Aims. The aims of this study were to determine the success of a reconstruction algorithm used in major acetabular bone loss, and to further define the indications for custom-made implants in major acetabular bone loss. Methods. We reviewed a consecutive series of Paprosky type III acetabular defects treated according to a reconstruction algorithm. IIIA defects were planned to use a superior augment and hemispherical acetabular component. IIIB defects were planned to receive either a hemispherical acetabular component plus augments, a cup-cage reconstruction, or a custom-made implant. We used national digital health records and registry reports to identify any reoperation or re-revision procedure and Oxford Hip Score (OHS) for patient-reported outcomes. Implant survival was determined via Kaplan-Meier analysis. Results. A total of 105 procedures were carried out in 100 patients (five bilateral) with a mean age of 73 years (42 to 94). In the IIIA defects treated, 72.0% (36 of 50) required a porous metal augment; the remaining 14 patients were treated with a hemispherical acetabular component alone. In the IIIB defects, 63.6% (35 of 55) underwent reconstruction as planned with 20 patients who actually required a hemispherical acetabular component alone. At mean follow-up of 7.6 years, survival was 94.3% (95% confidence interval 97.4 to 88.1) for all-cause revision and the overall dislocation rate was 3.8% (4 of 105). There was no difference observed in survival between type IIIA and type IIIB defects and whether a hemispherical implant alone was used for the reconstruction or not. The mean gain in OHS was 16 points. Custom-made implants were only used in six cases, in patients with either a mega-defect in which the anteroposterior diameter > 80 mm, complex pelvic discontinuity, and massive bone loss in a small pelvis. Conclusion. Our findings suggest that a reconstruction algorithm can provide a successful approach to reconstruction in major acetabular bone loss. The use of custom implants has been defined in this series and accounts for < 5% of cases. Cite this article: Bone Joint J 2024;106-B(5 Supple B):47–53

In March 2012, an algorithm for the treatment of intertrochanteric fractures of the hip was introduced in our academic department of Orthopaedic Surgery. It included the use of specified implants for particular patterns of fracture. In this cohort study, 102 consecutive patients presenting with an intertrochanteric fracture were followed prospectively (post-algorithm group). Another 117 consecutive patients who had been treated immediately prior to the implementation of the algorithm were identified retrospectively as a control group (pre-algorithm group). The total cost of the implants prior to implementation of the algorithm was $357 457 (mean: $3055 (1947 to 4133)); compared with $255 120 (mean: $2501 (1052 to 4133)) after its implementation. There was a trend toward fewer complications in patients who were treated using the algorithm (33% pre- versus 22.5% post-algorithm; p = 0.088). Application of the algorithm to the pre-algorithm group revealed a potential overall cost saving of $70 295. The implementation of an evidence-based algorithm for the treatment of intertrochanteric fractures reduced costs while maintaining quality of care with a lower rate of complications and re-admissions. Cite this article: Bone Joint J 2014;96-B:1192–7

Aims. The aim of this study was to evaluate the reliability and validity of a patient-specific algorithm which we developed for predicting changes in sagittal pelvic tilt after total hip arthroplasty (THA). Methods. This retrospective study included 143 patients who underwent 171 THAs between April 2019 and October 2020 and had full-body lateral radiographs preoperatively and at one year postoperatively. We measured the pelvic incidence (PI), the sagittal vertical axis (SVA), pelvic tilt, sacral slope (SS), lumbar lordosis (LL), and thoracic kyphosis to classify patients into types A, B1, B2, B3, and C. The change of pelvic tilt was predicted according to the normal range of SVA (0 mm to 50 mm) for types A, B1, B2, and B3, and based on the absolute value of one-third of the PI-LL mismatch for type C patients. The reliability of the classification of the patients and the prediction of the change of pelvic tilt were assessed using kappa values and intraclass correlation coefficients (ICCs), respectively. Validity was assessed using the overall mean error and mean absolute error (MAE) for the prediction of the change of pelvic tilt. Results. The kappa values were 0.927 (95% confidence interval (CI) 0.861 to 0.992) and 0.945 (95% CI 0.903 to 0.988) for the inter- and intraobserver reliabilities, respectively, and the ICCs ranged from 0.919 to 0.997. The overall mean error and MAE for the prediction of the change of pelvic tilt were -0.3° (SD 3.6°) and 2.8° (SD 2.4°), respectively. The overall absolute change of pelvic tilt was 5.0° (SD 4.1°). Pre- and postoperative values and changes in pelvic tilt, SVA, SS, and LL varied significantly among the five types of patient. Conclusion. We found that the proposed algorithm was reliable and valid for predicting the standing pelvic tilt after THA. Cite this article: Bone Joint J 2024;106-B(1):19–27

Aims. The aim of this study was to evaluate the ability of a machine-learning algorithm to diagnose prosthetic loosening from preoperative radiographs and to investigate the inputs that might improve its performance. Methods. A group of 697 patients underwent a first-time revision of a total hip (THA) or total knee arthroplasty (TKA) at our institution between 2012 and 2018. Preoperative anteroposterior (AP) and lateral radiographs, and historical and comorbidity information were collected from their electronic records. Each patient was defined as having loose or fixed components based on the operation notes. We trained a series of convolutional neural network (CNN) models to predict a diagnosis of loosening at the time of surgery from the preoperative radiographs. We then added historical data about the patients to the best performing model to create a final model and tested it on an independent dataset. Results. The convolutional neural network we built performed well when detecting loosening from radiographs alone. The first model built de novo with only the radiological image as input had an accuracy of 70%. The final model, which was built by fine-tuning a publicly available model named DenseNet, combining the AP and lateral radiographs, and incorporating information from the patient’s history, had an accuracy, sensitivity, and specificity of 88.3%, 70.2%, and 95.6% on the independent test dataset. It performed better for cases of revision THA with an accuracy of 90.1%, than for cases of revision TKA with an accuracy of 85.8%. Conclusion. This study showed that machine learning can detect prosthetic loosening from radiographs. Its accuracy is enhanced when using highly trained public algorithms, and when adding clinical data to the algorithm. While this algorithm may not be sufficient in its present state of development as a standalone metric of loosening, it is currently a useful augment for clinical decision making. Cite this article: Bone Joint J 2020;102-B(6 Supple A):101–106

Aims. In 2013, we introduced a specialized, centralized, and interdisciplinary team in our institution that applied a standardized diagnostic and treatment algorithm for the management of prosthetic joint infections (PJIs). The hypothesis for this study was that the outcome of treatment would be improved using this approach. Patients and Methods. In a retrospective analysis with a standard postoperative follow-up, 95 patients with a PJI of the hip and knee who were treated with a two-stage exchange between 2013 and 2017 formed the study group. A historical cohort of 86 patients treated between 2009 and 2011 not according to the standardized protocol served as a control group. The success of treatment was defined according to the Delphi criteria in a two-year follow-up. Results. Patients in the study group had a significantly higher Charlson Comorbidity Index (3.9 vs 3.1; p = 0.009) and rate of previous revisions for infection (52.6% vs 36%; p = 0.025), and tended to be older (69.0 vs 66.2 years; p = 0.075) with a broader polymicrobial spectrum (47.3% vs 33.7%; p = 0.062). The rate of recurrent infection (3.1% vs 10.4%; p = 0.048) and the mean time interval between the two stages of the procedure (66.6 vs 80.7 days; p < 0.001) were reduced significantly in the study group compared with the control group. Conclusion. We were able to show that the outcome following the treatment of PJIs of the hip and knee is better when managed in a separate department with an interdisciplinary team using a standard algorithm

Aims. The Oswestry-Bristol Classification (OBC) is an MRI-specific assessment tool to grade trochlear dysplasia. The aim of this study is to validate clinically the OBC by demonstrating its use in selecting treatments that are safe and effective. Methods. The OBC and the patellotrochlear index were used as part of the Oswestry Patellotrochlear Algorithm (OPTA) to guide the surgical treatment of patients with patellar instability. Patients were assigned to one of four treatment groups: medial patellofemoral ligament reconstruction (MPFLr); MPFLr + tibial tubercle distalization (TTD); trochleoplasty; or trochleoplasty + TTD. A prospective analysis of a longitudinal patellofemoral database was performed. Between 2012 and 2018, 202 patients (233 knees) with a mean age of 24.2 years (SD 8.1), with recurrent patellar instability were treated by two fellowship-trained consultant sports/knee surgeons at The Robert Jones and Agnes Hunt Orthopaedic Hospital. Clinical efficacy of each treatment group was assessed by Kujala, International Knee Documentation Committee (IKDC), and EuroQol five-dimension questionnaire (EQ-5D) scores at baseline, and up to 60 months postoperatively. Their safety was assessed by complication rate and requirement for further surgery. The pattern of clinical outcome over time was analyzed using mixed regression modelling. Results. In all, 135 knees (mean age 24.9 years (SD 9.4)) were treated using a MPFLr. Ten knees (7.4%) required additional surgery. A total of 50 knees (mean age 24.4 years (SD 6.3)) were treated using MPFLr + TTD. Ten (20%) required additional surgery. A total of 20 knees (mean age 19.5 years (SD 3.0)) were treated using trochleoplasty + TTD. Three patients (15%) required additional surgery. In each treatment group, there was a significant improvement in Kujala, IKDC, and EQ-5D at one year postoperatively (p < 0.001) with a recognized level of overall complication rate. Conclusion. The OBC is a valid assessment tool to grade patients with trochlear dysplasia and, when used as part of the OPTA, helps to determine treatments that are safe and effective. This fulfils the requirements for its application in mainstream clinical practice. Cite this article: Bone Joint J 2021;103-B(10):1586–1594

The use of plate-and-cable constructs to treat periprosthetic fractures around a well-fixed femoral component in total hip replacements has been reported to have high rates of failure. Our aim was to evaluate the results of a surgical treatment algorithm to use these lateral constructs reliably in Vancouver type-B1 and type-C fractures. The joint was dislocated and the stability of the femoral component was meticulously evaluated in 45 type-B1 fractures. This led to the identification of nine (20%) unstable components. The fracture was considered to be suitable for single plate-and-cable fixation by a direct reduction technique if the integrity of the medial cortex could be restored. Union was achieved in 29 of 30 fractures (97%) at a mean of 6.4 months (3 to 30) in 29 type-B1 and five type-C fractures. Three patients developed an infection and one construct failed. Using this algorithm plate-and-cable constructs can be used safely, but indirect reduction with minimal soft-tissue damage could lead to shorter times to union and lower rates of complications

The purpose of this study was to evaluate treatment results following arthroscopic triangular fibrocartilage complex (TFCC) debridement for recalcitrant ulnar wrist pain. According to the treatment algorithm, 66 patients (36 men and 30 women with a mean age of 38.1 years (15 to 67)) with recalcitrant ulnar wrist pain were allocated to undergo ulnar shortening osteotomy (USO; n = 24), arthroscopic TFCC repair (n = 15), arthroscopic TFCC debridement (n = 14) or prolonged conservative treatment (n = 13). The mean follow-up was 36.0 months (15 to 54). Significant differences in Hand20 score at 18 months were evident between the USO group and TFCC debridement group (p = 0.003), and between the TFCC repair group and TFCC debridement group (p = 0.029). Within-group comparisons showed that Hand20 score at five months or later and pain score at two months or later were significantly decreased in the USO/TFCC repair groups. In contrast, scores in the TFCC debridement/conservative groups did not decrease significantly. Grip strength at 18 months was significantly improved in the USO/TFCC repair groups, but not in the TFCC debridement/conservative groups. TFCC debridement shows little benefit on the clinical course of recalcitrant ulnar wrist pain even after excluding patients with ulnocarpal abutment or TFCC detachment from the fovea from the indications for arthroscopic TFCC debridement. Cite this article: Bone Joint J 2013;95-B:1687–96

We investigated the incidence of anomalies in the vertebral arteries and Circle of Willis with three-dimensional CT angiography in 55 consecutive patients who had undergone an instrumented posterior fusion of the cervical spine.

We recorded any peri-operative and post-operative complications. The frequency of congenital anomalies was 30.9%, abnormal vertebral artery blood flow was 58.2% and vertebral artery dominance 40%.

The posterior communicating artery was occluded on one side in 41.8% of patients and bilaterally in 38.2%. Variations in the vertebral arteries and Circle of Willis were not significantly related to the presence or absence of posterior communicating arteries. Importantly, 18.2% of patients showed characteristic variations in the Circle of Willis with unilateral vertebral artery stenosis or a dominant vertebral artery, indicating that injury may cause lethal complications. One patient had post-operative cerebellar symptoms due to intra-operative injury of the vertebral artery, and one underwent a different surgical procedure because of insufficient collateral circulation.

Pre-operative assessment of the vertebral arteries and Circle of Willis is essential if a posterior spinal fusion with instrumentation is to be carried out safely.

Cite this article: Bone Joint J 2014;96-B:535–40.

Aims. Machine learning (ML), a branch of artificial intelligence that uses algorithms to learn from data and make predictions, offers a pathway towards more personalized and tailored surgical treatments. This approach is particularly relevant to prevalent joint diseases such as osteoarthritis (OA). In contrast to end-stage disease, where joint arthroplasty provides excellent results, early stages of OA currently lack effective therapies to halt or reverse progression. Accurate prediction of OA progression is crucial if timely interventions are to be developed, to enhance patient care and optimize the design of clinical trials. Methods. A systematic review was conducted in accordance with PRISMA guidelines. We searched MEDLINE and Embase on 5 May 2024 for studies utilizing ML to predict OA progression. Titles and abstracts were independently screened, followed by full-text reviews for studies that met the eligibility criteria. Key information was extracted and synthesized for analysis, including types of data (such as clinical, radiological, or biochemical), definitions of OA progression, ML algorithms, validation methods, and outcome measures. Results. Out of 1,160 studies initially identified, 39 were included. Most studies (85%) were published between 2020 and 2024, with 82% using publicly available datasets, primarily the Osteoarthritis Initiative. ML methods were predominantly supervised, with significant variability in the definitions of OA progression: most studies focused on structural changes (59%), while fewer addressed pain progression or both. Deep learning was used in 44% of studies, while automated ML was used in 5%. There was a lack of standardization in evaluation metrics and limited external validation. Interpretability was explored in 54% of studies, primarily using SHapley Additive exPlanations. Conclusion. Our systematic review demonstrates the feasibility of ML models in predicting OA progression, but also uncovers critical limitations that currently restrict their clinical applicability. Future priorities should include diversifying data sources, standardizing outcome measures, enforcing rigorous validation, and integrating more sophisticated algorithms. This paradigm shift from predictive modelling to actionable clinical tools has the potential to transform patient care and disease management in orthopaedic practice. Cite this article: Bone Joint J 2024;106-B(11):1216–1222

Aims. Natural Language Processing (NLP) offers an automated method to extract data from unstructured free text fields for arthroplasty registry participation. Our objective was to investigate how accurately NLP can be used to extract structured clinical data from unstructured clinical notes when compared with manual data extraction. Methods. A group of 1,000 randomly selected clinical and hospital notes from eight different surgeons were collected for patients undergoing primary arthroplasty between 2012 and 2018. In all, 19 preoperative, 17 operative, and two postoperative variables of interest were manually extracted from these notes. A NLP algorithm was created to automatically extract these variables from a training sample of these notes, and the algorithm was tested on a random test sample of notes. Performance of the NLP algorithm was measured in Statistical Analysis System (SAS) by calculating the accuracy of the variables collected, the ability of the algorithm to collect the correct information when it was indeed in the note (sensitivity), and the ability of the algorithm to not collect a certain data element when it was not in the note (specificity). Results. The NLP algorithm performed well at extracting variables from unstructured data in our random test dataset (accuracy = 96.3%, sensitivity = 95.2%, and specificity = 97.4%). It performed better at extracting data that were in a structured, templated format such as range of movement (ROM) (accuracy = 98%) and implant brand (accuracy = 98%) than data that were entered with variation depending on the author of the note such as the presence of deep-vein thrombosis (DVT) (accuracy = 90%). Conclusion. The NLP algorithm used in this study was able to identify a subset of variables from randomly selected unstructured notes in arthroplasty with an accuracy above 90%. For some variables, such as objective exam data, the accuracy was very high. Our findings suggest that automated algorithms using NLP can help orthopaedic practices retrospectively collect information for registries and quality improvement (QI) efforts. Cite this article: Bone Joint J 2020;102-B(7 Supple B):99–104

Aims. Failure of irrigation and debridement (I&D) for prosthetic joint infection (PJI) is influenced by numerous host, surgical, and pathogen-related factors. We aimed to develop and validate a practical, easy-to-use tool based on machine learning that may accurately predict outcome following I&D surgery taking into account the influence of numerous factors. Methods. This was an international, multicentre retrospective study of 1,174 revision total hip (THA) and knee arthroplasties (TKA) undergoing I&D for PJI between January 2005 and December 2017. PJI was defined using the Musculoskeletal Infection Society (MSIS) criteria. A total of 52 variables including demographics, comorbidities, and clinical and laboratory findings were evaluated using random forest machine learning analysis. The algorithm was then verified through cross-validation. Results. Of the 1,174 patients that were included in the study, 405 patients (34.5%) failed treatment. Using random forest analysis, an algorithm that provides the probability for failure for each specific patient was created. By order of importance, the ten most important variables associated with failure of I&D were serum CRP levels, positive blood cultures, indication for index arthroplasty other than osteoarthritis, not exchanging the modular components, use of immunosuppressive medication, late acute (haematogenous) infections, methicillin-resistant Staphylococcus aureus infection, overlying skin infection, polymicrobial infection, and older age. The algorithm had good discriminatory capability (area under the curve = 0.74). Cross-validation showed similar probabilities comparing predicted and observed failures indicating high accuracy of the model. Conclusion. This is the first study in the orthopaedic literature to use machine learning as a tool for predicting outcomes following I&D surgery. The developed algorithm provides the medical profession with a tool that can be employed in clinical decision-making and improve patient care. Future studies should aid in further validating this tool on additional cohorts. Cite this article: Bone Joint J 2020;102-B(7 Supple B):11–19

Aims. The purpose of this study was to develop a convolutional neural network (CNN) for fracture detection, classification, and identification of greater tuberosity displacement ≥ 1 cm, neck-shaft angle (NSA) ≤ 100°, shaft translation, and articular fracture involvement, on plain radiographs. Methods. The CNN was trained and tested on radiographs sourced from 11 hospitals in Australia and externally validated on radiographs from the Netherlands. Each radiograph was paired with corresponding CT scans to serve as the reference standard based on dual independent evaluation by trained researchers and attending orthopaedic surgeons. Presence of a fracture, classification (non- to minimally displaced; two-part, multipart, and glenohumeral dislocation), and four characteristics were determined on 2D and 3D CT scans and subsequently allocated to each series of radiographs. Fracture characteristics included greater tuberosity displacement ≥ 1 cm, NSA ≤ 100°, shaft translation (0% to < 75%, 75% to 95%, > 95%), and the extent of articular involvement (0% to < 15%, 15% to 35%, or > 35%). Results. For detection and classification, the algorithm was trained on 1,709 radiographs (n = 803), tested on 567 radiographs (n = 244), and subsequently externally validated on 535 radiographs (n = 227). For characterization, healthy shoulders and glenohumeral dislocation were excluded. The overall accuracy for fracture detection was 94% (area under the receiver operating characteristic curve (AUC) = 0.98) and for classification 78% (AUC 0.68 to 0.93). Accuracy to detect greater tuberosity fracture displacement ≥ 1 cm was 35.0% (AUC 0.57). The CNN did not recognize NSAs ≤ 100° (AUC 0.42), nor fractures with ≥ 75% shaft translation (AUC 0.51 to 0.53), or with ≥ 15% articular involvement (AUC 0.48 to 0.49). For all objectives, the model’s performance on the external dataset showed similar accuracy levels. Conclusion. CNNs proficiently rule out proximal humerus fractures on plain radiographs. Despite rigorous training methodology based on CT imaging with multi-rater consensus to serve as the reference standard, artificial intelligence-driven classification is insufficient for clinical implementation. The CNN exhibited poor diagnostic ability to detect greater tuberosity displacement ≥ 1 cm and failed to identify NSAs ≤ 100°, shaft translations, or articular fractures. Cite this article: Bone Joint J 2024;106-B(11):1348–1360

Aims. Obtaining solid implant fixation is crucial in revision total knee arthroplasty (rTKA) to avoid aseptic loosening, a major reason for re-revision. This study aims to validate a novel grading system that quantifies implant fixation across three anatomical zones (epiphysis, metaphysis, diaphysis). Methods. Based on pre-, intra-, and postoperative assessments, the novel grading system allocates a quantitative score (0, 0.5, or 1 point) for the quality of fixation achieved in each anatomical zone. The criteria used by the algorithm to assign the score include the bone quality, the size of the bone defect, and the type of fixation used. A consecutive cohort of 245 patients undergoing rTKA from 2012 to 2018 were evaluated using the current novel scoring system and followed prospectively. In addition, 100 first-time revision cases were assessed radiologically from the original cohort and graded by three observers to evaluate the intra- and inter-rater reliability of the novel radiological grading system. Results. At a mean follow-up of 90 months (64 to 130), only two out of 245 cases failed due to aseptic loosening. Intraoperative grading yielded mean scores of 1.87 (95% confidence interval (CI) 1.82 to 1.92) for the femur and 1.96 (95% CI 1.92 to 2.0) for the tibia. Only 3.7% of femoral and 1.7% of tibial reconstructions fell below the 1.5-point threshold, which included the two cases of aseptic loosening. Interobserver reliability for postoperative radiological grading was 0.97 for the femur and 0.85 for the tibia. Conclusion. A minimum score of 1.5 points for each skeletal segment appears to be a reasonable cut-off to define sufficient fixation in rTKA. There were no revisions for aseptic loosening at mid-term follow-up when this fixation threshold was achieved or exceeded. When assessing first-time revisions, this novel grading system has shown excellent intra- and interobserver reliability. Cite this article: Bone Joint J 2024;106-B(5):468–474

Aims. Accurate diagnosis of chronic periprosthetic joint infection (PJI) presents a significant challenge for hip surgeons. Preoperative diagnosis is not always easy to establish, making the intraoperative decision-making process crucial in deciding between one- and two-stage revision total hip arthroplasty (THA). Calprotectin is a promising point-of-care novel biomarker that has displayed high accuracy in detecting PJI. We aimed to evaluate the utility of intraoperative calprotectin lateral flow immunoassay (LFI) in THA patients with suspected chronic PJI. Methods. The study included 48 THAs in 48 patients with a clinical suspicion of PJI, but who did not meet European Bone and Joint Infection Society (EBJIS) PJI criteria preoperatively, out of 105 patients undergoing revision THA at our institution for possible PJI between November 2020 and December 2022. Intraoperatively, synovial fluid calprotectin was measured with LFI. Cases with calprotectin levels ≥ 50 mg/l were considered infected and treated with two-stage revision THA; in negative cases, one-stage revision was performed. At least five tissue cultures were obtained; the implants removed were sent for sonication. Results. Calprotectin was positive (≥ 50 mg/l) in 27 cases; out of these, 25 had positive tissue cultures and/or sonication. Calprotectin was negative in 21 cases. There was one false negative case, which had positive tissue cultures. Calprotectin showed an area under the curve of 0.917, sensitivity of 96.2%, specificity of 90.9%, positive predictive value of 92.6%, negative predictive value of 95.2%, positive likelihood ratio of 10.6, and negative likelihood ratio of 0.04. Overall, 45/48 patients were correctly diagnosed and treated by our algorithm, which included intraoperative calprotectin measurement. This yielded a 93.8% concordance with postoperatively assessed EBJIS criteria. Conclusion. Calprotectin can be a valuable tool in facilitating the intraoperative decision-making process for cases in which chronic PJI is suspected and diagnosis cannot be established preoperatively. Cite this article: Bone Joint J 2024;106-B(5 Supple B):118–124

Aims. The aim of this study was to identify factors associated with five-year cancer-related mortality in patients with limb and trunk soft-tissue sarcoma (STS) and develop and validate machine learning algorithms in order to predict five-year cancer-related mortality in these patients. Methods. Demographic, clinicopathological, and treatment variables of limb and trunk STS patients in the Surveillance, Epidemiology, and End Results Program (SEER) database from 2004 to 2017 were analyzed. Multivariable logistic regression was used to determine factors significantly associated with five-year cancer-related mortality. Various machine learning models were developed and compared using area under the curve (AUC), calibration, and decision curve analysis. The model that performed best on the SEER testing data was further assessed to determine the variables most important in its predictive capacity. This model was externally validated using our institutional dataset. Results. A total of 13,646 patients with STS from the SEER database were included, of whom 35.9% experienced five-year cancer-related mortality. The random forest model performed the best overall and identified tumour size as the most important variable when predicting mortality in patients with STS, followed by M stage, histological subtype, age, and surgical excision. Each variable was significant in logistic regression. External validation yielded an AUC of 0.752. Conclusion. This study identified clinically important variables associated with five-year cancer-related mortality in patients with limb and trunk STS, and developed a predictive model that demonstrated good accuracy and predictability. Orthopaedic oncologists may use these findings to further risk-stratify their patients and recommend an optimal course of treatment. Cite this article: Bone Joint J 2023;105-B(6):702–710

Aims. To examine whether natural language processing (NLP) using a clinically based large language model (LLM) could be used to predict patient selection for total hip or total knee arthroplasty (THA/TKA) from routinely available free-text radiology reports. Methods. Data pre-processing and analyses were conducted according to the Artificial intelligence to Revolutionize the patient Care pathway in Hip and knEe aRthroplastY (ARCHERY) project protocol. This included use of de-identified Scottish regional clinical data of patients referred for consideration of THA/TKA, held in a secure data environment designed for artificial intelligence (AI) inference. Only preoperative radiology reports were included. NLP algorithms were based on the freely available GatorTron model, a LLM trained on over 82 billion words of de-identified clinical text. Two inference tasks were performed: assessment after model-fine tuning (50 Epochs and three cycles of k-fold cross validation), and external validation. Results. For THA, there were 5,558 patient radiology reports included, of which 4,137 were used for model training and testing, and 1,421 for external validation. Following training, model performance demonstrated average (mean across three folds) accuracy, F1 score, and area under the receiver operating curve (AUROC) values of 0.850 (95% confidence interval (CI) 0.833 to 0.867), 0.813 (95% CI 0.785 to 0.841), and 0.847 (95% CI 0.822 to 0.872), respectively. For TKA, 7,457 patient radiology reports were included, with 3,478 used for model training and testing, and 3,152 for external validation. Performance metrics included accuracy, F1 score, and AUROC values of 0.757 (95% CI 0.702 to 0.811), 0.543 (95% CI 0.479 to 0.607), and 0.717 (95% CI 0.657 to 0.778) respectively. There was a notable deterioration in performance on external validation in both cohorts. Conclusion. The use of routinely available preoperative radiology reports provides promising potential to help screen suitable candidates for THA, but not for TKA. The external validation results demonstrate the importance of further model testing and training when confronted with new clinical cohorts. Cite this article: Bone Joint J 2024;106-B(7):688–695

Aims. Ganz’s studies made it possible to address joint deformities on both the femoral and acetabular side brought about by Perthes’ disease. Femoral head reduction osteotomy (FHRO) was developed to improve joint congruency, along with periacetabular osteotomy (PAO), which may enhance coverage and containment. The purpose of this study is to show the clinical and morphological outcomes of the technique and the use of an implemented planning approach. Methods. From September 2015 to December 2021, 13 FHROs were performed on 11 patients for Perthes’ disease in two centres. Of these, 11 hips had an associated PAO. A specific CT- and MRI-based protocol for virtual simulation of the corrections was developed. Outcomes were assessed with radiological parameters (sphericity index, extrusion index, integrity of the Shenton’s line, lateral centre-edge angle (LCEA), Tönnis angle), and clinical parameters (range of motion, visual analogue scale (VAS) for pain, Merle d'Aubigné-Postel score, modified Harris Hip Score (mHHS), and EuroQol five-dimension five-level health questionnaire (EQ-5D-5L)). Early and late complications were reported. Results. The mean follow-up was 39.7 months (standard deviation (SD) 26.4). The mean age at surgery was 11.4 years (SD 1.6). No major complications were recorded. One patient required a total hip arthroplasty. Mean femoral head sphericity increased from 46.8% (SD 9.34%) to 70.2% (SD 15.44; p < 0.001); mean LCEA from 19.2° (SD 9.03°) to 44° (SD 10.27°; p < 0.001); mean extrusion index from 37.8 (SD 8.70) to 7.5 (SD 9.28; p < 0.001); and mean Tönnis angle from 16.5° (SD 12.35°) to 4.8° (SD 4.05°; p = 0.100). The mean VAS improved from 3.55 (SD 3.05) to 1.22 (1.72; p = 0.06); mean Merle d’Aubigné-Postel score from 14.55 (SD 1.74) to 16 (SD 1.6; p = 0.01); and mean mHHS from 60.6 (SD 18.06) to 81 (SD 6.63; p = 0.021). The EQ-5D-5L also showed significant improvements. Conclusion. FHRO associated with periacetabular procedures is a safe technique that showed improved functional, clinical, and morphological outcomes in Perthes’ disease. The newly introduced simulation and planning algorithm may help to further refine the technique. Cite this article: Bone Joint J 2024;106-B(5 Supple B):40–46

Aims. Total hip arthroplasty (THA) and total knee arthroplasty (TKA) are common orthopaedic procedures requiring postoperative radiographs to confirm implant positioning and identify complications. Artificial intelligence (AI)-based image analysis has the potential to automate this postoperative surveillance. The aim of this study was to prepare a scoping review to investigate how AI is being used in the analysis of radiographs following THA and TKA, and how accurate these tools are. Methods. The Embase, MEDLINE, and PubMed libraries were systematically searched to identify relevant articles. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for scoping reviews and Arksey and O’Malley framework were followed. Study quality was assessed using a modified Methodological Index for Non-Randomized Studies tool. AI performance was reported using either the area under the curve (AUC) or accuracy. Results. Of the 455 studies identified, only 12 were suitable for inclusion. Nine reported implant identification and three described predicting risk of implant failure. Of the 12, three studies compared AI performance with orthopaedic surgeons. AI-based implant identification achieved AUC 0.992 to 1, and most algorithms reported an accuracy > 90%, using 550 to 320,000 training radiographs. AI prediction of dislocation risk post-THA, determined after five-year follow-up, was satisfactory (AUC 76.67; 8,500 training radiographs). Diagnosis of hip implant loosening was good (accuracy 88.3%; 420 training radiographs) and measurement of postoperative acetabular angles was comparable to humans (mean absolute difference 1.35° to 1.39°). However, 11 of the 12 studies had several methodological limitations introducing a high risk of bias. None of the studies were externally validated. Conclusion. These studies show that AI is promising. While it already has the ability to analyze images with significant precision, there is currently insufficient high-level evidence to support its widespread clinical use. Further research to design robust studies that follow standard reporting guidelines should be encouraged to develop AI models that could be easily translated into real-world conditions. Cite this article: Bone Joint J 2022;104-B(8):929–937

In recent years, machine learning (ML) and artificial neural networks (ANNs), a particular subset of ML, have been adopted by various areas of healthcare. A number of diagnostic and prognostic algorithms have been designed and implemented across a range of orthopaedic sub-specialties to date, with many positive results. However, the methodology of many of these studies is flawed, and few compare the use of ML with the current approach in clinical practice. Spinal surgery has advanced rapidly over the past three decades, particularly in the areas of implant technology, advanced surgical techniques, biologics, and enhanced recovery protocols. It is therefore regarded an innovative field. Inevitably, spinal surgeons will wish to incorporate ML into their practice should models prove effective in diagnostic or prognostic terms. The purpose of this article is to review published studies that describe the application of neural networks to spinal surgery and which actively compare ANN models to contemporary clinical standards allowing evaluation of their efficacy, accuracy, and relatability. It also explores some of the limitations of the technology, which act to constrain the widespread adoption of neural networks for diagnostic and prognostic use in spinal care. Finally, it describes the necessary considerations should institutions wish to incorporate ANNs into their practices. In doing so, the aim of this review is to provide a practical approach for spinal surgeons to understand the relevant aspects of neural networks. Cite this article: Bone Joint J 2021;103-B(9):1442–1448