Explore the vast range of titles available.

Purpose The application of robotics in the operating theatre for total knee arthroplasty (TKA) remains controversial. As with all new technology, the introduction of new systems is associated with a learning curve and potentially associated with extra complications. Therefore, the aim of this study is to identify and predict the learning curve of robot-assisted (RA) TKA. Methods A RA TKA system (MAKO) was introduced in April 2018 in our service. A retrospective analysis was performed of all patients receiving a TKA with this system by six surgeons. Operative times, implant and limb alignment, intraoperative joint balance and robot-related complications were evaluated. Cumulative summation (CUSUM) analyses were used to assess learning curves for operative time, implant alignment and joint balance in RA TKA. Linear regression was performed to predict the learning curve of each surgeon. Results RA TKA was associated with a learning curve of 11–43 cases for operative time ( p  < 0.001). This learning curve was significantly affected by the surgical profile (high vs. medium vs. low volume). A complete normalisation of operative times was seen in four out of five surgeons. The precision of implant positioning and gap balancing showed no learning curve. An average deviation of 0.2° (SD 1.4), 0.7° (SD 1.1), 1.2 (SD 2.1), 0.2° (SD 2.9) and 0.3 (SD 2.4) for the mLDFA, MPTA, HKA, PDFA and PPTA from the preoperative plan was observed. Limb alignment showed a mean deviation of 1.2° (SD 2.1) towards valgus postoperatively compared to the intraoperative plan. One tibial stress fracture was seen as a complication due to suboptimal positioning of the registration pins. Conclusion RA TKA is associated with a learning curve for surgical time, which might be longer than reported in current literature and dependent on the profile of the surgeon. There is no learning curve for component alignment, limb alignment and gap balancing. Level of evidence IV.

Purpose The primary objective of this study was to determine the surgical team’s learning curve for robotic-arm assisted TKA through assessments of operative times, surgical team comfort levels, accuracy of implant positioning, limb alignment, and postoperative complications. Secondary objectives were to compare accuracy of implant positioning and limb alignment in conventional jig-based TKA versus robotic-arm assisted TKA. Methods This prospective cohort study included 60 consecutive conventional jig-based TKAs followed by 60 consecutive robotic-arm assisted TKAs performed by a single surgeon. Independent observers recorded surrogate markers of the learning curve including operative times, stress levels amongst the surgical team using the state-trait anxiety inventory (STAI) questionnaire, accuracy of implant positioning, limb alignment, and complications within 30 days of surgery. Cumulative summation (CUSUM) analyses were used to assess learning curves for operative time and STAI scores in robotic TKA. Results Robotic-arm assisted TKA was associated with a learning curve of seven cases for operative times ( p  = 0.01) and surgical team anxiety levels ( p  = 0.02). Cumulative robotic experience did not affect accuracy of implant positioning (n.s.) limb alignment (n.s.) posterior condylar offset ratio (n.s.) posterior tibial slope (n.s.) and joint line restoration (n.s.). Robotic TKA improved accuracy of implant positioning ( p  < 0.001) and limb alignment ( p  < 0.001) with no additional risk of postoperative complications compared to conventional manual TKA. Conclusion Implementation of robotic-arm assisted TKA led to increased operative times and heightened levels of anxiety amongst the surgical team for the initial seven cases but there was no learning curve for achieving the planned implant positioning. Robotic-arm assisted TKA improved accuracy of implant positioning and limb alignment compared to conventional jig-based TKA. The findings of this study will enable clinicians and healthcare professionals to better understand the impact of implementing robotic TKA on the surgical workflow, assist the safe integration of this procedure into surgical practice, and facilitate theatre planning and scheduling of operative cases during the learning phase. Level of evidence II.

IntroductionThere has been increased interest in assessing the surgeon learning curve for new skill acquisition. While there is no consensus around the best methodology, one of the most frequently used learning curve assessments in the surgical literature is the cumulative sum curve (CUSUM) of operative time. To demonstrate the limitations of this methodology, we assessed the CUSUM of console time across cohorts of surgeons with differing case acquisition rates while varying the total number of cases used to calculate the CUSUM.MethodsWe compared the CUSUM curves of the average console times of surgeons who completed their first 20 robotic-assisted (RAS) cases in 13, 26, 39, and 52 weeks, respectively, for their first 50 and 100 cases, respectively. This analysis was performed for prostatectomy (1094 surgeons), malignant hysterectomy (737 surgeons), and inguinal hernia (1486 surgeons).ResultsIn all procedures, the CUSUM curve of the cohort of surgeons who completed their first 20 procedures in 13 weeks demonstrated a lower slope than cohorts of surgeons with slower case acquisition rates. The case number at which the peak of the CUSUM curve occurs uniformly increases when the total number of cases used in generation of the CUSUM chart changes from 50 to 100 cases.ConclusionThe CUSUM analyses of these three procedures suggests that surgeons with fast initial case acquisition rates have less variability in their operative times over the course of their learning curve. The peak of the CUSUM curve, which is often used in surgical learning curve literature to denote “proficiency” is predictably influenced by the total number of procedures evaluated, suggesting that defining the peak as the point at which a surgeon has overcome the learning curve is subject to routine bias. The CUSUM peak, by itself, is an insufficient measure of “conquering the learning curve.”

The performance of a classification algorithm eventually reaches a point of diminishing returns, where the additional sample added does not improve the results. Thus, there is a need to determine an optimal sample size that maximizes performance while accounting for computational burden or budgetary concerns. This study aimed to determine optimal sample sizes and the relationships between sample size and dataset-level characteristics over a variety of binary classification algorithms. A total of 16 large open-source datasets were collected, each containing a binary clinical outcome. Furthermore, 4 machine learning algorithms were assessed: XGBoost (XGB), random forest (RF), logistic regression (LR), and neural networks (NNs). For each dataset, the cross-validated area under the curve (AUC) was calculated at increasing sample sizes, and learning curves were fit. Sample sizes needed to reach the observed full-dataset AUC minus 2 points (0.02) were calculated from the fitted learning curves and compared across the datasets and algorithms. Dataset-level characteristics, minority class proportion, full-dataset AUC, number of features, type of features, and degree of nonlinearity were examined. Negative binomial regression models were used to quantify relationships between these characteristics and expected sample sizes within each algorithm. A total of 4 multivariable models were constructed, which selected the best-fitting combination of dataset-level characteristics. Among the 16 datasets (full-dataset sample sizes ranging from 70,000-1,000,000), median sample sizes were 9960 (XGB), 3404 (RF), 696 (LR), and 12,298 (NN) to reach AUC stability. For all 4 algorithms, more balanced classes (multiplier: 0.93-0.96 for a 1% increase in minority class proportion) were associated with decreased sample size. Other characteristics varied in importance across algorithms-in general, more features, weaker features, and more complex relationships between the predictors and the response increased expected sample sizes. In multivariable analysis, the top selected predictors were minority class proportion among all 4 algorithms assessed, full-dataset AUC (XGB, RF, and NN), and dataset nonlinearity (XGB, RF, and NN). For LR, the top predictors were minority class proportion, percentage of strong linear features, and number of features. Final multivariable sample size models had high goodness-of-fit, with dataset-level predictors explaining a majority (66.5%-84.5%) of the total deviance in the data among all 4 models. The sample sizes needed to reach AUC stability among 4 popular classification algorithms vary by dataset and method and are associated with dataset-level characteristics that can be influenced or estimated before the start of a research study.

Objective The number of colonoscopies required to reach competency is not well established. The primary aim of this study was to determine the number of colonoscopies trainees need to perform to attain competency, defined by a caecal intubation rate (CIR) ≥90%. As competency depends on completion, we also investigated trainee factors that were associated with colonoscopy completion. Design The Joint Advisory Group on GI Endoscopy in the UK has developed a trainee e-portfolio from which colonoscopy data were retrieved. Inclusion criteria were all trainees who had performed a total of ≥20 colonoscopies and had performed ≤50 colonoscopies prior to submission of data to the e-portfolio. The primary outcome measure was colonoscopy completion. The number of colonoscopies required to achieve CIR ≥90% was calculated by the moving average method and learning curve cumulative summation (LC-Cusum) analysis. To determine factors which determine colonoscopy completion, a mixed effect logistic regression model was developed which allowed for nesting of patients within trainees and nesting of patients within hospitals, with various patient, trainee and training factors entered as fixed effects. Results 297 trainees undertook 36 730 colonoscopies. By moving average analysis, the cohort of trainees reached a CIR of 90% at 233 procedures. By LC-Cusum analysis, 41% of trainees were competent after 200 procedures. Of the trainee factors, the number of colonoscopies, intensity of training and previous flexible sigmoidoscopy experience were significant factors associated with colonoscopy completion. Conclusions This is the largest study to date investigating the number of procedures required to achieve competency in colonoscopy. The current training certification benchmark in the UK of 200 procedures does not appear to be an inappropriate minimum requirement. The LC-Cusum chart provides real time feedback on individual learning curves for trainees. The association of training intensity and flexible sigmoidoscopy experience with colonoscopy completion could be exploited in training programmes.

Purpose Patient-specific cutting guides (PSCGs) have been advocated to improve the accuracy of deformity correction in opening-wedge high-tibial osteotomies (HTO). It was hypothesized that PSCGs for HTO would have a short learning curve. Therefore, the goals of this study were to determine the surgeons learning curve for PSCGs used for opening-wedge HTO assessing: the operating time, surgeons comfort levels, number of fluoroscopic images, accuracy of post-operative limb alignment and functional outcomes. Methods This prospective cohort study included 71 consecutive opening-wedge HTO with PSCGs performed by three different surgeons with different experiences. The operating time, the surgeon’s anxiety levels evaluated using the Spielberger State-Trait Anxiety Inventory (STAI), the number of fluoroscopic images was systematically and prospectively collected. The accuracy of the postoperative alignment was defined by the difference between the preoperative targeted correction and the final post-operative correction both measured on standardized CT-scans using the same protocol (ΔHKA, ΔMPTA, ΔPPTA). Functional outcomes were evaluated at 1 year using the different sub-scores of the KOOS. Cumulative summation (CUSUM) analyses were used to assess learning curves. Results The use of PSCGs in HTO surgery was associated with a learning curve of 10 cases to optimize operative time (mean operative time 26.3 min ± 8.8), 8 cases to lessen surgeon anxiety levels, and 9 cases to decrease the number of fluoroscopic images to an average of 4.3 ± 1.2. Cumulative PSCGs experience did not affect accuracy of post-operative limb alignment with a mean: ΔHKA = 1.0° ± 1.0°, ΔMPTA = 0.5° ± 0.6° and ΔPPTA = 0.4° ± 0.8°. No significant difference was observed between the three surgeons for these three parameters. There was no statistical correlation between the number of procedures performed and the patient’s functional outcomes. Conclusion The use of PSCGs requires a short learning curve to optimize operating time, reduce the use of fluoroscopy and lessen surgeon’s anxiety levels. Additionally, this learning phase does not affect the accuracy of the postoperative correction and the functional results at 1 year. Level of evidence II: prospective observational study.

The existence of large volumes of data has considerably alleviated concerns regarding the availability of sufficient data instances for machine learning experiments. Nevertheless, in certain contexts, addressing limited data availability may demand distinct strategies and efforts. Analyzing COVID-19 predictions at pandemic beginning emerged a question: how much data is needed to make reliable predictions? When does the volume of data provide a better understanding of the disease’s evolution and, in turn, offer reliable forecasts? Given these questions, the objective of this study is to analyze learning curves obtained from predicting the incidence of COVID-19 in Brazilian States using ARIMA models with limited available data. To fulfill the objective, a retrospective exploration of COVID-19 incidence across the Brazilian States was performed. After the data acquisition and modeling, the model errors were assessed by employing a learning curve analysis. The asymptotic exponential curve fitting enabled the evaluation of the errors in different points, reflecting the increased available data over time. For a comprehensive understanding of the results at distinct stages of the time evolution, the average derivative of the curves and the equilibrium points were calculated, aimed to identify the convergence of the ARIMA models to a stable pattern. We observed differences in average derivatives and equilibrium values among the multiple samples. While both metrics ultimately confirmed the convergence to stability, the equilibrium points were more sensitive to changes in the models’ accuracy and provided a better indication of the learning progress. The proposed method for constructing learning curves enabled consistent monitoring of prediction results, providing evidence-based understandings required for informed decision-making.

There is an important challenge in systematically interpreting the internal representations of deep neural networks (DNNs). Existing techniques are often less effective for non-tabular tasks, or they primarily focus on qualitative, ad-hoc interpretations of models. In response, this study introduces a cognitive science-inspired, multi-dimensional quantification and visualization approach that captures two temporal dimensions of model learning: the “information-processing trajectory” and the “developmental trajectory.” The former represents the influence of incoming signals on an agent’s decision-making, while the latter conceptualizes the gradual improvement in an agent’s performance throughout its lifespan. Tracking the learning curves of DNNs enables researchers to explicitly identify the model appropriateness of a given task, examine the properties of the underlying input signals, and assess the model’s alignment (or lack thereof) with human learning experiences. To illustrate this method, we conducted 750 runs of simulations on two temporal tasks: gesture detection and sentence classification, showcasing its applicability across different types of deep learning tasks. Using four descriptive metrics to quantify the mapped learning curves— start , end - start , max , t max —, we identified significant differences in learning patterns based on data sources and class distinctions (all p’s  <  .0001), the prominent role of spatial semantics in gesture learning, and larger information gains in language learning. We highlight three key insights gained from mapping learning curves: non-monotonic progress , pairwise comparisons , and domain distinctions . We reflect on the theoretical implications of this method for cognitive processing, language models and representations from multiple modalities.

BackgroundUltrasound (US) guidance is not yet commonly used for cardiac device implantation, despite the first report of US-guided axillary access over twenty years ago and recent randomised trials demonstrating similar success rates to subclavian puncture or cephalic access. Changes to workflow and the learning curve represent barriers to more widespread adoption. There is limited real-world experience of the learning curve for ultrasound (US) guided axillary venous access for cardiac device implantation.PurposeWe described US-guided axillary venous access adapted to standard implant workflow, including application to device upgrade procedures, and using a standard vascular US probe. We investigated its learning curve, radiation exposure, safety, and efficacy.MethodsUS-guided access was performed by an experienced electrophysiologist with no prior application of the technique. Patients underwent standard skin preparation and draping. A standard vascular ultrasound probe was placed in the deltopectoral grove, medially and upwards towards the clavicle until the axillary vein and artery could be seen in the out-of-plane projection, with the vein in the middle of the imaging field. Lidocaine was injected and transdermal incision was made from below the midpoint of the probe inferiorly and parallel to the deltopectoral groove. The punctures were made after only the dermis had been incised which maintained optimal image quality. Access (needle to wire) and fluoroscopy times for US-guided access were compared to fluoroscopy guided access in ten control patients.Results147 US-guided punctures were performed in 74 patients for one (8%), two (71%) or three (17%) leads, or upgrades (4%). Access was successful in 97% (n=72). There were no access related peri-procedural complications.First US-guided access time was 30 seconds (interquartile range, IQR: 17,60), and was similar to fluoroscopy guided access time (43 seconds, IQR: 24,58; p=0.45). Access time stabilised after 45 procedures, decreasing from 81 (IQR: 61,90) to 16 (IQR: 10,20) seconds from the first to the last fifteen procedures (p<0.001).96% (n=69) did not require fluoroscopy. 4% (n=3) required 1 second fluoroscopy to confirm wire position after difficult passage. Radiation exposure saving estimated from controls was 29 (IQR: 17,56) seconds of fluoroscopy, resulting in 0.25 (IQR: 0,1.4) mGy cumulative skin dose, and 0.03 (IQR 0.02,0.5) Gy.cm2 effective dose area product.DiscussionThis study describes a novel technique for ultrasound guided axillary venous access for cardiac device implantation using standard equipment and minimal modification to the workflow. The technique demonstrated a high success rate in a cohort of 147 punctures in 74 patients, including device upgrade procedures. For an experienced operator, the learning curve stabilized after 45 procedures which is acceptable and safe even during training. Radiation dose-saving was approximately equivalent to one chest radiograph.ConclusionUS-guided axillary venous access for cardiac device implantation is a feasible alternative to fluoroscopy guided access. It can be performed safely by using a standard vascular ultrasound probe with a high success rate, even during the short learning curve.Abstract 95 Figure 1Ultrasound guided axillary vein anatomy. Ultrasound images are obtained by placing a vascular probe below and perpendicular to the clavicleAbstract 95 Figure 2Illustration of procedural steps (step 1: position of the ultrasound transducer in the deltopectoral grove, moved medially and upwards towards the clavicle until the axillary vein and artery could be seen in the out-of-plane projection, step 2: inject the lidocaine and make transdermal incision from below the midpoint of the probe inferiorly and parallel to the deltopectoral groove, step 3: puncture at superior end of incision, step 4: dissect down to pre-pectoral fascia, step 5: create pre-pectoral pocket)Abstract 95 Figure 3Access times for first puncture per patient decrease with experience. Statistically similar fluoroscopy guided access times for controls are shown in green for reference. One procedure requiring 506 seconds is not represented on the graphConflict of InterestNo