Citations

Background and Objectives: The number of patients who undergo multiple operations on a knee is increasing. The objective of this study was to develop a deep learning algorithm that could detect 17 different surgical implants on plain knee radiographs. Materials and Methods: An internal dataset consisted of 5206 plain knee antero-posterior X-rays from a single, tertiary institute for model development. An external set contained 238 X-rays from another tertiary institute. A total of 17 different types of implants including total knee arthroplasty, unicompartmental knee arthroplasty, plate, and screw were labeled. The internal dataset was approximately split into a train set, a validation set, and an internal test set at a ratio of 7:1:2. You Only look Once (YOLO) was selected as the detection network. Model performances with the validation set, internal test set, and external test set were compared. Results: Total accuracy, total sensitivity, total specificity value of the validation set, internal test set, and external test set were (0.978, 0.768, 0.999), (0.953, 0.810, 0.990), and (0.956, 0.493, 0.975), respectively. Means ± standard deviations (SDs) of diagonal components of confusion matrix for these three subsets were 0.858 ± 0.242, 0.852 ± 0.182, and 0.576 ± 0.312, respectively. True positive rate of total knee arthroplasty, the most dominant class of the dataset, was higher than 0.99 with internal subsets and 0.96 with an external test set. Conclusion: Implant identification on plain knee radiographs could be automated using a deep learning technique. The detection algorithm dealt with overlapping cases while maintaining high accuracy on total knee arthroplasty. This could be applied in future research that analyzes X-ray images with deep learning, which would help prompt decision-making in clinics.

AbstractMean, median, and mode are among the most basic and consistently used measures of central tendency in statistical analysis and are crucial for simplifying data sets to a single value. However, there is a lack of understanding of when to use each metric and how various factors can impact these values. The aim of this article is to clarify some of the confusion related to each measure and explain how to select the appropriate metric for a given data set. The authors present this work as an educational resource, ensuring that these common statistical concepts are better understood throughout the Orthopedic research community.

AbstractSupervised learning is the most common form of machine learning utilized in medical research. It is used to predict outcomes of interest or classify positive and/or negative cases with a known ground truth. Supervised learning describes a spectrum of techniques, ranging from traditional regression modeling to more complex tree boosting, which are becoming increasingly prevalent as the focus on “big data” develops. While these tools are becoming increasingly popular and powerful, there is a paucity of literature available that describe the strengths and limitations of these different modeling techniques. Typically, there is no formal training for health care professionals in the use of machine learning models. As machine learning applications throughout medicine increase, it is important that physicians and other health care professionals better understand the processes underlying application of these techniques. The purpose of this study is to provide an overview of commonly used supervised learning techniques with recent case examples within the orthopedic literature. An additional goal is to address disparities in the understanding of these methods to improve communication within and between research teams.

AbstractPurposeAccurate prediction of outcome following hip arthroscopy is challenging and machine learning has the potential to improve our predictive capability. The purpose of this study was to determine if machine learning analysis of the Danish Hip Arthroscopy Registry (DHAR) can develop a clinically meaningful calculator for predicting the probability of a patient undergoing subsequent revision surgery following primary hip arthroscopy.MethodsMachine learning analysis was performed on the DHAR. The primary outcome for the models was probability of revision hip arthroscopy within 1, 2, and/or 5 years after primary hip arthroscopy. Data were split randomly into training (75%) and test (25%) sets. Four models intended for these types of data were tested: Cox elastic net, random survival forest, gradient boosted regression (GBM), and super learner. These four models represent a range of approaches to statistical details like variable selection and model complexity. Model performance was assessed by calculating calibration and area under the curve (AUC). Analysis was performed using only variables available in the pre‐operative clinical setting and then repeated to compare model performance using all variables available in the registry.ResultsIn total, 5581 patients were included for analysis. Average follow‐up time or time‐to‐revision was 4.25 years (± 2.51) years and overall revision rate was 11%. All four models were generally well calibrated and demonstrated concordance in the moderate range when restricted to only pre‐operative variables (0.62–0.67), and when considering all variables available in the registry (0.63–0.66). The 95% confidence intervals for model concordance were wide for both analyses, ranging from a low of 0.53 to a high of 0.75, indicating uncertainty about the true accuracy of the models.ConclusionThe association between pre‐surgical factors and outcome following hip arthroscopy is complex. Machine learning analysis of the DHAR produced a model capable of predicting revision surgery risk following primary hip arthroscopy that demonstrated moderate accuracy but likely limited clinical usefulness. Prediction accuracy would benefit from enhanced data quality within the registry and this preliminary study holds promise for future model generation as the DHAR matures. Ongoing collection of high‐quality data by the DHAR should enable improved patient‐specific outcome prediction that is generalisable across the population.Level of evidenceLevel III.

Background: In professional sports, injuries resulting in loss of playing time have serious implications for both the athlete and the organization. Efforts to quantify injury probability utilizing machine learning have been met with renewed interest, and the development of effective models has the potential to supplement the decision-making process of team physicians. Purpose/Hypothesis: The purpose of this study was to (1) characterize the epidemiology of time-loss lower extremity muscle strains (LEMSs) in the National Basketball Association (NBA) from 1999 to 2019 and (2) determine the validity of a machine-learning model in predicting injury risk. It was hypothesized that time-loss LEMSs would be infrequent in this cohort and that a machine-learning model would outperform conventional methods in the prediction of injury risk. Study Design: Case-control study; Level of evidence, 3. Methods: Performance data and rates of the 4 major muscle strain injury types (hamstring, quadriceps, calf, and groin) were compiled from the 1999 to 2019 NBA seasons. Injuries included all publicly reported injuries that resulted in lost playing time. Models to predict the occurrence of a LEMS were generated using random forest, extreme gradient boosting (XGBoost), neural network, support vector machines, elastic net penalized logistic regression, and generalized logistic regression. Performance was compared utilizing discrimination, calibration, decision curve analysis, and the Brier score. Results: A total of 736 LEMSs resulting in lost playing time occurred among 2103 athletes. Important variables for predicting LEMS included previous number of lower extremity injuries; age; recent history of injuries to the ankle, hamstring, or groin; and recent history of concussion as well as 3-point attempt rate and free throw attempt rate. The XGBoost machine achieved the best performance based on discrimination assessed via internal validation (area under the receiver operating characteristic curve, 0.840), calibration, and decision curve analysis. Conclusion: Machine learning algorithms such as XGBoost outperformed logistic regression in the prediction of a LEMS that will result in lost time. Several variables increased the risk of LEMS, including a history of various lower extremity injuries, recent concussion, and total number of previous injuries.

Abstract Machine learning (ML) and artificial intelligence (AI) may be described as advanced statistical techniques using algorithms to “learn” to evaluate and predict relationships between input and results without explicit human programming, often with high accuracy. The potentials and pitfalls of ML continue to be explored as predictive modeling grows in popularity. While use of and optimism for AI continues to increase in orthopaedic surgery, there remains little high‐quality evidence of its ability to improve patient outcome. It is up to us as clinicians to provide context for ML models and guide the use of these technologies to optimize the outcome for our patients. Barriers to widespread adoption of ML include poor quality data, limits to compliant data sharing, few clinicians who are expert in ML statistical techniques, and computing costs including technology, infrastructure, personnel, energy, and updates.

Purpose To determine whether conventional logistic regression or machine learning algorithms were more precise in identifying the risk factors for unplanned overnight admission after medial patellofemoral ligament (MPFL) reconstruction. Methods A retrospective review of the prospectively collected National Surgical Quality Improvement Program database was performed to identify patients who underwent outpatient MPFL reconstruction from 2006–2018. Patients admitted overnight were identified as those with length of stay of 1 or more days. Models were generated using random forest, extreme gradient boosting, adaptive boosting, or elastic net penalized logistic regression, and an additional model was produced as a weighted ensemble of the 4 final algorithms. The predictive capacity of these models was compared to that of logistic regression. Results Of the 1307 patients identified, 221 (16.9%) required at least one overnight stay after MPFL reconstruction. Multivariate logistic regression found the following variables to be predictors of inpatient admission: age (odds ratio [OR] = 1.03 [95% confidence interval {CI} 1.02‐1.04]; P <.001), spinal anesthesia (OR = 3.42 [95% CI 1.98‐6.08]; P < .001), American Society of Anesthesiologists (ASA) class 3/4 (OR = 1.96 [95% CI 1.25‐3.06]; P < .001), history of chronic obstructive pulmonary disease (COPD) (OR = 6.44 [95% CI 1.58‐26.17]; P = .02), and body mass index (BMI) (OR = 1.03 [95% CI 1.01‐1.05]; P < .001). The ensemble model achieved the best performance based on discrimination assessed via internal validation (area under the curve = 0.722). The variables determined most important by the ensemble model were increasing BMI, increasing age, ASA class, anesthesia, smoking, hypertension, lateral release, and history of COPD. Conclusions An internally validated machine learning algorithm outperformed logistic regression modeling in predicting the need for unplanned overnight hospitalization after MPFL reconstruction. In this model, the most significant risk factors for admission were age, BMI, ASA class, smoking status, hypertension, lateral release, and history of COPD. This tool can be deployed to augment provider assessment to identify high‐risk candidates and appropriately set postoperative expectations for patients. Clinical Relevance Identifying and mitigating patient risk factors to prevent adverse surgical outcomes and hospitalizations is one of our primary goals. There may be a key role for machine learning algorithms to help successfully and efficiently risk stratify patients to decrease costs, appropriately set postoperative expectations, and increase the quality of delivered care.

AbstractThe application of machine learning (ML) to the field of orthopaedic surgery is rapidly increasing, but many surgeons remain unfamiliar with the nuances of this novel technique. With this editorial, we address a fundamental topic—the differences between ML techniques and traditional statistics. By doing so, we aim to further familiarize the reader with the new opportunities available thanks to the ML approach.

AbstractPurposeExternal validation of machine learning predictive models is achieved through evaluation of model performance on different groups of patients than were used for algorithm development. This important step is uncommonly performed, inhibiting clinical translation of newly developed models. Machine learning analysis of the Norwegian Knee Ligament Register (NKLR) recently led to the development of a tool capable of estimating the risk of anterior cruciate ligament (ACL) revision (https://swastvedt.shinyapps.io/calculator_rev/). The purpose of this study was to determine the external validity of the NKLR model by assessing algorithm performance when applied to patients from the Danish Knee Ligament Registry (DKLR).MethodsThe primary outcome measure of the NKLR model was probability of revision ACL reconstruction within 1, 2, and/or 5 years. For external validation, all DKLR patients with complete data for the five variables required for NKLR prediction were included. The five variables included graft choice, femur fixation device, KOOS QOL score at surgery, years from injury to surgery, and age at surgery. Predicted revision probabilities were calculated for all DKLR patients. The model performance was assessed using the same metrics as the NKLR study: concordance and calibration.ResultsIn total, 10,922 DKLR patients were included for analysis. Average follow‐up time or time‐to‐revision was 8.4 (± 4.3) years and overall revision rate was 6.9%. Surgical technique trends (i.e., graft choice and fixation devices) and injury characteristics (i.e., concomitant meniscus and cartilage pathology) were dissimilar between registries. The model produced similar concordance when applied to the DKLR population compared to the original NKLR test data (DKLR: 0.68; NKLR: 0.68–0.69). Calibration was poorer for the DKLR population at one and five years post primary surgery but similar to the NKLR at two years.ConclusionThe NKLR machine learning algorithm demonstrated similar performance when applied to patients from the DKLR, suggesting that it is valid for application outside of the initial patient population. This represents the first machine learning model for predicting revision ACL reconstruction that has been externally validated. Clinicians can use this in‐clinic calculator to estimate revision risk at a patient specific level when discussing outcome expectations pre‐operatively. While encouraging, it should be noted that the performance of the model on patients undergoing ACL reconstruction outside of Scandinavia remains unknown.Level of evidenceIII.