valgus stress test technique, regardless of surgeon variability and experience. Soft-tissue balance, while being a key determinant in improving outcomes in TKA, is difficult to objectively assess at the time of the surgery. This study established that the acquisition of the knee joint laxities using an instrumented technique was consistently associated with a significantly higher reliability than the conventional technique.
Abstract. Total knee replacement (TKA) represents a well-established reconstructive procedure for end-stage knee joint disorders with the balancing of soft-tissue envelope throughout the full arc of motion as a newly emerging possibility. This cadaveric study evaluated the ability to achieve targeted mediolateral (ML) gap balance throughout the arc of motion using conventional mechanical instrumentation versus a computer-assisted orthopaedic surgery (CAOS) system featuring an intraarticular distractor while considering surgeon experience level. For the CAOS system, an intraarticular distractor applied a quasi- constant distraction force to the joint (instrumented) while the conventional system involved conventional spacers. Regardless of experience level, the instrumented TKAs were associated with a significantly lower ML gap differential than the conventional TKAs. In contrast, regardless of the type of instrumentation, there were no significant differences between the junior and senior surgeon mean gaps. Historically, soft tissue balancing during TKA has been reported as an art rather than a science. In this regard, the addition of dedicated technology to characterize the soft-tissue envelope during TKA has the potential to provide an augmented perspective to the surgeon and can be particularly beneficial for junior surgeons. The present study established that the usage of instrumented CAOS led to significantly lower ML gap differences than conventional instrumentation.
Abstract. Appropriate management of the soft tissue envelope at the time of the surgery is critical to the long- term success of total knee arthroplasty (TKA). In this regard, this
computer-assisted orthopedic surgery (
a force-controlled intraarticular distractor. The first 150 cases performed by 16 surgeons were reported without any exclusions, and for each of these cases, the final mediolateral (ML) laxity was compared to the predicted ML laxity. The average signed ML laxity was well aligned with a neutral differential throughout the full arc of motion and ranged from -0.05mm at 35° of flexion to 0.37mm at 85° of flexion. The signed ML laxity curves tend to be surgeon specific. The average unsigned ML laxity was linear throughout the full arc motion and ranged from 1.14mm at 85° of flexion to 1.27mm
at 30° of flexion.
the targeted ML gap balance when using a
Despite data from all the users (not only design surgeons) involved with this pilot
release were considered and the learning curve cases were not excluded, it was observed a high ability
study evaluated the ability to achieve
CAOS) system
to achieve the targeted ML laxity using the proposed method.
Abstract. Proper soft tissue balancing during total knee arthroplasty (TKA) is critical to ensure successful clinical outcomes. As an attempt to offer an intra-operative characterization of the soft-tissue envelope, a novel method enables the possibility of acquiring the joint laxities under a quasi-constant distraction force throughout the entire range of motion. TKAs were performed using a computer-assisted orthopaedic surgery (CAOS) system on a fresh-frozen human cadaveric specimen. A total of 60 laxity acquisitions were performed by 5 surgeons using the CAOS system. There was an excellent interobserver reliability of the laxity acquisitions (ICC=0.913-0.992). Similarly, the intraobserver reliability was also excellent (ICC=0.846-0.984). These findings demonstrated that the acquisition of the knee joint laxities under the proposed controlled load environment is highly reliable.
Abstract. Remote patient monitoring, using wearable devices and connected patient engagement platforms has the potential to improve timely clinical decisions. Data collected from multiple patients, including using the remote engagement platforms themselves, can be used to produce evidence-based reference to support clinical decisions. While some normative references for functional measure currently exist for total knee arthroplasty (TKA), these are still lacking for VAS pain scores. Therefore, VAS pain scores on a 10-point Likert scale were analyzed for 66 patients, each reporting at least five scores in the 180 days following surgery. These were used to produce a normative recovery model for total knee arthroplasty patients. A nonlinear mixed effects model was fitted, whereby the response variable is assumed to be distributed following a beta-binomial distribution. The population mean trend showed a with wide dispersion in the first few days following surgery, showing scores ranging throughout the 10-point scale. After the first week, the expected pain score steadily decreases, resulting in a score no higher than one in 50% of the population beyond 90 days after surgery. The fitted model allows referencing individual patient's pain scores at different stages of recovery, against the model’s predicted distribution. This approach can support early detection of patients that significantly deviate from the reference model and be a useful integration into clinical decision support software tools.
Abstract. For the complex clinical issue of treatment decision for proximal humeral fractures, dedicated software based on three-dimensional (3D) computer tomography (CT) models would potentially allow for a more accurate fracture classification and help to plan the surgical strategy needed to reduce the fracture in the operating theatre. The aim of this study was to elaborate the feasibility of implementation of such software using state-of-the-art cloud technology to enable access to its functionalities in a distributed manner. Feasibility was studied by implementation of a prototype application, which was tested in a usability study with five biomedical engineers.
Implementation of a cloud-based solution was feasible using state-of-the-art technology under application of a specific software architectural approach allowing to distribute computational load between client and server. Mean System Usability Scale (SUS) Score for the developed application was determined to be 63 (StDev 20.4). These results can be interpreted as a medium low usability with high standard deviation of the measured SUS score. We conclude that more test subjects should be included in future studies and the developed application should be evaluated with a representative user group such as orthopaedic shoulder surgeons in a clinical setting.
Abstract. The automated and robust segmentation of bone surfaces in ultrasound (US) images can open up new fields of application for US imaging in computer-assisted orthopedic surgery, e.g. for the patient-specific planning process in computer-assisted knee replacement. For the automated, deep learning-based segmentation of medical images, CNN-based methods have been the state of the art over the last years, while recently Transformer-based methods are on the rise in computer vision. To compare these methods with respect to US image segmentation, in this paper the recent Transformer- based Swin-UNet is exemplarily benchmarked against the commonly used CNN-based nnUNet on the application of in-vivo 2D US knee segmentation.
Trained and tested on our own dataset with 8166 annotated images (split in 7155 and 1011 images respectively), both the nnUNet and the pre-trained Swin-UNet show a Dice coefficient of 0.78 during testing. For distances between skeletonized labels and predictions, a symmetric Hausdorff distance of 44.69 pixels and a symmetric surface distance of 5.77 pixels is found for nnUNet as compared to 42.78 pixels and 5.68 pixels respectively for the Swin-UNet. Based on qualitative assessment, the Transformer-based Swin-UNet appears to benefit from its capability of learning global relationships as compared to the CNN-based nnUNet, while the latter shows more consistent and smooth predictions on a local level, presumably due to the character of convolution operation. Besides, the Swin-UNet requires generalized pre-training to be competitive.
Since both architectures are evenly suited for the task at hand, for our future work, hybrid architectures combining the characteristic advantages of Transformer-based and CNN-based methods seem promising for US image segmentation.
Abstract. Reference axis based on Friedman’s approach is widely recognized as an anatomic landmark from which to measure and compare implant parameters within preoperative planning software for total shoulder arthroplasty. Equinoxe Planning Application (ExactechInc.) offers 3D measurements techniques for glenoid version and inclination requiring meticulous placement of trigonum and glenoid center. We propose as automatic determination of this reference axis, based on deep learning that shown a median error of less than 1°.
Abstract. Introduction
Patient recovery from neuromuscular injuries that cause upper limb dysfunction is commonly assessed via manual methods. Manual muscle testing is subjective, time consuming and requires extensive training. Existing dynamometers are more objective, but they are prohibitively expensive and impractically large, making them inaccessible to most clinics and patients with disabilities. Our aim is to develop a table-top upper limb muscle dynamometer that provides standard positioning, ease of use and portability while giving clinicians consistent and reliable quantitative data on a patient’s isotonic and isometric muscle power and strength, respectively.

The device consists of a lever arm, a brushless DC motor, a load sensor and an ergonomic cuff. It outputs analog data via standard BNC connectors. The device can be intuitively controlled by the operator to test various upper limb joints and motions. Isometric measurement repeatability was assessed by recording the maximal voluntary contractions of 18 healthy participants over three trials.

The repeatability across 3 trials was 2.70±2.27 Nm (95th percentile: 6.74 Nm) for elbow flexion, and 2.83±2.13 Nm (95th percentile: 5.65 Nm) for elbow extension.

The dynamometer demonstrates a marked improvement in repeatability relative to manual muscle testing. Its small footprint and low cost can make it an easily accessible, standardized testing tool that requires little training to use. Future research and development will focus on using field-oriented control to measure isotonic muscle power in addition to isometric strength.
Abstract. Progress in machine learning and artificial intelligence (AI) opens the way to the devel- opment of smart clinical-assistance systems and decision-support tools for the operating room (OR). Yet, before deploying these algorithms in the OR, assessment of their perfor- mances in real clinical conditions is necessary. Gathering intraoperative data for training and testing is hard, and robustness to the challenging conditions of the OR is not always demonstrated. In this paper we introduce a unique multi-patient dataset of images cap- tured during Total Knee Arthroplasty (TKA) surgery. We use this dataset to compare five deep learning-based image segmentation approaches and provide quantitative and qualita- tive results. We hope that this work will help bringing light on the performances of AI in a real surgical environment.
Abstract. The purpose of the present study was to associate the intraoperative kinematics acquired with a computer navigation system with long-term clinical outcomes and survivorship in patients undergoing TKA to investigate the role of constraint in patients’ satisfaction.
A surgical navigation system was used to verify bone resections, gaps, and implant positioning during TKA. Kinematic examination, i.e. varus-valgus at full-extended knee (VV0), varus-valgus at 30° of flexion (VV30), anterior/posterior displacement at 90° of flexion (AP90), passive range of motion (ROM) were performed. Long-term clinical assessment interviews were performed. The Knee Injury and Osteoarthritis Outcome Score (KOOS) was used to investigate patients’ clinical and functional status.
Out of 165 patients, 120 met the inclusion criteria. The average follow-up time was 7.7±2.8 years. 7 patients had undergone revision surgery and were considered as a surgical failure with an overall survival rate of 94.2%, while the survival rate at 6, 8, 10 years was 98.8%, 97.4%, 93.6%, respectively. Clinical failure (KOOS score <70) was detected in 11 (9.2%), 10 (8.3%), 21 (17.5%), 39 (32.5%), 113 (94.2%) patients for the Symptoms, Pain, ADL, QoL, and Sport sub-scores, respectively. A statistically significant difference was found in KOOS-QoL between patients with and without clinical failure for the VV0 test (ES=0.58, p=0.022), with lower laxity for patients with score<70.
Over-constraint kinematics during TKA surgery leads to worse clinical outcomes at long-term follow-up. Surgeons should be aware of the intraoperative ligament balancing and avoid over-constraint, especially in PS TKA designs.
Abstract. Minimally invasive intervention requires accuracy and practice as it can be vital in complex and narrow places. In this paper we propose a solution based on augmented reality (AR) for the ablation of bone tumors. Our proposal deals with the preoperative and intraoperative phases of the procedure. The first part consists of the segmentation and 3D reconstruction of the structures of interest. The second part consists of the visualization in AR. This solution is intended to facilitate the tasks of surgeons and radiologists when planning RF needle insertion and trajectory in order to avoid excessive exposure to X-rays, which is a phase that requires more precision and knowledge of the morphology of the mass tumor. The second part offers AR assistance based on the planning of the preoperative phase. The solution we proposed is based on the use of HoloLens 2 headsets to provide better AR visualization and assistance.
Abstract. Aims: Several studies have been performed that compare the accuracy of Robotic-Assisted Total Knee Arthroplasty (RATKA) to conventional instrumentation as well as navigation to conventional instrumentation, yet there is a lack of studies comparing RATKA to navigation. The purpose of this study is to evaluate the accuracy of a contemporary image free navigation system for TKA in a cadaveric study using the same methodology as used previously to access the accuracy of a RATKA system and conventional instrumentation. Methods: Four orthopaedic surgeons performed bi-lateral TKA on 18 pelvis-to-toe cadaveric specimens without implantation using the BrainLab Knee3 navigation system. Preoperative and postoperative computed tomography (CT) scans were taken to access the resection accuracy of the navigation system relative to the planned alignment targets recorded intraoperatively. Results: The mean error in femoral coronal angle was 1.08° ± 0.87° compared to 1.39° ± 0.95° conventional and 0.63° ± 0.50° RATKA; the differences between navigation and RATKA were statistically significant. The mean error in the tibial coronal angle was 1.24° ± 1.13° compared to 1.65° ± 1.29° conventional and 0.93° ± 0.72° RATKA. The mean error in femoral flexion was 2.13° ± 1.87° compared to 3.27° ± 2.51° conventional and 1.21° ± 0.90° RATKA; the differences between navigation and manual and navigation and RATKA were statistically significant. The mean errors in the femoral rotation (navigation 1.30° ± 1.38°, conventional 1.00° ± 0.70°, RATKA 1.04° ± 0.81°) and tibial slope (navigation 1.89° ± 1.28°, conventional 1.63° ± 1.39°, RATKA 1.62° ± 1.13°) were similar between the groups. Conclusion: This study showed that for some metrics navigation improves resection accuracy compared to conventional instrumentation and RATKA further improves resection accuracy compared to CAS.
Abstract. The objective of the current paper is to present a pipeline designed to reduce the pre-processing time required to build subject-specific finite element knee models and facilitate their clinical integration. The pipeline involves development and validation of an atlas model of the knee joint and features of the TwInsight software suit that use novel methodologies such as: 1) deep learning for automatic segmentation of the bones from computed tomography scans, 2) automatic generation of finite element meshes with hexahedral elements, and 3) anatomical inference algorithm to adapt the atlas model to the morphology of a subject and result in the subject’s personalized biomechanical model.
Abstract. In clinical routine, the capture of three-dimensional (3D) bone geometry is crucial for surgical planning, implant placement and postoperative evaluation. Nevertheless, accurate 3D reconstruction of the knee joint for the estimation of patient-specific features remains a challenge, although it has been widely studied. In this context, statistical shape models (SSM) have been used to reconstruct a global shape from partial observations, based on their ability to capture the anatomical variation from different patients. However, these studies incorporate single object SSMs which limit their application for analyzing local bone morphology and thus they lack the capacity to analyze the human anatomy at the joint level. In this paper, we present a multi-object based framework for the 3D reconstruction of the knee joint using a dynamic multi-object Gaussian process model (DMO-GPM) and an adapted Markov Chain Monte Carlo (MCMC) based model fitting algorithm.
The knees were reconstructed with an average mean square error of 1.81±0.37 mm and maximum error of 3.31 mm corresponding to the surface-to-surface distance between the predicted and original knees. The results show that the knee is accurately reconstructed, especially around the joint contact surfaces. This is crucial because most of the patient- specific features required for the implant design, use landmarks in this area. The results suggest that the approach is robust and accurate to design personalized knee implants.
Abstract. Preoperative anatomic measurements in total shoulder arthroplasty (TSA) influence a surgeon’s decision-making process in deciding treatment options for a given patient. Glenoid retroversion is one of the most significant measurements and can be highly subject to intra- and inter-observer variability in measurement technique. This study compares surgeon measured retroversion values to semi-automated software measured retroversion values on the same 1862 computed tomography scans, showing consistent measurements with an average absolute mean error between the two techniques of 3.1 ± 3.6°
Abstract. Two-year minimum clinical outcomes were collected on anatomic and reverse total shoulder arthroplasty patients enrolled in a single implant global registry that were performed using an intraoperative computer navigated surgery system. Age, gender, and follow-up matched cohorts were created from the same registry for comparison purposes for both anatomic and reverse total shoulder arthroplasty. The navigated cohorts exhibited as good or better clinical outcomes compared to the non-navigated cohorts as well as reductions in postoperative complications, revision rates, and adverse events.
Abstract. Interbone parameters of the knee are of relevance in clinical practice, e.g. for the assessment of the functional anatomy of the individual patient. However, respective landmark identification and parameter derivation is mostly done manually. An automated analysis could enable the processing of large datasets, which could again enable the derivation of reference ranges or safe zones for various populations. Hence, the aim of this study was to automate the derivation of interbone parameters from 3D surface data of the knee and to evaluate the method’s robustness against a large dataset.
A dataset of 414 knees from patients scheduled for total knee arthroplasty (TKA) was available for the analysis. For each case, knee surface models derived from CT as well as coordinates of the hip and ankle joint centers were available. Eight interbone parameters of the knee were identified in a literature research and an existing framework for morphological analysis of the knee was extended, in order to automatically calculate those parameters.
The interbone analysis succeeded for 405 (97.8%) cases. After the exclusion of implausible cases, 373 (90.1%) parameter sets remained for statistical analysis.
Differences in methodology, populations, imaging technique etc. complicate the comparison with values from the literature. However, for similar studies a good agreement in parameter values was found.
The workflow presented proved robust against a large dataset of knee surface models. In the future, information about the bones’ relative position in the active, weight-bearing situation should be incorporated, in order to assess the impact on knee interbone parameters.
Abstract. Mismatch between the patient’s knee morphology and the implant geometry is linked to poorer clinical outcome after total knee arthroplasty (TKA). Hence, patients whose knee morphology differs strongly from the norm may have a higher risk to be dissatisfied after surgery. Consequently, a preoperative risk assessment regarding differences between individual knee morphology and implant geometry is favorable. For adequate availability and limited radiation dose, this should be based on standard imaging in TKA, being conventional radiographs.
We reviewed morphological measures of the knee to be evaluated on X-ray images. Only measures of the articulating areas, without connections to pathologies such as patellar instability or pain, were included. In addition, the accuracy of 2D-3D knee reconstruction was reviewed, in order to assess the potential use for 3D X-ray based analysis.
Various parameter definitions for the evaluation on anterior-posterior and lateral X-rays exist in the literature. If given, the inter- and intraobserver reliability can be interpreted as moderate to excellent. Several authors have reported on 2D-3D reconstruction accuracies with maximum absolute errors of ~5-6 mm for in vitro studies.
Mismatch between the bone morphology implant geometry can partly be assessed in 2D, using single X-rays. Methods for 2D-3D reconstruction demonstrated potential for enabling 3D X ray-based analyses. However, improvements regarding accuracy and larger in vivo validation studies are pending.
A basic preoperative risk assessment using X-rays is possible. Future steps could include the automation of the parameter derivation and an enhancement of 2D-3D reconstruction for enabling a more comprehensive assessment.
Abstract. The usual safe zone for cup orientation in THA is not suitable for all patients, as the pelvic tilt varies with the movements of daily activities. A new Functional Safe Zone (FSZ) is proposed that considers the pelvic tilt in different positions. The aims of this study were to validate the proposed FSZ and to evaluate how the pelvic mobility impact it.
We measured the pelvic tilts of 30 patients when standing, sitting and supine, using our ultrasound-based device and computed their FSZs. The FSZs accuracy was assessed using a Computer-Aided-Design (CAD) software. The pelvic mobility influence onto the FSZ was assessed by jointly analysing the patients’ FSZs and their pelvic tilt difference between positions.
The true FSZ provided by the CAD software and the estimated FSZ were similar by 92% and differed by less than 0.5◦ at borders and at the mean orientation. Patients with stiff pelvic mobility obtained small FSZs, and conversely, patients with large pelvic tilt variations between positions obtained large FSZs.
The proposed method allows the computation of a patient-specific FSZ without requir- ing additional X-ray or CT images. Patients having a low pelvic mobility with a higher risk of postoperative instability could be better managed using this FSZ.
Abstract. Rotator cuff tears (RCT) are one of the most common sources of shoulder pain. Many factors can be considered to choose the right surgical treatment procedure. Of the most important factors are the tear retraction and tear width, assessed manually on preoperative MRI.
A novel approach to automatically quantify a rotator cuff tear, based on the segmentation of the tear from MRI images, was developed and validated. For segmentation, a neural network was trained and methods for the automatic calculation of the tear width and retraction from the segmented tear volume were developed.
The accuracy of the automatic segmentation and the automated tear analysis were evaluated relative to manual consensus segmentations by two clinical experts. Variance in the manual segmentations was assessed in an interrater variability study of two clinical experts.
The accuracy of the tear retraction calculation based on the developed automatic tear segmentation was 5.3 mm ± 5.0 mm in comparison to the interrater variability of tear retraction calculation based on manual segmentations of 3.6 mm ± 2.9 mm.
These results show that an automatic quantification of a rotator cuff tear is possible. The large interrater variability of manual segmentation-based measurements highlights the difficulty of the tear segmentations task in general.
Abstract. Fat fraction of the rotator cuff muscles has been shown to be a predictor of rotator cuff repair failure. In clinical diagnosis, fat fraction of the affected muscle is typically assessed visually on the oblique 2D Y-view and categorized according to the Goutallier scale on T1 weighted MRI. To enable a quantitative fat fraction measure of the rotator cuff muscles, an automated analysis of the whole muscle and Y-view slice was developed utilizing 2-point Dixon MRI. 3D nn-Unet were trained on water only 2-point Dixon data and corresponding annotations for the automatic segmentation of the supraspinatus, humerus and scapula and the detection of 3 anatomical landmarks for the automatic reconstruction of the Y-view slice. The supraspinatus was segmented with a Dice coefficient of 90% (N=24) and automatic fat fraction measurements with a difference from manual measurements of 1.5 % for whole muscle and 0.6% for Y-view evaluation (N=21) were observed. The presented automatic analysis demonstrates the feasibility of a 3D quantification of fat fraction of the rotator cuff muscles for the investigation of more accurate predictors of rotator cuff repair outcome.
Abstract. Introduction: Primary robotic total knee arthroplasty (TKA) is associated with favorable outcomes. To further understand robotic TKA learning curve, we evaluated early postoperative outcomes of robotics vs. manual TKA, based on surgeon experience. Methods: Patients (> 64 years) from the Medicare database, with primary, unilateral, elective TKA (“index”) from October 2015 to December 2019 were identified and categorized based on robotic vs. manual surgery, and surgeon experience: low-count surgeons had < 40 cases in the 12-months pre-index; medium-low, medium- high and high count surgeons had 41-80, 81-120 and 121-160 cases, respectively. The low-count robotic cohort (RC) was compared to the low, medium-low, medium-high, and high count manual cohort (MC) for the length of the hospital stay (LOS), and rates of home and skilled nursing facility (SNF) discharge. Descriptive statistics (means and proportion with 95% confidence intervals) were performed. Results: 296 low-count robotic cases were compared to 209,494 low-count manual and 252,905 medium-low, medium-high and high-count manual cases. The low-count RC had an average LOS of 2.03 days (95% confidence intervals (CI): 1.86-2.20) vs. 2.20 days (95%CI: 2.20-2.21) for the low-count MC. 82.4% patients (95%CI: 78.1%-86.8%) from the low-count RC were discharged home vs. 74.2% (95%CI: 74.0%-74.4%) in the low-count MC and 83.6% (95%CI: 83.3%-84.0%) in the high-count MC. Discharge to SNF affected 15.2% (95%CI: 11.1%-19.3%) in the low-count RC vs. 21.0% (95%CI: 20.9%-21.2%) and 15.2% (95%CI: 14.9%-15.4%) in the low-count and medium-high MC, respectively. Conclusion: Patients operated with robotic surgery by surgeons with low yearly volume had a LOS and probability of home discharge similar to that of patients operated with manual surgery by high-volume surgeons. Patients in the robotic group also had a lower rate of SNF discharge compared to the patients in the manual surgery group, with surgeons of similar experience.
Abstract. Fully-automatic and reliable segmentation of bone surface in volumetric ultrasound images could enable the use of this imaging technique for a variety of tasks, including diagnosis of hip dysplasia, ACL injuries in the knee as well as patient-specific instrumentation and implants in total hip or knee arthroplasty. Interpretation of volumetric data is a hard task, even for humans. In this study, we investigate the benefit of using the spatial information of a third dimension on the task of segmentation of the distal femoral bone. A data set of 52 volumetric image with 12771 image slices is split into a training and test set. We employ 2D and 3D variants of the nnUNet architecture and compare the accuracy in terms of dice coefficient and performance in terms of inference time. Note that processing of 2D data allows for a bigger model due to less memory consumption. Both architectures achieve a Dice of about 82% while the 2D variant shows less false positive segmentation and achieves a surface distance error of 0.44mm, in contrast to 0.81mm for the 3D variant. At the same time, the former infers three times faster at about 10 seconds per volume image. Apparently, model size has a bigger positive effect than the additional spatial information. Thus, we recommend considering 2D segmentation architectures even for volumetric segmentation tasks.
Abstract. Native extension and flexion joint gaps are primarily measured intraoperatively using devices such as navigation systems or tensioners, but there are advantages to being able to pre-operatively plan to such gaps. This study aims to validate the ability of a novel distracted joint gap radiology protocol to measure pre-operative extension and
flexion joint gaps. A retrospective study comprised of 42 knees was performed. Patient imaging was obtained and used to perform segmentation, landmarking and 3D-to-2D registration. The pre-operative medial and lateral joint gaps were determined in extension and flexion. Intraoperatively, a range of motion analysis was conducted using the Brainlab Knee 3 navigation system to measure the joint gaps in extension and flexion.
In extension, both medial and lateral pre-operative radiological and intraoperative navigated gaps displayed moderate and statistically significant correlations (r=0.45; p=0.003 for medial and r=0.4; p=0.01 for lateral). In flexion, only the medial radiological and navigated joint gaps correlated (r=0.54, p<0.001), with a not statistically significant trend for the lateral flexion joint gaps.
The moderate and statistically significant correlations between these joint gaps to those measured intraoperatively suggests they are reflective of on the table experience with patients. Although further work is required to understand if differences are attributable to variability in the radiological or intra-operative assessments, the pre- operative analysis technique described in this study provides the opportunity to develop a more holistic pre-operative surgical plan which considers the state of both hard and soft tissue within the joint.
Abstract. A key goal of all TKA alignment strategies is to achieve joint balance. This study aims to compare the alignments achieved by preoperatively planning to a novel distracted joint gap protocol to common alignment strategies as well as to the alignment of a healthy non-arthritic population.
A retrospective study comprised of 145 knees was performed. A long-leg supine CT scan, weightbearing AP knee X-ray and two distracted knee X-rays (one each in extension and flexion, making use of an ankle weight to open the joint) were taken pre-operatively. This imaging was used to perform segmentation, landmarking and 3D-to-2D registration. The medial and lateral joint gaps were determined in extension and flexion.
The mean weightbearing, KA planned and distracted joint planned HKA were 4.7° (±5.9°) varus, 0.3° (±3.2°) varus, and 2.2° (±3.5°) varus. This compares to a healthy adult HKA of 1.3° (±2.3°) varus. A patient level comparison between the planned KA and distracted joint HKA found that the coronal angles of the two alignments are within 3° of each other for 64% patients, within 3-5° for 26% of patients and greater than 5° for the remaining 10% of patients.
Of those compared, the planned distracted HKA was the closest to the constitutional varus HKA of a healthy population. Patient level analysis highlighted the fundamental differences between the planned KA and joint distracted alignments. By considering both hard and soft tissue, the planned joint distracted alignment allows for a more holistic foundation for pre-operative surgical planning for a given patient.
Abstract. A key goal of all TKA alignment strategies is to achieve joint balance. This study aims to compare the alignments achieved by preoperatively planning to a novel distracted joint gap protocol to common alignment strategies as well as to the alignment of a healthy non-arthritic population.
A retrospective study comprised of 145 knees was performed. A long-leg supine CT scan, weightbearing AP knee X-ray and two distracted knee X-rays (one each in extension and flexion, making use of an ankle weight to open the joint) were taken pre-operatively. This imaging was used to perform segmentation, landmarking and 3D-to-2D registration. The medial and lateral joint gaps were determined in extension and flexion.
The mean weightbearing, KA planned and distracted joint planned HKA were 4.7° (±5.9°) varus, 0.3° (±3.2°) varus, and 2.2° (±3.5°) varus. This compares to a healthy adult HKA of 1.3° (±2.3°) varus. A patient level comparison between the planned KA and distracted joint HKA found that the coronal angles of the two alignments are within 3° of each other for 64% patients, within 3-5° for 26% of patients and greater than 5° for the remaining 10% of patients.
Of those compared, the planned distracted HKA was the closest to the constitutional varus HKA of a healthy population. Patient level analysis highlighted the fundamental differences between the planned KA and joint distracted alignments. By considering both hard and soft tissue, the planned joint distracted alignment allows for a more holistic foundation for pre-operative surgical planning for a given patient.
Abstract. There is controversy regarding the effect of different approaches on recovery after THR. Collecting detailed relevant data with satisfactory compliance is difficult.
Our retrospective observational multi-center study aimed to find out if the data collected via a remote coaching app can be used to monitor the speed of recovery after THR using the anterolateral (ALA), posterior (PA) and the direct anterior approach (DAA).
771 patients undergoing THR from 13 centers using the moveUP platform were identified. 239 had ALA, 345 DAA and 42 PA. There was no significant difference between the groups in the sex of patients or in preoperative HOOS Scores. There was however a significantly lower age in the DAA (64,1y) compared to ALA (66,9y), and a significantly lower Oxford Hip Score in the DAA (23,9) compared to PA (27,7). Step count measured by an activity tracker, pain killer and NSAID use was monitored via the app. We recorded when patients started driving following surgery, stopped using crutches, and their HOOS and Oxford hip scores at 6 weeks.
Overall compliance with data request was 80%. Patients achieved their preoperative activity level after 25.8, 17,7 and 23.3 days, started driving a car after 33.6, 30.3 and 31.7 days, stopped painkillers after 27.5, 20.2 and 22.5 days, NSAID after 30.3, 25.7, and 24.7 days for ALA, DAA and PA respectively. Painkillers were stopped and preoperative activity levels were achieved significantly earlier favoring DAA over ALA. Similarly, crutches were abandoned significantly earlier (39.9, 29.7 and 24.4 days for ALA, DAA and PA respectively) favoring DAA and PA over ALA. HOOS scores and Oxford Hip scores improved significantly in all 3 groups at 6 weeks, without any statistically significant difference between groups in either Oxford Hip or HOOS subscores.
No final conclusion can be drawn as to the superiority of either approach in this study but the remote coaching platform allowed the collection of detailed data which can be used to advise patients individually, manage expectations, improve outcomes and identify areas for further research.
Abstract. The Spine Cobot System (eCential Robotics, France) is a new platform which unifies 2D/3D imaging, navigation and a robotic arm. The intent is to increase patient and surgeon safety without adding time or complexity to the surgical workflow. The primary endpoint of this cadaveric trial is to assess the precision and safety of pedicular screw positioning. The secondary endpoint is to confirm the system’s usability by the operative team. The Spine Cobot System is composed of a C-arm, a station which includes the software, an infrared camera and a collaborative robotic arm (cobot). Screw placement and neural safety were assessed. Precision of screw placement was determined by comparing the final 3D acquisition to the surgeon’s planned trajectory. Safety was quantified by 3 blinded surgeons using the Gertzbein-Robbins classification. Additionally, the usability of the integrated system for spine surgery was assessed. A system evaluation was performed in compliance with international standards (IEC, FDA). Three experienced surgeons placed 90 pedicular screws in 3 prone cadavers. 100% (90/90) of the screws were accurately placed according to the Gertzbein-Robbins classification. 97% (87/90) were classified as Grade 0 and 3% (3/90) as Grade 1. The average pilot hole middle point distance deviation is 1.3mm±0.88 mm. The average pilot hole angular deviation is 0.6°±0.6°. Only 2 usability errors were observed during the workflow assessment, and none was critical for patient safety. This preliminary study shows the efficiency of the system for pedicular screw placement, with precision and safety results. This confirms the functionality of a unified system for usability and effectiveness.
Abstract. Custom implants in Total Knee Arthroplasty (TKA) could improve prosthesis’ durability and patient’s comfort, but designing such personalized implants requires a simplified and thus automatic workflow to be easily integrated in the clinical routine. A good knowledge of the shape of the patient's femur and tibia is necessary to design it, but segmentation is still today a key issue. We present here an automatic segmentation approach of the three joints of the lower limb: hip, knee and ankle, using convolutional neural networks (CNNs) on successive transverse views from CT images. Our three 2D CNNs are built on the U-net model, and their specialization each on one joint allowed us to achieve promising results presented here. This could be integrated in a TKA planning software allowing the automatic design of TKA custom implants.
Abstract. Ultrasound (US) bone segmentation is an important component of US-guided or- thopaedic procedures. While there are many published segmentation techniques, there is no direct way to compare their performance. We present a solution to this, by curating a multi-institutional set of US images and corresponding segmentations, and systematically evaluating six previously-published bone segmentation algorithms using consistent metric definitions. We find that learning-based segmentation methods outperform traditional al- gorithms that rely on hand-crafted image features, as measured by their Dice scores, RMS distance errors and segmentation success rates. However, there is no single best performing algorithm across the datasets, emphasizing the need for carefully evaluating techniques on large, heterogenous datasets. The datasets and evaluation framework described can be used to accelerate development of new segmentation algorithms.
Abstract. Balancing soft tissues in the knee with the patella in place and with regularly applied force helps surgeons make decisions for positioning knee components in a manner that is friendly to soft tissues. A novel intraarticular device has been developed for achieving a balanced knee joint over the range of motion of the knee without requiring manual adjustments during surgery. Quasi-Constant force output was generated by the device at usual joint gaps for the knee sizes encountered during total knee arthroplasty.
Abstract. Despite the success of total knee arthroplasty (TKA), malalignment continues to be a problem which often leads to post-operative complications. The aim of this study was to investigate the accuracy of a novel, imageless, optical surgical navigation tool to assist with the alignment of femoral and tibial cuts performed during total knee arthroplasty. Six board-certified orthopedic surgeons performed TKA procedures on 9 cadavers (17 knees total), using a novel, imageless navigation system (Intellijoint KNEE, Intellijoint Surgical). Varus/valgus, femoral flexion, tibial slope, and rotation measurements from the device were compared with angular measurements calculated from post-operative computed tomography (CT) images. Navigation measurements were highly correlated with those obtained from CT scan in all three axes. For the femoral cuts, the absolute mean difference in varus/valgus was 0.83° (SD 0.46°, r = 0.76), in flexion was 1.91° (SD 1.16°, r = 0.85), and in rotation was 1.29° (SD 1.01°, r = 0.88) relative to Whiteside’s line and 0.97° (SD 0.56°, r = 0.81) relative to the posterior condylar axis. For the tibia, the absolute mean difference in varus/valgus was 1.08° (SD 0.64°, r = 0.85), anterior/posterior slope was 2.78° (SD 1.40°, r = 0.60), and rotation was 2.98° (SD 2.54°, r = 0.79). Intraoperative monitoring with the imageless navigation tool accurately measures femoral and tibial cuts in TKA and may help to increase component alignment.
Abstract. This paper describes Signature Robot, a cooperative haptic robot for knee surgery. Designed to address the lessons learned from the pioneering Acrobot Company ltd, this novel platform allows low and even impedance motion across 3 degrees of freedom, whilst the implementation of active constraints ensures patient safety throughout surgery. The robot was demonstrated to have an average positional accuracy of 0.82mm.
Abstract. Purpose
Distal radius fractures (DRF) are common types of fractures with a high incident rate. DRF can be treated either by cast or surgery. To determine the clinical procedure and the operative management, standardized guidelines have become increasingly common. As operative indications are controversial, radiographic parameters (RPs) can provide objective support for effective decision making. Calculating the RPs manually from radiographs is time consuming and subject to observer variability and clinician experience. Our aim was to develop an automatic method for accurately and reliably computing 10 RPs associated with DRF in anteroposterior (AP) and lateral radiographs of a fractured hand with and without cast.
The inputs are the AP and lateral radiographs of the fractured hand with or without cast. The outputs are 10 RP values and composite images showing the landmark points and axes used in the RPs computation on the radiographs. Our method comprises three main steps: 1) segmentation of the radius and the ulna with a deep learning radiograph pixel classifier; 2) landmark points and axis extraction from the segmentations using geometric model-based methods; 3) RPs computation from the landmarks and generation of composite images. Our study tested the accuracy of step 2.
The dataset consists of 20 pairs of AP and lateral radiographs. Ground truth radius and ulna segmentations were manually performed by an expert clinician co-author. Ground truth landmarks were manually located and annotated by the two expert clinician co-authors. The computed RP was considered accurate (in range) when its value was inside the inter and intra observer variability range of the manual annotation. The overall accuracy of the AP and lateral measurements was obtained by averaging the accuracy of each RP.
The accuracy of the computed AP RPs is 92.7%. The Radial Length and Radial Shift are within the observer variability range; for the Radial Angle, Ulnar Variance and Step all cases are within range except for one outlier; the Gap has two outlier cases. The accuracy of the computed lateral RPs is 100%: all four Palmer Tilt, Dorsal Shift, Gap, and Step are within the clinician observer variability.

Automatic computation of distal radius fractures RPs from AP and lateral radiographs of hands with and without cast can be performed accurately. Precise and consistent measurement of RPs may improve the clinical decision making process.
Abstract. AIMS: The Pixee Knee+ system offers intraoperative assistance through augmented reality glasses. This allows the surgeon to see the tibial and femoral axis depicted on the surgical field, providing real-time information during surgery.
METHODS: 122 patients received TKA surgery with the Pixee Knee+ system, and were matched based on gender and age to 122 patients who received conventional surgery. PROMs (Oxford knee Score, KOOS, and Forgotten Joint score) were collected preoperatively, at 6 weeks and 3 months. The difference between the scores at 6 weeks versus preoperative (Delta) was analyzed over time of surgery, in order to evaluate any possible surgeon learning curve.
RESULTS: Pixee patients scored significantly lower on the symptoms sub-scale of the KOOS score at 6 weeks. Similarly, at 3 months, the Quality of life sub-score, Forgotten Joint score and Oxford Knee Score were all significantly worse for the Pixee group. When analyzing the Delta KOOS over time, a clear increase in the linear model could be established for the Pixee group, whereas the Delta KOOS outcomes in the conventional group remained at a plateau.
CONCLUSION: The use of the Pixee Knee+ system results in an initial inferior clinical outcome when comparing the average of the two groups. This is likely explained by a learning curve, which shows an increase over time of the Delta KOOS at 6 weeks in the Pixee group. To what extent this increase over time will persist remains to be investigated
Abstract. A procedure with subvastus lateral approach has been utilized routinely on 60 patients, navigation was used due to the reduced exposure of this technique. Purpose of this study was to evaluate pain, function, and implant kinematics at early follow up of this surgical technique.
Tibial and femoral implant planning was based on ligament balance, gaps, and intraoperative kinematics. This approach, on pain and function, was verified at early follow- up. KSS and pain score were obtained at pre-op, 1, 3, 12 months. Data were analyzed with ANOVA for KSS and Chi-square for Pain.
No intraoperative complications were registered, no patellar tendon lesion or avulsion was noted. Preoperative average leg alignment was 4±6° varus (range 16; -14), corrected to 0° (range 2; -1). Kinematic analysis showed rollback on lateral compartment, while on medial compartment rollback was lower or negligible until 70° of flexion. Less than 5% had a “Fair” or “Poor” KSS score after 3 months. Preop pain was: 41% severe; 50% moderate; 8% mild and 0% none. At 1 month pain was: 2% severe; 18% moderate; 55% mild and 25% none. After 3 months 50% of patients had mild and 50% had no pain. This data was maintained after 1 year, with 31% of patients with mild and 69% of patients no pain (p<0.05).
This approach produced promising early outcomes in terms of pain, ROM and knee function, with less than 5% of patients presenting sub-optimal clinical results at 3- months. On symmetrical implant, medial pivot behavior was observed. Medial ligamental envelope preservation and navigated ligament balancing allow to optimize the medial stability and minimize the post-operative pain.
Abstract. The accuracy of image-based computer assisted orthopedic surgery highly depends on the accuracy of the registration step as well as image acquisition, planning and tool calibration. In this paper the accuracy of those steps is evaluated exemplarily for a robotic laminectomy.
A high-resolution test bench was designed to compare the actual location of an object and the position to which the robotic system guides the surgical tool according to the image-based plan.
Depending on the distance between the patient reference array and the tool array, average accuracies from 0.14 mm ± 0.17 mm to 0.42 mm ± 0.15 mm with a maximum error of 0.59 mm were measured.
This very high accuracy is in the range of the thickness of the spinal dura mater.
Abstract. Computer technology is ubiquitous and relied upon in virtually all professional activities. Confounding this is orthopaedic surgery where less than 5% of surgeons are using computer-assisted technologies routinely. However, the impact of Computer Assisted Orthopaedic Surgery (CAOS) may go beyond adoption in theatre.
We searched pubmed for all knee arthroplasty papers concerning knee alignment and balancing between 1976 and 2016, dividing the results into those related to CAOS and those not. Results were grouped by technology.
Between 2001 and 2008, the number of publications regarding knee navigation multiplied by 20 mainly focused on this topic of alignment and balancing, with alignment papers paralleled between navigation and non-navigation until 2010. After 2010, when navigation publications decline the number of articles related to the knee alignment and balancing without navigation increased granting the value of assessing accurately intraoperative kinematic data to improve Total Knee Arthroplasty (TKA) outcomes. From 2008, patient specific instrumentation (PSI) publications greatly increase, but navigation decreases, while robotic publications rise from 2014.
CAOS surgery publications on the search topic of alignment and balancing increased greatly between 2001 and 2018 which may suggest the impact of CAOS technology on this important knee orthopaedic forum segment.
Abstract. Computer assisted and Robotic technology in orthopaedic surgery is still not commonplace compared to un-assisted, conventional orthopaedic surgery. We analysed the relationship between patents and publications trend and question whether we could recognise a pattern which would confirm industry-driven innovation in orthopaedic surgery.
Following the same methodology used by Dalton et al. in 2016, we searched pubmed for publications between 1980 and 2018 concerning unicompartmental, patient specific instrumentation, navigation and robotic knee arthroplasty, and patents registered under the “knee arthroplasty” or “knee replacement” label over the same period. Data was plotted using 4 point moving averages.
Between 2004 and 2008, the number of publications regarding navigation multiplied by 20 following the number of patents registered during the same period. From 2008 onwards, the number of navigation publications declined while Patient Specific Instrumentation (PSI) publications increased also following patent investments from orthopaedic companies. Finally, robotic publications grew significantly pulled by massive patent registrations after 2012.
It seems that the industry has finally found a lucrative economical model after many years of trial and errors and sustained driving innovations.
Abstract. The primary objective of this study was to obtain a reliable method of automatic segmentation of shoulder muscles. The secondary objective of this study was to define a new computed tomography (CT)-based quantitative 3-dimensional (3D) measure of muscle loss (3DML) based on the rationale of the 2-dimensional (2D) qualitative Goutallier score. 102 CT scans were manually segmented and an algorithm of automated segmentation of the muscles was created. The volume of muscle fibers without intramuscular fat was then calculated for each rotator cuff muscle and normalized to the patient's scapular volume to account for the effect of body size (NVfibers). 3D muscle mass (3DMM) was calculated by dividing the NVfibers value of a given muscle by the mean expected volume in healthy shoulders. 3D muscle loss (3DML) was defined as 1 - (3DMM). Automated segmentation of the muscles was possible with a mean Dice of 0.904 ± 0.01 for the deltoid, 0.887 ± 0.014 for the infraspinatus (ISP), 0.892 ± 0.008 for the subscapularis (SSC), 0.885 for the supraspinatus (SSP) and 0.796 ± 0.006 for the teres minor (TM). The mean values of 3DFI and 3DML were 0.9% and 5.3% for Goutallier 0, 2.9% and 25.6% for Goutallier 1, 11.4% and 49.5% for Goutallier 2, 20.7% and 59.7% for Goutallier 3, and 29.3% and 70.2% for Goutallier 4, respectively. 3DML measurements obtained automatically incorporate both atrophy and fatty infiltration, thus they could become a very reliable index for assessing shoulder muscle function which could help in the decision process in shoulder surgery
Abstract. Background: Restoration of the hip-knee-ankle (HKA) angle to within 3 of the neutral mechanical axis is considered a well-aligned total knee arthroplasty (TKA), with outliers associated with higher failure rates. Thus, efforts to improve intraoperative surgical accuracy are of strong clinical interest. This study evaluated the accuracy and safety of a novel, imageless, computer-assisted navigation system (CAS) for TKA.
Methods: 112 consecutive patients who underwent primary TKA between January-December 2020 with 2 board-certified, high-volume orthopedic surgeons using the same imageless CAS were retrospectively reviewed. Patient age, BMI, sex, postoperative complications, and reoperations were collected. Two trained reviewers independently assessed tibial and femoral component mechanical alignment measurements in a standardized manner on postoperative full-leg AP and lateral radiographs. The primary outcome was mean absolute degrees of difference for each measurement compared to intraoperative CAS measurements. Outcomes were reported as means  standard deviation.
Results: 38%(N=43/112) of patients were male. Mean age was 698 years and mean BMI was 31.15.9. 71%(N=79/112) of patients had a well-aligned TKA (HKA within 3).
The mean absolute difference was 1.51.2 for femoral coronal alignment, 1.00.8 for tibial coronal alignment, 2.21.5 for femoral flexion, and 1.81.6 for tibial slope.
Two patients(1.8%) underwent reoperation; specifically, 1 patient received a 1-stage revision for periprosthetic joint infection 5 months postoperatively and the other underwent lysis of adhesions 9 months postoperatively for arthrofibrosis.
Conclusions: This novel imageless CAS provides accurate readings within 2 for tibial and femoral coronal and sagittal alignment, and patients have low complication rates at early follow-up.