To what extent can mastication functionality be restored following mandibular reconstruction surgery? A computer modeling approach.

STATEMENT OF PROBLEM: Advanced cases of head and neck cancer involving the mandible often require surgical removal of diseased sections and subsequent replacement with donor bone. During the procedure, the surgeon must make decisions regarding which bones or tissues to resect. This requires balancing tradeoffs related to issues such as surgical access and post-operative function; however, the latter is often difficult to predict, especially given that long-term functionality also depends on the impact of post-operative rehabilitation programs. PURPOSE: To assist in surgical decision-making, we present an approach for estimating the effects of reconstruction on key aspects of post-operative mandible function. MATERIAL AND METHODS: We develop dynamic biomechanical models of the reconstructed mandible considering different defect types and validate them using literature data. We use these models to estimate the degree of functionality that might be achieved following post-operative rehabilitation. RESULTS: We find significant potential for restoring mandibular functionality, even in cases involving large defects. This entails an average trajectory error below 2 mm, bite force comparable to a healthy individual, improved condyle mobility, and a muscle activation change capped at a maximum of 20%. CONCLUSION: These results suggest significant potential for adaptability in the masticatory system and improved post-operative rehabilitation, leading to greater restoration of jaw function.

Computational models and their applications in biomechanical analysis of mandibular reconstruction surgery.

Advanced head and neck cancers involving the mandible often require surgical removal of the diseased parts and replacement with donor bone or prosthesis to recreate the form and function of the premorbid mandible. The degree to which this reconstruction successfully replicates key geometric features of the original bone critically affects the cosmetic and functional outcomes of speaking, chewing, and breathing. With advancements in computational power, biomechanical modeling has emerged as a prevalent tool for predicting the functional outcomes of the masticatory system and evaluating the effectiveness of reconstruction procedures in patients undergoing mandibular reconstruction surgery. These models offer cost-effective and patient-specific treatment tailored to the needs of individuals. To underscore the significance of biomechanical modeling, we conducted a review of 66 studies that utilized computational models in the biomechanical analysis of mandibular reconstruction surgery. The majority of these studies employed finite element method (FEM) in their approach; therefore, a detailed investigation of FEM has also been provided. Additionally, we categorized these studies based on the main components analyzed, including bone flaps, plates/screws, and prostheses, as well as their design and material composition.

Design of an ultrasound-emitting drill guide for freehand pedicle screw navigation.

BACKGROUND: Accurate pedicle screw placement in spinal surgery is critical as inaccuracies can lead to morbidity and suboptimal outcomes. Navigation and robotics have reduced malplacement rates, but their adoption is limited by high costs, learning curves, surgical time, and radiation. The authors propose an ultrasound-emitting and self-localising drill guide for precise screw placement that overcomes the limitations of current techniques. MATERIALS AND METHODS: The preliminary configuration analysis involves systematically varying design parameters and assessing localization performance using lumbar spine MRI based simulations. The authors evaluate localization techniques based on accuracy and optimization capture range. RESULTS: Results suggest that feasible designs can accurately estimate position. A promising design features a 5 mm radius cannula with ten 35mm-long ultrasound strips, 32 elements per strip, and a fanned-out emission profile. A multi-start active-set optimization algorithm with six initial estimates ensures reliable and efficient localization. CONCLUSIONS: The simulation suggests that the proposed design can achieve sufficient localization accuracy for pedicle screw navigation. These findings will guide the fabrication of a novel ultrasound-emitting drill guide for further evaluation and physical testing.

Using acoustic feedback and analyses for removal of bone cement during revision knee replacement surgeries.

PURPOSE: The purpose of this study was to analyze the scraping sounds generated during revision knee replacement surgeries to discriminate between the inner cortical bone and the cement, with the goal of minimizing bone removal and increasing the structural integrity of the revision. METHODS: We prepared seven porcine femurs by partially filling them with bone cement, and recorded scraping sounds produced by a surgical scraping tool. We used a hierarchical machine learning approach to first detect a contact and then classify it as either bone or cement. This approach was based on a Support Vector Machine learning algorithm that was fed with temporal and spectral features of the sounds. A Leave-One-Bone-Out validation method was used to assess the performance of the proposed method. RESULTS: The average recall for the noncontact, bone, and cement classes was 98%, 75%, and 72%, respectively. The corresponding precision for the respective classes was 99%, 67%, and 61%. CONCLUSION: The scraping sound that is generated during revision replacement surgeries carries significant information about the material that is being scraped. Such information can be extracted using a supervised machine learning algorithm. The scraping sound produced during revision replacement procedures can potentially be used to enhance cement removal during knee revision surgery. Future work will assess whether such monitoring can increase the structural integrity of the revision.

What are the optimal targeting visualizations for performing surgical navigation of iliosacral screws? A user study.

INTRODUCTION: Complex orthopaedic procedures, such as iliosacral screw (ISS) fixations, can take advantage of surgical navigation technology to achieve accurate results. Although the impact of surgical navigation on outcomes has been studied, no studies to date have quantified how the design of the targeting display used for navigation affects ISS targeting performance. However, it is known in other contexts that how task information is displayed can have significant effects on both accuracy and time required to perform motor tasks, and that this can be different among users with different experience levels. This study aimed to investigate which visualization techniques helped experienced surgeons and inexperienced users most efficiently and accurately align a surgical tool to a target axis. METHODS: We recruited 21 participants and conducted a user study to investigate five proposed 2D visualizations (bullseye, rotated bullseye, target-fixed, tool-fixed in translation, and tool-fixed in translation and rotation) with varying representations of the ISS targets and tool, and one 3D visualization. We measured the targeting accuracy achieved by each participant, as well as the time required to perform the task using each of the visualizations. RESULTS: We found that all 2D visualizations had equivalent translational and rotational errors, with mean translational errors below 0.9 mm and rotational errors below 1.1[Formula: see text]. The 3D visualization had statistically greater mean translational and rotational errors (4.29 mm and 5.47[Formula: see text], p < 0.001) across all users. We also found that the 2D bullseye view allowed users to complete the simulated task most efficiently (mean 30.2 s; 95% CI 26.4-35.7 s), even when combined with other visualizations. CONCLUSIONS: Our results show that 2D bullseye views helped both experienced orthopaedic trauma surgeons and inexperienced users target iliosacral screws accurately and efficiently. These findings could inform the design of visualizations for use in a surgical navigation system for screw insertions for both training and surgical practice.

Can Technology Assistance be Cost Effective in TKA? A Simulation-Based Analysis of a Risk-prioritized, Practice-specific Framework.

BACKGROUND: Robotic, navigated, and patient-specific instrumentation (PSI) TKA procedures have been introduced to improve component placement precision and improve implant survivorship and other clinical outcomes. However, the best available evidence has shown that these technologies are ineffective in reducing revision rates in the general TKA patient population. Nonetheless, it seems plausible that these technologies could be an effective and cost-effective means of reducing revision risk in clinical populations that are at an elevated risk of revision because of patient-specific demographics (such as older age at index surgery, elevated BMI, and being a man). Since clinical trials on this topic would need to be very large, a simulation approach could provide insight on which clinical populations would be the most promising for analysis. QUESTIONS/PURPOSES: We conducted a simulation-based analysis and asked: (1) Given key demographic parameters characterizing a patient population, together with estimates of the precision achievable with selected forms of technology assistance in TKA, can we estimate the expected distributions of anticipated reductions in lifetime revision risk for that population and the associated improvements in quality-adjusted life years (QALYs) that would be expected to result? (2) Are there realistic practice characteristics (such as combinations of local patient demographics and capital and per-procedure costs) for which applying a per-patient risk-prioritized policy for using technology-assisted TKA could be considered cost-effective based on projected cost savings from reductions in revision rates? METHODS: We designed simulations of hypothetical practice-specific clinical scenarios, each characterized by patient volume, patient demographics, and technology-assisted surgical technique, using demographic information drawn from other studies to characterize two contrasting simulated clinical scenarios in which the distributions of factors describing patients undergoing TKA place one population at a comparatively elevated risk of revision (elevated-risk population) and the second at a comparatively reduced risk of revision (lower-risk population). We used results from previous systematic reviews and meta-analyses to estimate the implant precision in coronal plane alignment for patient-specific instrumentation, navigated, and robotic technology. We generated simulated TKA patient populations based on risk estimates from large clinical studies, structured reviews, and meta-analyses and calculated the patient-specific reduction in the revision risk and the change in QALYs attributable to the technology-assisted intervention in each of the two simulated clinical scenarios. We also incorporated a sensitivity analysis, incorporating variations in the effect size of deviations from overall coronal alignment on revision risk and difference in health state utilities acquired through a structured review process. We then simulated the outcomes of 25,000 operations per patient using the precisions associated with the conventional TKA technique, the three technology-assisted techniques, and a hypothetical technology-assisted intervention that could consistently deliver perfectly neutral overall coronal alignment, which is unachievable in practice. A risk-prioritized treatment policy was emulated by ordering the simulated patients from the highest to lowest predicted increase in QALYs, such that simulated patients who would see the greatest increase in the QALYs (and therefore the greatest reduction in lifetime revision risk) were the patients to receive technology-assisted TKA intervention in a practice. We used cost estimates acquired through a structured review process and calculated the net added costs of each of the three technology-assisted techniques as a function of the percent utilization (proportion of patients treated with technology assistance in a practice), factoring in fixed costs, per-procedure variable costs, and savings occurring from the prevention of future revision surgery. Finally, we calculated the incremental cost-effectiveness ratio (ICER) and marginal cost-effectiveness ratio (MCER) for each technology-assisted technique for the two clinical scenarios. We then used a Monte Carlo approach to simulate variations in key patient risk, health state, and economic factors as well as to obtain a distribution of estimates for cost-effectiveness. We considered an intervention to be cost effective if either the ICER or MCER values were below USD/QALY 63,000. RESULTS: For the lower-risk population, the median reduction in the revision risk was 0.9% (0.4% to 2.2%, extrema from the sensitivity analysis) and 1.8% (0.9% to 4.4%) for PSI and robotic TKA, respectively, and 1.9% (1.0% to 4.6%) for ideal TKA. In contrast, the median reduction in the revision risk in the elevated-risk clinical scenario was 2.0% (1.2% to 3.4%) and 4.6% (2.7% to 8.5%) for PSI and robotic TKA and 5.1% (3.0% to 9.4%) for ideal TKA. Estimated differences in the cumulative gain in QALYs attributable to technology-assisted TKA ranged from 0.6 (0.2 to 1.8) to 4.0 (1.8 to 10.0) QALYs per 100 patients, depending on the intervention type and clinical scenario. For PSI, we found treating 15% of patients in the lower-risk population and 77% in the elevated-risk population could meet the threshold for being considered cost effective. For navigated TKA systems offering high alignment precision, we found the intervention could meet this threshold for practice sizes of at least 300 patients per year and a percent utilization of 27% in the lower-risk population. In the elevated-risk population, cost-effectiveness could be achieved in practice volumes as small as 100 patients per year with a percent utilization of at least 6%, and cost savings could be achieved with a percent utilization of at least 45%. We found that robotic TKA could only meet the threshold for being considered cost-effectiveness in the lower-risk population if yearly patient volumes exceeded 600 and for a limited range of percent utilization (27% to 32%). However, in the elevated-risk patient population, robotic TKA with high alignment precision could potentially be cost effective for practice sizes as small as 100 patients per year and a percent utilization of at least 20% if a risk-prioritized treatment protocol were used. CONCLUSION: Based on these simulations, a selective-use policy for technology-assisted TKA that prioritizes using technology assistance for those patients at a higher risk of revision based on patient-specific factors could potentially meet the cost-effectiveness threshold in selected circumstances (for example, primarily in elevated-risk populations and larger practice sizes). Whether it does meet that threshold would depend significantly on the surgical precision that can be achieved in practice for a given proposed technology as well as on the true local costs of using the proposed technology. We further recommend that any future randomized trials seeking to demonstrate possible effects of technology assistance on revision risk focus on clinical populations that are at higher risk of revision (such as, patient populations that are relatively younger, have higher BMIs, and higher proportions of men). CLINICAL RELEVANCE: This study suggests that technology assistance is only likely to prove cost effective in selected circumstances rather than in all clinical populations and practice settings. In general, we project that surgical navigation is most likely to prove cost effective in the widest range of circumstances, that PSI may be cost effective or cost neutral in a moderate range of circumstances, and that robotic surgery is only likely to be cost effective in moderately large practices containing patients who are on average at an intrinsically elevated risk of revision.

Monitoring Arm Movements Post-Stroke for Applications in Rehabilitation and Home Settings.

Optimal recovery of arm function following stroke requires patients to perform a large number of functional arm movements in clinical therapy sessions, as well as at home. Technology to monitor adherence to this activity would be helpful to patients and clinicians. Current approaches to monitoring arm movements are limited because of challenges in distinguishing between functional and non-functional movements. Here, we present an Arm Rehabilitation Monitor (ARM), a device intended to make such measurements in an unobtrusive manner. The ARM device is based on a single Inertial Measurement Unit (IMU) worn on the wrist and uses machine learning techniques to interpret the resulting signals. We characterized the ability of the ARM to detect reaching actions in a functional assessment dataset (functional assessment tasks) and an Activities-of-Daily-Living (ADL) dataset (pizza-making and walking task) from 12 participants with stroke. The Convolutional Neural Network (CNN) and Random Forests (RF) classifiers had a Matthews Correlation Coefficient score of 0.59 and 0.58 when trained and tested on the functional dataset, 0.50 and 0.49 when trained and tested on the ADL dataset, and 0.37 and 0.36 when trained on the functional dataset and tested on the ADL dataset, respectively. The latter is the most relevant scenario for the intended application of training during a clinical visit for monitoring movements in the in-home setting. The classifiers showed good performance in estimating the time spent reaching and number of reaching gestures and showed low sensitivity to irrelevant arm movements produced during walking. We conclude that the ARM has sufficient accuracy and robustness to merit being used in preliminary studies to monitor arm activity in rehabilitation or home applications.

Verifying a C-arm-based roentgen stereophotogrammetric analysis protocol for assessing tibial implant movement in total knee arthroplasty.

Roentgen stereophotogrammetric analysis (RSA) is the "gold standard" technique for measuring sub-millimetric relative motion between implant and bone to quantify post-operative implant migration over time. The vast majority of RSA studies addressing implant motion in knee replacements, however, have been conducted using expensive biplanar radiography systems and commercial software that are not readily available at many institutions. In this study, we evaluated the feasibility of performing RSA using ordinary, readily available C-arm fluoroscopes and open-source software to assess tibial component migration.We developed an assessment protocol using a Siemens Arcadis Orbic C-arm and the open-source XROMM software and evaluated its accuracy and precision through a series of phantom-based verification tests. The results were highly promising: accuracies were in the range of - 39 to 11 µm for translations and - 0.025 to 0.029° for rotations, while system precisions ranged between 16 to 27 µm for translations and 0.041 to 0.059° for rotations. This performance is comparable to specialized RSA systems reported in the literature. The proposed RSA protocol is therefore capable of accurately measuring the relative motion of knee replacement implants in phantom scenarios, which justifies further the development of the protocol towards use in prospective clinical assessments of new implant designs and surgical techniques.

Can uncertainty estimation predict segmentation performance in ultrasound bone imaging?

PURPOSE: Segmenting bone surfaces in ultrasound (US) is a fundamental step in US-based computer-assisted orthopaedic surgeries. Neural network-based segmentation techniques are a natural choice for this, given promising results in related tasks. However, to gain widespread use, we must be able to know how much to trust segmentation networks during clinical deployment when ground-truth data is unavailable. METHODS: We investigated alternative ways to measure the uncertainty of trained networks by implementing a baseline U-Net trained on a large dataset, together with three uncertainty estimation modifications: Monte Carlo dropout, test time augmentation, and ensemble learning. We measured the segmentation performance, calibration quality, and the ability to predict segmentation performance on test data. We further investigated the effect of data quality on these measures. RESULTS: Overall, we found that ensemble learning with binary cross-entropy (BCE) loss achieved the best segmentation performance (mean Dice: 0.75-0.78 and RMS distance: 0.62-0.86mm) and the lowest calibration errors (mean: 0.22-0.28%). In contrast to previous studies of area or volumetric segmentation, we found that the resulting uncertainty measures are not reliable proxies for surface segmentation performance. CONCLUSION: Our experiments indicate that a significant performance and confidence calibration boost can be achieved with ensemble learning and BCE loss, as tested on 13,687 US images containing various anatomies and imaging parameters. However, these techniques do not allow us to reliably predict future segmentation performance. The results of this study can be used to improve the calibration and performance of US segmentation networks.

Targeting repeatability of a less obtrusive surgical navigation procedure for total shoulder arthroplasty.

PURPOSE: Surgical navigation systems have demonstrated improvements in alignment accuracy in a number of arthroplasty procedures, but they have not yet been widely adopted for use in total shoulder arthroplasty (TSA). We believe this is due in part to the obtrusiveness of conventional optical tracking systems, as well as the need for additional intraoperative steps such as calibration and registration. The purpose of this study is to evaluate the feasibility of adapting a less-intrusive dental navigation system for use in TSA. METHODS: We developed a proof-of-concept system based on validated laser-engraved surgical tools recently introduced for use in dental surgery that are calibrated once when manufactured and not recalibrated at time of use. The design also features a notably smaller bone-mounted tracker that can be tracked from a wide range of viewing angles. To assess our system's performance, we modified the dental surgical software to support guidance of a TSA procedure. We then conducted a user study in which three participants with varying surgical experience used the system to drill 30 holes in a glenoid model. Using a coordinate measuring machine, we determined the resulting drilled trajectory and compared this to the pre-planned trajectory. Since we used a model glenoid rather than anatomical specimens, we report on targeting precision rather than overall procedure precision or accuracy. RESULTS: We found targeting precision < 1 mm (standard deviation) for locating the entry hole and < ~ 1° (SD) for both version and inclination. The latter value was markedly lower than the end-to-end angular precision achieved by previously reported TSA navigation systems (approximately 3°-5° SD). CONCLUSION: We conclude that variability during the targeting phase represents a small fraction of the overall variability exhibited by existing systems, so a less obtrusive navigation system for TSA based on laser-engraved tooling is likely feasible, which could improve the uptake rates of surgical navigation for TSA, thereby potentially leading to improved overall surgical outcomes.

The Distal Free Achilles Tendon Is Longer in People with Tendinopathy than in Controls: A Retrospective Case-Control Study.

Objectives: The free Achilles tendon is defined as the region of tendon distal to the soleus which is "unbuttressed," i.e., unsupported by muscular tissue. We reasoned that a relative lack of distal buttressing could place the tendon at a greater risk for developing Achilles tendinopathy. Therefore, our primary goal was to compare the free Achilles tendon length between those with midportion or insertional Achilles tendinopathy and healthy controls. Design: This is a retrospective case-control study. Setting. Hospital in Vancouver, Canada. Participants. 66 cases with Achilles tendinopathy (25 insertional, 41 midportion) consecutively drawn from a hospital database within a 5-year period and matched to 66 controls (without tendinopathy) based on sex, age, and weight. Main outcome measures. Odds ratio of the risk of developing Achilles tendinopathy given the length of free tendon, defined anatomically on MRI, after adjustment for confounders. Results: MRI-defined free Achilles tendon length is a statistically significant predictor of having midportion Achilles tendinopathy (odds ratio = 0.53, 95% confidence interval 1.13 to 2.07). Midportion Achilles tendinopathy cases had significantly longer free tendons (Mdn = 51.2 mm, IQR = 26.9 mm) compared to controls (Mdn = 40.8 mm, IQR = 20.0 mm), p = 0.007. However, there was no significant difference between the free Achilles tendon lengths in insertional AT cases (Mdn = 47.9 mm, IQR = 15.1 mm) and controls (Mdn = 39.2 mm, IQR = 17.9 mm), p = 0.158. Free Achilles tendon length was also correlated with the tendon thickness among those with Achilles tendinopathy, rτ = 0.25, and p = 0.003. Conclusions: The MRI-defined length of the free Achilles tendon is positively associated with the risk of midportion Achilles tendinopathy. A relative lack of distal muscular buttressing of the Achilles tendon may therefore influence the development of tendinopathy.

Clinicians' perceptions of a potential wearable device for capturing upper limb activity post-stroke: a qualitative focus group study.

BACKGROUND: There is growing interest in the use of wearable devices that track upper limb activity after stroke to help determine and motivate the optimal dose of upper limb practice. The purpose of this study was to explore clinicians' perceptions of a prospective wearable device that captures upper limb activity to assist in the design of devices for use in rehabilitation practice. METHODS: Four focus groups with 18 clinicians (occupational and physical therapists with stroke practice experience from a hospital or private practice setting) were conducted. Data were analyzed thematically. RESULTS: Our analysis revealed three themes: (1) "Quantity and quality is ideal" emphasized how an ideal device would capture both quantity and quality of movement; (2) "Most useful outside therapy sessions" described how therapists foresaw using the device outside of therapy sessions to monitor homework adherence, provide self-monitoring of use, motivate greater use and provide biofeedback on movement quality; (3) "User-friendly please" advocated for the creation of a device that was easy to use and customizable, which reflected the client-centered nature of their treatment. CONCLUSIONS: Findings from this study suggest that clinicians support the development of wearable devices that capture upper limb activity outside of therapy for individuals with some reach to grasp ability. Devices that are easy to use and capture both quality and quantity may result in greater uptake in the clinical setting. Future studies examining acceptability of wearable devices for tracking upper limb activity from the perspective of individuals with stroke are needed.

How Large a Study Is Needed to Detect TKA Revision Rate Reductions Attributable to Robotic or Navigated Technologies? A Simulation-based Power Analysis.

BACKGROUND: Robotic and navigated TKA procedures have been introduced to improve component placement precision in the hope of improving implant survivorship and other clinical outcomes. Although numerous comparative studies have shown enhanced precision and accuracy in placing components, most comparative studies have not shown that such interventions result in improved implant survival. Given what we know about effect sizes from large arthroplasty registries, large cohort studies, and large randomized controlled trials (RCTs), we wondered how large randomized trials would need to be to detect such small differences, and if the number is very high, what that would tell us about the value of these treatments for preventing revision surgery. QUESTIONS/PURPOSES: In this simulation study, we asked: Given that survivorship differences between technology-assisted TKA (TA-TKA, which we defined as either navigated or robot-assisted TKA) and conventional TKA are either small or absent based on large arthroplasty registries, large cohort studies, and large RCTs, how large would randomized trials need to be to detect small differences between TA-TKA and conventional TKA if they exist, and how long would the follow-up period need to be to have a reasonable chance to detect those differences? METHODS: We used estimated effect sizes drawn from previous clinical and registry studies, combined with estimates of the accuracy and precision of various navigation and robotic systems, to model and simulate the likely outcomes of potential comparative clinical study designs. To characterize the ranges of patients enrolled and general follow-up times associated with traditional RCT studies, we conducted a structured search of previously published studies evaluating the effect of robotics and navigation on revision rates compared with that of conventional TKA. The structured search of the University of British Columbia's library database (which automatically searches medical publication databases such as PubMed, Embase, Medline, and Web of Science) and subsequent searching through included studies' reference lists yielded 103 search results. Only clinical studies assessing implant survival differences between patient cohorts of TA-TKA and conventional TKA were included. Studies analyzing registry data, using cadaver specimens, assessing revision TKA, conference proceedings, and preprint services were excluded. Twenty studies met all our inclusion criteria, but only one study reported a statistically significant difference between the conventional and robotic or navigated groups. Next, we generated a large set of patients with simulated TKA (1.5 million), randomly assigning each simulated patient a set of patient-specific factors (age at the index surgery, gender, and BMI) drawn from data from registries and published information. We divided this set of simulated procedures into four groups, each associated with a coronal alignment precision reported for different types of surgical procedures, and randomly assigned each patient an overall coronal alignment consistent with their group's precision. TA procedures were modeled based on the alignment precision that an intervention could deliver, regardless of whether the technology used was navigation- or robot-assisted. To evaluate the power associated with using different cohort sizes, we ran a Monte Carlo simulation generating 3000 simulated populations that were drawn (with replacement) from the large set of simulated patients with TKA. We simulated the time to revision for aseptic loosening for each patient, computed the corresponding Kaplan-Meier survival curves, and applied a log-rank test to each study for statistical differences in revision rates at concurrent follow-up timepoints (1-25 years). From each simulation associated with a given cohort size, we determined the percentage of simulated studies that found a statistically significant difference at each follow-up interval. For each alternative precision, we then also calculated the expected reduction in revision rates (effect size) attributable to TA-TKA intervention and the number needed to treat (NNT) using TA-TKA to prevent one revision at 2, 5, 10, and 15 years after index surgery for the entire set of Kaplan-Meier survival analyses. RESULTS: The results from our simulation found survivorship differences favoring TA-TKA ranging from 1.4% to 2.0% at 15 years of follow-up. Comparative studies would need to enroll between 2500 and 4000 patients in each arm of the study, depending on the precision of the navigated or robotic procedure, to have an 80% chance of showing this reduction in revision rates at 15 years of follow-up. For the highest precision simulated intervention, the NNT using TA-TKA to prevent one revision was 1000 at 2 years, 334 at 5 years, 100 at 10 years, and 50 at 15 years post-index surgery. CONCLUSION: Based on these simulations, it appears that TA-TKA interventions could potentially result in a relative reduction in revision rates as large as 27% (from 7.5% down to about 5.5% at 15 years for the intervention with the most precise coronal alignment); however, since this 2% absolute reduction in revision rates is relatively small in comparison with the baseline success rate of TKA and would not be realized until 15 years after the index surgery, traditional RCT studies would require excessively large numbers of patients to be enrolled and excessively long follow-up times to demonstrate whether such a reduction actually exists. CLINICAL RELEVANCE: Given that the NNTs to avoid revisions at various time points are predicted to be high, it would require correspondingly low system costs to justify broad adoption of TA-TKA based on avoided revision costs alone, though we speculate that technology assistance could perhaps prove to be cost effective in the care of patients who are at an elevated risk of revision.

Automatically Delineating Key Anatomy in 3-D Ultrasound Volumes for Hip Dysplasia Screening.

Developmental dysplasia of the hip (DDH) metrics based on 3-D ultrasound have proven more reliable than those based on 2-D images, but to date have been based mainly on hand-engineered features. Here, we test the performance of 3-D convolutional neural networks for automatically segmenting and delineating the key anatomical structures used to define DDH metrics: the pelvis bone surface and the femoral head. Our models are trained and tested on a data set of 136 volumes from 34 participants. For the pelvis, a 3D-U-Net achieves a Dice score of 85%, outperforming the confidence-weighted structured phase symmetry algorithm (Dice score = 19%). For the femoral head, the 3D-U-Net had centre and radius errors of 1.42 and 0.46 mm, respectively, outperforming the random forest classifier (3.90 and 2.01 mm). The improved segmentation may improve DDH measurement accuracy and reliability, which could reduce misdiagnosis.

Leveraging voxel-wise segmentation uncertainty to improve reliability in assessment of paediatric dysplasia of the hip.

PURPOSE: Estimating uncertainty in predictions made by neural networks is critically important for increasing the trust medical experts have in automatic data analysis results. In segmentation tasks, quantifying levels of confidence can provide meaningful additional information to aid clinical decision making. In recent work, we proposed an interpretable uncertainty measure to aid clinicians in assessing the reliability of developmental dysplasia of the hip metrics measured from 3D ultrasound screening scans, as well as that of the US scan itself. In this work, we propose a technique to quantify confidence in the associated segmentation process that incorporates voxel-wise uncertainty into the binary loss function used in the training regime, which encourages the network to concentrate its training effort on its least certain predictions. METHODS: We propose using a Bayesian-based technique to quantify 3D segmentation uncertainty by modifying the loss function within an encoder-decoder type voxel labeling deep network. By appending a voxel-wise uncertainty measure, our modified loss helps the network improve prediction uncertainty for voxels that are harder to train. We validate our approach by training a Bayesian 3D U-Net with the proposed modified loss function on a dataset comprising 92 clinical 3D US neonate scans and test on a separate hold-out dataset of 24 patients. RESULTS: Quantitatively, we show that the Dice score of ilium and acetabulum segmentation improves by 5% when trained with our proposed voxel-wise uncertainty loss compared to training with standard cross-entropy loss. Qualitatively, we further demonstrate how our modified loss function results in meaningful reduction of voxel-wise segmentation uncertainty estimates, with the network making more confident accurate predictions. CONCLUSION: We proposed a Bayesian technique to encode voxel-wise segmentation uncertainty information into deep neural network optimization, and demonstrated how it can be leveraged into meaningful confidence measures to improve the model's predictive performance.

Deep learning-based X-ray inpainting for improving spinal 2D-3D registration.

BACKGROUND: Two-dimensional (2D)-3D registration is challenging in the presence of implant projections on intraoperative images, which can limit the registration capture range. Here, we investigate the use of deep-learning-based inpainting for removing implant projections from the X-rays to improve the registration performance. METHODS: We trained deep-learning-based inpainting models that can fill in the implant projections on X-rays. Clinical datasets were collected to evaluate the inpainting based on six image similarity measures. The effect of X-ray inpainting on capture range of 2D-3D registration was also evaluated. RESULTS: The X-ray inpainting significantly improved the similarity between the inpainted images and the ground truth. When applying inpainting before the 2D-3D registration process, we demonstrated significant recovery of the capture range by up to 85%. CONCLUSION: Applying deep-learning-based inpainting on X-ray images masked by implants can markedly improve the capture range of the associated 2D-3D registration task.

3-D Ultrasound Imaging Reliability of Measuring Dysplasia Metrics in Infants.

Developmental dysplasia of the hip is a hip abnormality that ranges from mild acetabular dysplasia to irreducible femoral head dislocations. While 2-D B-mode ultrasound (US)-based dysplasia metrics or disease metrics are currently used clinically to diagnose developmental dysplasia of the hip, such estimates suffer from high inter-exam variability. In this work, we propose and evaluate 3-D US-derived dysplasia metrics that are automatically computed and demonstrate that these automatically derived dysplasia metrics are considerably more reproducible. The key features of our automatic method are (i) a random forest-based learning technique to remove regions across the coronal axis that do not contain bone structures necessary for dysplasia-metric extraction, thereby reducing outliers; (ii) a bone segmentation method that uses rotation-invariant and intensity-invariant filters, thus remaining robust to signal dropout and varying bone morphology; (iii) a novel slice-based learning and 3-D reconstruction strategy to estimate a probability map of the hypoechoic femoral head in the US volume; and (iv) formulae for calculating the 3-D US-derived dysplasia metrics. We validate our proposed method on real clinical data acquired from 40 infant hip examinations. Results show a considerable (around 70%) reduction in variability in two key 3-D US-derived dysplasia metrics compared with their 2-D counterparts.