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. The aim of this study was to investigate the relationship between the Orthopaedic Trauma Society (OTS) classification of open fractures and economic costs. Methods. Resource use was measured during the six months that followed open fractures of the lower limb in 748 adults recruited as part of two large clinical trials within the UK Major Trauma Research Network. Resource inputs were valued using unit costs drawn from primary and secondary sources. Economic costs (GBP sterling, 2017 to 2018 prices), estimated from both a NHS and Personal Social Services (PSS) perspective, were related to the degree of complexity of the open fracture based on the OTS classification. Results. Adjusted mean total NHS and PSS costs were £13,785 following treatment of complex fractures and £3,550 following treatment of simple fractures, where the open fracture wound is closed at the end of the first wound debridement, generating a mean difference of £10,235 (95% confidence interval £8,074 to £12,396). Conclusion. Following previous work correlating clinical outcomes with the OTS classification of open fractures, this study suggests that the new OTS classification also correlates with economic costs estimated from alternative study perspectives. Cite this article: Bone Joint J 2022;104-B(3):408–412

Aims. To describe a new objective classification for open fractures of the lower limb and to correlate the classification with patient-centred outcomes. Methods. The proposed classification was investigated within a cohort of adults with open fractures of the lower limb who were recruited as part of two large clinical trials within the UK Major Trauma Network. The classification was correlated with patient-reported Disability Rating Index (DRI) and EuroQol five-dimension questionnaire (EQ-5D) health-related quality of life in the year after injury, and with deep infection at 30 days, according to the Centers for Disease Control and Prevention definition of a deep surgical site infection. Results. A total of 748 participants were included in the analysis. Of these, 288 (38.5%) had a simple open fracture and 460 (61.5%) had a complex fracture as defined by the new classification system. At 12 months, the mean DRI in the simple fracture group was 32.5 (SD 26.8) versus 43.9 (SD 26.1) in the complex fracture group (odds ratio (OR) 8.19; 95% confidence interval (CI) 3.69 to 12.69). At 12 months the mean health-related quality of life (EQ-5D utility) in the simple fracture group was 0.59 (SD 0.29) versus 0.56 (SD 0.32) in the complex fracture group (OR -0.03; 95% CI -0.09 to 0.02). The differences in the rate of deep infection at 30 days was not statistically significant. Conclusion. The Orthopaedic Trauma Society open fracture classification is based upon objective descriptors of the injury and correlates with patient-centred outcomes in a large cohort of open fractures of the lower limb. Cite this article: Bone Joint J 2020;102-B(11):1469–1474

Aims. This study explores data quality in operation type and fracture classification recorded as part of a large research study and a national audit with an independent review. Patients and Methods. At 17 centres, an expert surgeon reviewed a randomly selected subset of cases from their centre with regard to fracture classification using the AO system and type of operation performed. Agreement for these variables was then compared with the data collected during conduct of the World Hip Trauma Evaluation (WHiTE) cohort study. Both types of surgery and fracture classification were collapsed to identify the level of detail of reporting that achieved meaningful agreement. In the National Hip Fracture Database (NHFD), the types of operation and fracture classification were explored to identify the proportion of “highly improbable” combinations. Results. The records were reviewed for 903 cases. Agreement for the subtypes of extracapsular fracture was poor; most centres achieved no better than “fair” agreement. When the classification was collapsed to a single option for “extracapsular” fracture, only four centres failed to have at least “moderate” agreement. There was only “moderate” agreement for the subtypes of intracapsular fracture, which improved to “substantial” when collapsed to “intracapsular”. Subtrochanteric fracture types were well reported with “substantial” agreement. There was near “perfect” agreement for internal fixation procedures. “Perfect” or “substantial” agreement was achieved when the type of arthroplasty surgery was reported at the level of “hemiarthroplasty” and “total hip replacement”. When reviewing data submitted to the NHFD, a minimum of 5.2% of cases contained “highly improbable” procedures for the stated fracture classification. Conclusion. The complexity of collecting fracture classification data at a national scale compromises the accuracy with which detailed classification systems can be reported. Data around type of surgery performed show similar tendencies. Data capture, reporting, and interpretation in future studies must take this into account. Cite this article: Bone Joint J 2019;101-B:1292–1299

Restoration of native Coronal Plane Alignment of the Knee (CPAK) phenotype is a strategy suggested to achieve better satisfaction. The aim of this study was to investigate the influence of changes in CPAK classification on patient-reported outcome measures (PROMs) and survivorship in a large cohort of manual mechanically aligned (MA) cemented TKAs. A retrospective analysis of 1062 consecutive cemented TKAs using MA philosophy at a single institution. Pre- and post-operative hip-knee-ankle radiographs were classified using the CPAK classification. Oxford Knee Score (OKS) and patient satisfaction (4-point-Likert scale) were collected prospectively. Implant survival data was obtained from our national arthroplasty database. We compared the outcomes of patients who maintained or changed their CPAK classification following TKA. Satisfaction was analysed using chi-square test, and OKS was analysed using Mann-Whitney test. Pre-operatively, most patients were CPAK type-I (38.8%). 85.5% of patients changed their CPAK type post-operatively, with CPAK type-V observed in 41.2% of these. Significantly better satisfaction (p=0.033) and OKS (p=0.021) were observed at one-year follow-up in patients who changed CPAK type, although the difference was below OKS minimally important clinical difference. There was no difference in satisfaction (p=0.73) and OKS (p=0.26) at one year between CPAK-V and non-V classifications. Post-operative CPAK type had no correlation with satisfaction and OKS. 12 TKAs (1.1%) were revised within 10 years (3 septic). In this large cohort of MA-TKA, excellent survivorship was observed at 10 years, with no demonstrable difference in outcome related to the final CPAK phenotype or change in phenotype

Though knee osteoarthritis (OA) is diagnosed and monitored radiographically, full thickness cartilage loss (FTCL) has rarely been correlated with radiographic classification. This study aims to analyse which classification system correlates best with FTCL and assessing their reliability. Prospective study of 300 consecutive patients undergoing total knee arthroplasty (TKA) for OA. Two blinded examiners independently graded preoperative radiographs using 5 systems: Kellgren-Lawrence (KL); International Knee Documentation Committee (IKDC); Fairbank; Brandt; and Ahlback. Interobserver agreement was assessed using the intraclass correlation coefficient. Intraoperatively, anterior cruciate ligament (ACL) status and FTCL in 16 regions of interest were recorded. Radiographic classification and FTCL were correlated using the Spearman correlation coefficient. On average, each knee had 6.8±3.1 regions of FTCL, most common medially. The commonest patterns of FTCL were medial with patellofemoral (48%) and tricompartmental (30%). ACL status was associated with pattern of FTCL (p=0.02). All classification systems demonstrated moderate ICC, but this was highest for IKDC: whole knee 0.68 (95%CI 0.60–0.74); medial compartment 0.84 (0.80–0.87); and lateral compartment 0.79 (0.73–0.83). Correlation with FTCL was strongest for Ahlback (Spearman rho 0.27–0.39) and KL (0.30–0.33), though all systems demonstrated medium correlation. The Ahlback was the most discriminating in severe OA. Osteophyte presence in the medial compartment had high positive predictive value for FTCL, but not in the lateral compartment. The Ahlback and KL systems had the highest correlation with confirmed cartilage loss. However, the IKDC system displayed best interobserver reliability, with favourable correlation with FTCL in medial and lateral compartments, though less discriminating in severe disease

The most widely used classification system for acetabular fractures was developed by Judet, Judet and Letournel over 50 years ago primarily to aid surgical planning. As population demographics and injury mechanisms have altered over time, the fracture patterns also appear to be changing. We conducted a retrospective review of the imaging of 100 patients with a mean age of 54.9 years (19 to 94) and a male to female ratio of 69:31 seen between 2010 and 2013 with acetabular fractures in order to determine whether the current spectrum of injury patterns can be reliably classified using the original system. Three consultant pelvic and acetabular surgeons and one senior fellow analysed anonymous imaging. Inter-observer agreement for the classification of fractures that fitted into defined categories was substantial, (κ = 0.65, 95% confidence interval (CI) 0.51 to 0.76) with improvement to near perfect on inclusion of CT imaging (κ = 0.80, 95% CI 0.69 to 0.91). However, a high proportion of injuries (46%) were felt to be unclassifiable by more than one surgeon; there was moderate agreement on which these were (κ = 0.42 95% CI 0.31 to 0.54). Further review of the unclassifiable fractures in this cohort of 100 patients showed that they tended to occur in an older population (mean age 59.1 years; 22 to 94 vs 47.2 years; 19 to 94; p = 0.003) and within this group, there was a recurring pattern of anterior column and quadrilateral plate involvement, with or without an incomplete posterior element injury. . Cite this article: Bone Joint J 2015;97-B:1139–43

Our study was undertaken to assess the inter- and intra-observer variability of the classification system of Sanders for calcaneal fractures. Five consultant orthopaedic surgeons with different subspecialty interests classified CT scans of 28 calcaneal fractures using this classification system. After six months, they reclassified the scans. Kappa statistics were used to analyse the two groups. The interobserver variability of the classification system was 0.32 (95% confidence interval (CI) 0.26 to 0.38). The subclasses were then combined and assessment of agreement between the general classes as a whole gave a kappa value of 0.33 (95% CI 0.25 to 0.41). The mean kappa value for intra-observer variability of the classification system was 0.42 (95% CI 0.22 to 0.62). When the subclasses were combined, it was 0.45 (95% CI 0.21 to 0.65). Our results show that, despite its popularity, the classification system of Sanders has only fair agreement among users

Aims. To evaluate interobserver reliability of the Orthopaedic Trauma Association’s open fracture classification system (OTA-OFC). Patients and Methods. Patients of any age with a first presentation of an open long bone fracture were included. Standard radiographs, wound photographs, and a short clinical description were given to eight orthopaedic surgeons, who independently evaluated the injury using both the Gustilo and Anderson (GA) and OTA-OFC classifications. The responses were compared for variability using Cohen’s kappa. Results. The overall interobserver agreement was ĸ = 0.44 for the GA classification and ĸ = 0.49 for OTA-OFC, which reflects moderate agreement (0.41 to 0.60) for both classifications. The agreement in the five categories of OTA-OFC was: for skin, ĸ = 0.55 (moderate); for muscle, ĸ = 0.44 (moderate); for arterial injury, ĸ = 0.74 (substantial); for contamination, ĸ = 0.35 (fair); and for bone loss, ĸ = 0.41 (moderate). Conclusion. Although the OTA-OFC, with similar interobserver agreement to GA, offers a more detailed description of open fractures, further development may be needed to make it a reliable and robust tool. Cite this article: Bone Joint J 2018;100-B:242–6

In this study, we describe a morphological classification for greater tuberosity fractures of the proximal humerus. We divided these fractures into three types: avulsion, split and depression. We retrospectively reviewed all shoulder radiographs showing isolated greater tuberosity fractures in a Level I trauma centre between July 2007 and July 2012. We identified 199 cases where records and radiographs were reviewed and included 79 men and 120 women with a mean age of 58 years (23 to 96). The morphological classification was applied to the first 139 cases by three reviewers on two occasions using the Kappa statistic and compared with the AO and Neer classifications. The inter- and intra-observer reliability of the morphological classification was 0.73 to 0.77 and 0.69 to 0.86, respectively. This was superior to the Neer (0.31 to 0.35/0.54 to 0.63) and AO (0.30 to 0.32/0.59 to 0.65) classifications. The distribution of avulsion, split and depression type fractures was 39%, 41%, and 20%, respectively. This classification of greater tuberosity fractures is more reliable than the Neer or AO classifications. These distinct fracture morphologies are likely to have implications in terms of pathophysiology and surgical technique. Cite this article: Bone Joint J 2014;96-B:646–51

Specific radiological features identified by Brandser and Marsh were selected for the analysis of acetabular fractures according to the classification of Letournel and Judet. The method employs a binary approach that requires the observer to allocate each radiological feature to one of two groups. The inter- and intra-observer variances were assessed. The presence of articular displacement, marginal impaction, incongruity, intra-articular fragments and osteochondral injuries to the femoral head were analysed by a similar method. These factors were termed ‘modifiers’ and are generally considered when planning operative intervention and, critically, they may influence prognosis. Six observers independently assessed 30 sets of plain radiographs and CT scans on two separate occasions, 12 weeks apart. They were asked to determine the presence or absence of specific radiological features. This simple binary approach to classification yields an inter- and intra-observer agreement which ranges from moderate to near-perfect (κ = 0.49 to 0.88 and κ = 0.57 to 0.88, respectively). A similar approach to the modifiers yields only slight to fair inter-observer agreement (κ = 0.20 to 0.34) and slight to moderate intra-observer agreement (κ = 0 to 0.55)

Existing classifications of fractures of the head of the humerus are inadequate in terms of interobserver reliability and the predictability of the clinical outcome. From a combined study of 73 fracture specimens in museums and 84 CT-three-dimensional reconstructions in patients, we have devised a classification which appears to be more useful clinically. Common patterns of fracture and a plausible mechanism of injury were observed. In 3-D most proximal humeral fractures can be organised into five basic types. These correspond in some degree to the Codman/Neer classification, but differ significantly in regard to the more complex patterns of fracture. We observed a logical progression from simple to complex fractures. An interobserver reliability study was carried out which indicated the improved usefulness of this new 3-D concept in providing a common language among clinicians for classifying these injuries. When surgery is indicated, the 3-D concept is also invaluable in guiding the restitution of anatomy through either open or percutaneous means

The lateral pillar classification (LPC) is a widely used tool in determining prognosis and planning treatment in patients who are in the fragmentation stage of Perthes disease. The original classification has been modified to help increase the accuracy of the classification system by the Herring group. The purpose of our study was to independently assess this modified Herring classification. 35 standardized true antero-posterior radiographs of children in various stages of fragmentation were independently assessed by 6 senior observers on 2 separate occasions (6 weeks apart). Kappa analysis was used to assess the inter and intraobserver agreement between observations made. The degrees of agreement were as follows: poor, fair, moderate, good and very good. Intraobserver analysis revealed at best only moderate agreement for two observers. 3 observers showed fair consistency, whilst 1 remaining observer showed poor consistency between repeated observations (p<0.01). The highest scores for interobserver agreement varying between moderate to good could only be established between 2 observers. For the remaining observers results were just fair (p<0.01). This study highlights the lack of agreement between senior clinicians when applying the modified LPC. This has clinical implications when applying the classification to the decision making process in treating patients at risk of developing adverse outcomes from the disease. To our knowledge, this is the first time the modified LPC has been independently tested for its reproducibility by another specialist paediatric orthopaedic unit

We describe the routine imaging practices of Level 1 trauma centres for patients with severe pelvic ring fractures, and the interobserver reliability of the classification systems of these fractures using plain radiographs and three-dimensional (3D) CT reconstructions. Clinical and imaging data for 187 adult patients (139 men and 48 women, mean age 43 years (15 to 101)) with a severe pelvic ring fracture managed at two Level 1 trauma centres between July 2007 and June 2010 were extracted. Three experienced orthopaedic surgeons classified the plain radiographs and 3D CT reconstruction images of 100 patients using the Tile/AO and Young–Burgess systems. Reliability was compared using kappa statistics. A total of 115 patients (62%) had plain radiographs as well as two-dimensional (2D) CT and 3D CT reconstructions, 52 patients (28%) had plain films only, 12 (6.4%) had 2D and 3D CT reconstructions images only, and eight patients (4.3%) had no available images. The plain radiograph was limited to an anteroposterior pelvic view. Patients without imaging, or only plain films, were more severely injured. A total of 72 patients (39%) were imaged with a pelvic binder in situ. Interobserver reliability for the Tile/AO (Kappa 0.10 to 0.17) and Young–Burgess (Kappa 0.09 to 0.21) was low, and insufficient for clinical and research purposes. Severe pelvic ring fractures are difficult to classify due to their complexity, the increasing use of early treatment such as with pelvic binders, and the absence of imaging altogether in important patient sub-groups, such as those who die early of their injuries. Cite this article: Bone Joint J 2013;95-B:1396–1401

We sought to quantify agreement by different assessors of the AO classification for distal fractures of the radius. Thirty radiographs of acute distal radial fractures were evaluated by 36 assessors of varying clinical experience. Our findings suggest that AO ‘type’ and the presence or absence of articular displacement are measured with high consistency when classification of distal radial fractures is undertaken by experienced observers. Assessors at all experience levels had difficulty agreeing on AO ‘group’ and especially AO ‘subgroup’. To categorise distal radial fractures according to joint displacement and AO type is simple and reproducible. Our study examined only whether distal radial fractures could be consistently classified according to the AO system. Validation of the classification as a predictor of outcome will require a prospective clinical study

Introduction and objectives. High-energy pelvic fractures are life-threatening injuries. Approximately 15% to 30% of patients with high-energy pelvic injuries are hemodynamic unstable, hemorrhagic shock remains the main cause of death in patients with pelvic fractures, with an overall mortality rate from 6% to 35%. The correlation between fracture pattern and mortality in polytrauma with pelvic fracture has been previously investigated. However, the purpose of our investigation was to evaluate the relationship of hemodynamic instability with the pelvic fracture pattern according to different classifications. Materials and Methods. A retrospective study of high-energy pelvic fractures was performed for consecutive patients admitted to the emergency Level I trauma center in the polytrauma unit of our institution from June 2007 to June 2010. A total of 759 patients polytrauma were attended, whom 100 had a pelvic fracture and were included in our study. Demographic data, mechanism of injury and associated injuries were recorded. The patients were classified as hemodynamic stable or unstable according to the ATLS protocol. The pelvic fracture patterns were divided into stable and unstable according to Young-Burgess and Tile classifications. Statistical analysis was performed to determine the relationship between fracture pattern and hemodynamic stability. Secondary outcomes were obtained: the relationship with TCE and pulmonary injury, usefulness of the external fixation, relationship between fracture pattern and embolization requests. Chi-square test was used for the analysis and OR test. Results. Male was the gender most frequent (70%), the average age was 45.2 years and the mortality rate was 24%, the main mechanism of injury was motor vehicle injury (41%). Pelvic fracture pattern (neither Tile classification nor Young-Burgess classification) showed no correlation with hemodynamic situation (p>0.05). Neither death could not be predicted on the basis of pelvic fracture pattern (p>0.05). We found statistical association between patients affected by pelvic fracture in polytrauma and head injury, and death (p<0.01). Conclusion. Pelvic fracture pattern is not useful to predict hemodynamic instability in polytrauma, regardless the classification system used. Pelvic fracture is not significantly associated with high risk of mortality; however it contributes to increase mortality risk in cases of head injury associated

Our aim was to assess the reproducibility and the reliability of the Weber classification system for fractures of the ankle based on anteroposterior and lateral radiographs. Five observers with varying clinical experience reviewed 50 sets of blinded radiographs. The same observers reviewed the same radiographs again after an interval of four weeks. Inter- and intra-observer agreement was assessed based on the proportion of agreement and the values of the kappa coefficient. For inter-observer agreement, the mean kappa value was 0.61 (0.59 to 0.63) and the proportion of agreement was 78% (76% to 79%) and for intra-observer agreement the mean kappa value was 0.74 (0.39 to 0.86) with an 85% (60% to 93%) observed agreement. These results show that the Weber classification of fractures of the ankle based on two radiological views has substantial inter-observer reliability and intra-observer reproducibility

Materials and method. This study included 309 patients, over the age of 60 (44 men, 256 women, mean age 85.6 years) who underwent osteosynthesis using a short femoral nail(Gamma)for femoral intertrochanteric fractures. The new reduction classification was defined, in AP view by the degree of contact between the medial cortex of the proximal fragment and that of the medial cortex of the distal fragment. The classification in AP view is divided into three Types: Anatomical Type, Medial Type, and Lateral Type. As for the ML view was defined based on the degree of anterior cortex contact between proximal and distal fragment, was divided also into Anatomical Type, Intramedullary Type, and Extramedullary Type. We assessed lag screw sliding amount at 2 weeks period after surgery as the post-operative displacement. Results. The sliding at 2 weeks was in AP Anatomical Type 3.7mm, in Medial Type 3.2mm and in Lateral Type 9.4mm. The sliding in the Lateral Type was much greater than those of Anatomical Type and Medial Type (Significant p-value, p=0.01, 0.01). Likewise, in the ML view, Anatomical Type 3.0mm, Intramedullary Type 6.2mm and Extramedullary Type 4.7mm respectively. There was a significant difference between ML Anatomical Type and Intramedullary Type(p=0.01). Conclusion. The AP Lateral Type and the ML intramedullary Type were the most post-operative displacement. A new reduction classification could be useful to assist in the prediction of post-operation displacement

Interobserver reliability of the AO system of classification of fractures of the distal radius was assessed using plain radiographs and CT. Five observers classified 30 Colles’-type fractures using only plain radiographs; two months later they were reclassified using CT in addition. Interobserver reliability was poor in both series when detailed classification was used. By reducing the categories to five, interobserver reliability was slightly improved, but was still poor. When only two AO types were used, the reliability was moderate using plain radiographs and good to excellent with the addition of CT. The use of CT as well as plain radiographs brings interobserver reliability to a good level in assessment of the presence or absence of articular involvement, but is otherwise of minor value in improving the interobserver reliability of the AO system of classification of fractures of the distal radius

We evaluated the impact of stereo-visualisation of three-dimensional volume-rendering CT datasets on the inter- and intraobserver reliability assessed by kappa values on the AO/OTA and Neer classifications in the assessment of proximal humeral fractures. Four independent observers classified 40 fractures according to the AO/OTA and Neer classifications using plain radiographs, two-dimensional CT scans and with stereo-visualised three-dimensional volume-rendering reconstructions. Both classification systems showed moderate interobserver reliability with plain radiographs and two-dimensional CT scans. Three-dimensional volume-rendered CT scans improved the interobserver reliability of both systems to good. Intraobserver reliability was moderate for both classifications when assessed by plain radiographs. Stereo visualisation of three-dimensional volume rendering improved intraobserver reliability to good for the AO/OTA method and to excellent for the Neer classification. These data support our opinion that stereo visualisation of three-dimensional volume-rendering datasets is of value when analysing and classifying complex fractures of the proximal humerus