There is increasing popularity in the use of artificial intelligence and machine-learning techniques to provide diagnostic and prognostic models for various aspects of Trauma & Orthopaedic surgery. However, correct interpretation of these models is difficult for those without specific knowledge of computing or health data science methodology. Lack of current reporting standards leads to the potential for significant heterogeneity in the design and quality of published studies. We provide an overview of machine-learning techniques for the lay individual, including key terminology and best practice reporting guidelines. Cite this article: Bone Joint J 2021;103-B(12):1754–1758

Plain radiographs remain important diagnostic tools in the initial assessment of patients with suspected fractures or dislocations of the foot. Analysis of foot radiograph is a difficult task even in the hands of experts, which is misinterpreted in 2.1% cases. Human visual system is more sensitive to horizontal and vertical stimuli & cardinal orientations (Vertical and Horizontal) perceived more accurately than others. Any absolute judgement task is subject to two types of error: Systematic error and random error. Hence a system of assessment with very high random error of measurement is unlikely to be a reliable assessment method. We tried to find out the influence of experience and orientation on interpretation of foot radiograph. AP and lateral radiograph from 25 patients who had some form of foot injury were included in this study. The radiographs were first analysed by an experienced foot surgeons for fractures with clinical details to facilitate the identification of the fracture & then presented in two different orientations (vertical and horizontal) to 38 orthopaedics surgeons of different grades and blinded about the nature of the study. We used Kappa analysis & logistic regression to find out the influence of orientation and experience on interpretation skills. Overall agreement between the foot surgeon and other observers was 0.4. Overall agreement between the observers and the foot surgeon, in the vertical orientation, was.43 and in the horizontal orientation was.32. When experience was taken into consideration agreement between orthopaedics surgeons and the foot surgeon was.416 and agreement between registrar and foot surgeon was.367 and the SHO and foot surgeon was.369. We found that the regression coefficient for horizontal orientation was −.302 and the regression coefficient for the experience was 0.067. Hence horizontal orientation increases the chance of misinterpreting a foot radiograph by 30.2% and with increasing experience the chance of correctly identifying the fracture increases by 6.7%. From this study it is evident with all the other variables like experience, quality and adequacy of radiographs optimised, orientation of the foot radiograph alone can significantly affect the interpretation skills of the observer

The study evaluated and compared the three-dimensional (3-D) changes in geometry of the first metatarsal following scarf osteotomy. All osteotomies were performed on standardised Sawbone® models by consultant orthopaedic surgeons with a sub-specialist interest in foot and ankle surgery. The study considered the inter-surgeon variances in interpretation and performance of the scarf osteotomy with respect to intra-surgeon variances. The analysis used an accurate digitising system to measure and record points on the Sawbone® models in 3-D space. Computer software performed vector analysis to calculate 3-D rotations and translations of the first metatarsal head as well as the inter-metatarsal angle. Bone cut lengths and displacements were measured using a digital Vernier caliper. One surgeon performed the osteotomy ten times to form an intra-surgeon control dataset, while ten different surgeons each did one scarf osteotomy to form an inter-surgeon test dataset. Both surgical groups produced reductions in the 3-D inter-metatarsal angle with non-significant differences between the groups (p> 0.05). In contrast, the test group demonstrated highly significant (p=0.000) greater variance compared with the control dataset for all of the variables (bone cut length, proximal and distal metatarsal displacements plus angulation of the distal fragment) associated with surgical technique. In addition, there were highly significant (p=0.02 and p=0.002) greater variances in the interpretation of the degree to which the metatarsal head should be translated medially (X) and inferiorly (Z). There was also a significant (p=0.001) increase in variances in the rotations about the dorsi/plantarflexion (X) axis. The only significant differences (all p=0.000) attributable solely to differences in mean values were in proximal-distal (Y) translation, pronation (Y) rotation and medial (Z) rotation. The test group applied greater medial and plantarflexion rotation of the metatarsal head than the control surgeon and significantly less (p=0.000) shortening of the first metatarsal than the control surgeon. The results of this geometric study demonstrate the versatility of the scarf osteotomy. In addition, it indicated notable out-of-plane metatarsal head rotations and translations effected by the scarf osteotomy. As a result of the multi-planar nature of the osteotomy, there is a potential risk of producing unintended rotational mal-unions in all three planes. These rotational mal-unions may account for some of the poorer outcomes documented within the literature

Aim. Current guidelines for the diagnosis of prosthetic joint infection (PJI) recommend collecting 4–5 independent tissue specimens, with isolation of indistinguishable organisms from two or more specimens. The same principle has been applied to other orthopaedic device-related infections (DRI) including fracture-related infections. However there are few published data validating this approach in DRI other than PJI. We evaluated the performance of different diagnostic cutoffs and varying numbers of tissue specimens for microbiological sampling in fracture-related infections. Method. We used standard protocols for tissue sample collection and laboratory processing, and a standard clinical definition of fracture-related infection. We explored how tissue culture sensitivity and specificity varied with the number of tissue specimens obtained; and with the number of specimens from which an identical isolate was required (diagnostic cutoff). To model the effect of the number of specimens taken we randomly sampled n specimens from those obtained at each procedure, excluding procedures from which less than n specimens were collected, and calculated sensitivity and specificity based on this sample. For each value of n we repeated this process 100 times to estimate the mean sensitivity and specificity for n specimens. Results. We analysed data for 246 cases of suspected fracture-related infection. 77 (31%) met the clinical definition of infection. A median of 4 independent tissue samples were obtained from each procedure (IQR 4–5). Culture sensitivity was highest and specificity lowest using a diagnostic cutoff of 1 specimen for isolation of an organism; specificity increased at the expense of sensitivity with diagnostic cutoffs of 2 or 3 specimens. Culture sensitivity increased as the number of tissue specimens obtained increased from 1 to 4. Although there was a corresponding decline in specificity with increasing numbers of tissue specimens obtained, this was negligible when a diagnostic cutoff of 2 or 3 specimens with identical organisms was used. Using a cutoff of 2 specimens with identical organisms, obtaining 4 specimens gave a sensitivity of 68% (55–78%) and a specificity of 95% (86–99%). Small numbers prevented meaningful analysis of the diagnostic performance of five or more specimens. Conclusions. These data are analogous to findings in prosthetic joint infections, and suggest similar principles may be applied to tissue sampling and culture interpretation in other orthopaedic DRI including fracture-related infection. A larger study is underway to evaluate the performance of greater numbers of tissue specimens

We performed a systematic review of the literature to evaluate the use and interpretation of generic and disease-specific functional outcome instruments in the reporting of outcome after the surgical treatment of disruptions of the pelvic ring. A total of 28 papers met our inclusion criteria, with eight reporting only generic outcome instruments, 13 reporting only pelvis-specific outcome instruments, and six reporting both. The Short-Form 36 (SF-36) was by far the most commonly used generic outcome instrument, used in 12 papers, with widely variable reporting of scores. The pelvis-specific outcome instruments were used in 19 studies; the Majeed score in ten, Iowa pelvic score in six, Hannover pelvic score in two and the Orlando pelvic score in one. Four sets of authors, all testing construct validity based on correlation with the SF-36, performed psychometric testing of three pelvis-specific instruments (Majeed, IPS and Orlando scores). No testing of responsiveness, content validity, criterion validity, internal consistency or reproducibility was performed. The existing literature in this area is inadequate to inform surgeons or patients in a meaningful way about the functional outcomes of these fractures after fixation

We propose a grading system for contrast free MRI images of tennis elbow and evaluate the inter and intra observer variability of their interpretation. Methods: Three senior orthopaedic surgeons were asked to blindly grade 0.2T dedicated extremity contrast free MRI images of elbows of patients who presented with varying degrees of symptomatic tennis elbow. Our proposed grading system of 1 to 5 based on the pattern around the common extensor tendon was used. Images of the symptomatic and contralateral non symptomatic elbows were graded blindly twice with an interval of 1 month by each surgeon. Each surgeon graded 176 MRI images twice. The grades were subsequently grouped into (I) grades 1 to 2 and (II) grades 3 to 5. Results: With regards to the intra observer agreement, consultant A showed 90.1% agreement, consultant B showed 90.6% agreement and consultant C 96.0% agreement. The mean intra observer agreement rate was 92.2%. The inter observer agreement between consultant A and B was 82.46%, between A and C 67.1% and between B and C 80.1%. It was also noted that there were systematic differences to the inter observer variability. Consultant A graded the images 3 to 5 on both occasions 52.9% of the time, consultant B graded 3 to 5 on both occasions 37.8% of the time and consultant C graded 3 to 5 on both occasions 23.3% of the time. Conclusion: The intra observer agreement rate is high. There is however a greater inter observer variation but this variation is consistent. We suggest that the inter observer differences can be improved by (1) reducing the grades to positive or negative and (2) by group reeducation of the observers

Introduction. High Tibial Osteotomy has become an increasingly popular management option for patients with painful medial compartment osteoarthritis. The Fujisawa method used to calculate the angle of correction is well-documented but there have been no studies to look at the reliability and accuracy of web-based systems to calculate this angle. Patients and Methods. Patients undergoing valgus high tibial osteotomy between October 2004 and February 2010 who had full-length lower-limb views on the Picture Archiving and Communications System (PACS). The Fujisawa angle and length of osteotomy were calculated by the surgeon and two Orthopaedic registrars who had been appropriately trained. Results. Thirty X-rays were reviewed in 28 patients. Mean difference between measurements was 0.43 mm (SD 2.45) There was a statistically significant correlation between all three raters (P < 0.001). The greatest correlation was between the Consultant and the more senior trainee (r=0.86) with the lowest correlation between the Consultant and the more junior trainee (r=0.70). Concordance correlation coefficient between raters varied from 0.81 to 0.63. Bland-Altman plot of agreement between the Consultant and senior trainee was excellent showing only two values lying outside 1.96 SD. Discussion. Comparison of measurements between raters showed increased accuracy with greater experience of the measurement technique. Despite this we show high correlation between raters with measurement accuracy of all raters found within clinically acceptable limits (< 1mm). Conclusion. Measurement of tibial osteotomy using the fujisawa method on a web-based X-ray interpretation system is accurate and reliable

Study Design: A retrospective cross sectional cohort study of degeneration of the lumbar spine, using pre- and post-discography MRI scans of 28 patients, as compared to two consecutive MRI scans of an age and sex matched control group of 32 patients. Objective: To determine whether injection of steroid into a lumbar intervertebral disc causes degeneration, as assessed by magnetic resonance imaging (MRI). Methods: Twenty-eight patients with chronic discogenic low back pain were selected. Each had been investigated with an MRI, discography (with intradiscal injection of methylprednisolone), and a post-discography MRI scan. A randomly selected control group of thirty-two age and sex matched patients, having been examined on two occasions with MRI, was established. Two interpreters blinded to the patient groups assessed the degree of lumbar disc degeneration on the MRI scans on two separate occasions, using the Pfirrmann grading system. Results: Kappa values proved interpretation consistency as compared with the published Pfirrmann paper. Variables of age and sex in the two groups showed no true variation in whether the discs improved, stayed the same or degenerated. The difference in the proportions for those with and without the injection gave a test statistic of 11.92 (p-value=0.002), indicating a discrepancy in the degeneration between those discs with and without an injection. Conclusion: Previous studies on intradiscal steroid injections have shown variable results. Animal studies have shown that steroid administered intradiscally causes degeneration and primary calcification in discs. Two prospective double blind clinical trials using intradiscal steroids identified no significant benefit or improvement in the clinical outcome. This study indicates that intradiscal steroid injections cause MRI visible disc degeneration. In association with the results of the clinical trials, this study questions the indications for the use of intradiscal steroids in the management of discogenic low back pain

We obtained information from the Elective Orthopaedic Centre on 1523 patients with baseline and six-month Oxford hip scores (OHS) after undergoing primary hip replacement (THR) and 1784 patients with Oxford knee scores (OKS) for primary knee replacement (TKR) who completed a six-month satisfaction questionnaire.

Receiver operating characteristic curves identified an absolute change in OHS of 14 points or more as the point that discriminates best between patients’ satisfaction levels and an 11-point change for the OKS. Satisfaction is highest (97.6%) in patients with an absolute change in OHS of 14 points or more, compared with lower levels of satisfaction (81.8%) below this threshold. Similarly, an 11-point absolute change in OKS was associated with 95.4% satisfaction compared with 76.5% below this threshold. For the six-month OHS a score of 35 points or more distinguished patients with the highest satisfaction level, and for the six-month OKS 30 points or more identified the highest level of satisfaction. The thresholds varied according to patients’ pre-operative score, where those with severe pre-operative pain/function required a lower six-month score to achieve the highest levels of satisfaction.

Our data suggest that the choice of a six-month follow-up to assess patient-reported outcomes of THR/TKR is acceptable. The thresholds help to differentiate between patients with different levels of satisfaction, but external validation will be required prior to general implementation in clinical practice.

The histopathology of periprosthetic tissues has been important to understanding the relationship between wear debris and arthroplasty outcome. In a landmark 1977paper, Willert and Semlitsch (1) used a semiquantitative rating to show that tissue reactions largely reflected the extent of particulate debris. Notably, small amounts of debris, including metal, could be eliminated without “overstraining the tissues” but excess debris led to deleterious changes. Currently, a plethora of terms is used to describe tissues from metal-on-metal (M-M) hips and corroded modular connections. We reviewed the evaluation and reporting of local tissue reactions over time, and asked if a dose response has been found between metal and tissue features, and how the use of more standardized terms and quantitative methodologies could reduce the current confusion in terminology.

Purpose

The ability to correctly interpret quantitative results is a crucial skill developed in medical school and surgical residency. It demands a basic understanding of epidemiological principles and modes of presenting data. Yet, there has been little investigation into the efficacy of current teaching methods and areas of difficulty among orthopaedic residents.

Ultrasonography for rotator cuff disease is a cheap and non-invasive investigation. Our study investigates the tendon specific pathologies leading to misinterpretation of ultrasound findings and their implication for the surgical management.

On hundred and five consecutive patients who had an ultrasound scan by a single musculoskeletal radiologist and then underwent shoulder arthroscopy by a single shoulder surgeon for rotator cuff pathologies were included.

Surpraspinatus Tendon (SST): There was a sensitivity of 90%. The relatively low positive predictive value (76%) and specificity (42.5%) were influenced by a high number of false positives. This was a mixed group of 23 cases, in which ultrasonography had described either a full-thickness (FTT) or partial-thickness (PTT) tear when arthroscopy did not show any evidence for a cuff tear. Seven of these cases were described as FTT with dimensions less than 1 cm and in ten cases the radiologist described a “possible sub-centimetre tear”. Subscapularis Tendon (SSC): There was a specificity of 100%. The poor negative predictive value (78%) and sensitivity (26%) were caused by a high number of false negatives. Further analysis of the 20 “false negative” patients showed four FTT and sixteen PTT. All partial thickness tears involved the superior fibres of the subscapularis tendon.

Our results confirm that USG is a reliable investigation in larger full thickness tears, particularly of the superior rotator cuff (SST). The reliability is significantly reduced in sub-centimetre tears and partial thickness tears, particularly of the subscapularis tendon. Associated tendon pathologies like intra-tendinous calcifications and intra-substance tears make an accurate diagnosis even more difficult and add to the tendency to ‘over-diagnose’ tears of the rotator cuff with use of ultrasonography.

The shoulder surgeon should be aware of the potential misinterpretation of ultrasonography findings and be prepared to adjust the surgical procedure accordingly.

Background

The increasing desire to protect the periarticular structures led the need of a Tissue Sparing Surgery. The accesses most widely used are the direct-lateral approach and the postero-lateral one, both with patient in lateral decubitus. Aim: This accesses require however an incision of tendons and muscles even in their minimally invasive technique, so we looked for an approach that would wholly protect the periarticular structures and allow us not to revise our experience in patient positioning, preparation of the operating field and surgeon's position during surgery. Our intent was to leave the acquired knowledge unchanged and to preserve unaltered the anatomical landmarks that we had previously identified and consolidated for the correct positioning of the components.

Purpose

The main predictors in the literature of local control for patients operated on for a soft tissue sarcoma are age, local presentation status, depth, grade, size, surgical margins and radiation. However, due to the competing effect of death (patients who die are withdrawn from the risk of local recurrence), the influence of these predictors on the cumulative probabilities may have been misinterpreted so far. The objective of the study was to interpret the influence of known predictors of local recurrence in a competing risks setting.

Left-right skeletal length asymmetries in upper limbs related to curve side have been detected with adolescent thoracic idiopathic scoliosis (AIS). In school screening referrals with thoracic scoliosis we find apical vertebral rotation (AVR, Perdriolle) is associated significantly with upper arm length asymmetry. Sixty-nine of 218 consecutive adolescent patients referred routinely during 1988–1999 had idiopathic thoracic scoliosis of whom 61 had left and right upper arm lengths measured with a Holtain anthropometer (right curves 49, left curves 12, mean age 14.9 years, girls 38 postmenarcheal 34, boys 23). The controls are 278 normal girls and 281 boys (11–18 years, mean age 13.5 years). The mean value for Cobb angle is 18 degrees (range 4–42 degrees), AVR 13 (range 0–34 degrees), Cobb angle (CA) and AVR are each positively associated with upper arm length asymmetry (p=0.001 & p< 0.0001 respectively) and after correcting for each of Cobb side, apical level, sex and handedness, AVR and upper arm length asymmetry are still significantly associated (p=0.004 ANOVA). Partial correlation analysis shows AVR is associated with upper arm length asymmetry after controlling for CA (p=0.033); but not CA and upper arm length asymmetry after controlling for AVR (p=0.595). The reason why a larger AVR to the right is associated with a relatively longer right upper arm is unknown. Possibilities include neuromuscular and skeletal mechanisms, the latter relative concave overgrowth of neurocentral synchondrosis and/or of periapical ribs. We suggest consideration be given to combining convex vertebral body stapling (Betz) with concave periapical rib resection (Sevastik and Xiong) for right thoracic AIS in girls.

Diagnostic interpretation error of paediatric musculoskeletal (MSK) radiographs can lead to late presentation of injuries that subsequently require more invasive surgical interventions with increased risks of morbidity. We aimed to determine the radiograph factors that resulted in diagnostic interpretation challenges for emergency physicians reviewing pediatric MSK radiographs. Emergency physicians provided diagnostic interpretations on 1,850 pediatric MSK radiographs via their participation in a web-based education platform. From this data, we derived interpretation difficulty scores for each radiograph using item response theory. We classified each radiograph by body region, diagnosis (fracture/dislocation absent or present), and, where applicable, the specific fracture location(s) and morphology(ies). We compared the interpretation difficulty scores by diagnosis, fracture location, and morphology. An expert panel reviewed the 65 most commonly misdiagnosed radiographs without a fracture/dislocation to identify normal imaging findings that were commonly mistaken for fractures. We included data from 244 emergency physicians, which resulted in 185,653 unique radiograph interpretations, 42,689 (23.0%) of which were diagnostic errors. For humerus, elbow, forearm, wrist, femur, knee, tibia-fibula radiographs, those without a fracture had higher interpretation difficulty scores relative to those with a fracture; the opposite was true for the hand, pelvis, foot, and ankle radiographs (p < 0 .004 for all comparisons). The descriptive review demonstrated that specific normal anatomy, overlapping bones, and external artefact from muscle or skin folds were often mistaken for fractures. There was a significant difference in difficulty score by anatomic locations of the fracture in the elbow, pelvis, and ankle (p < 0 .004 for all comparisons). Ankle and elbow growth plate, fibular avulsion, and humerus condylar were more difficult to diagnose than other fracture patterns (p < 0 .004 for all comparisons). We identified actionable learning opportunities in paediatric MSK radiograph interpretation for emergency physicians. We will use this information to design targeted education to referring emergency physicians and their trainees with an aim to decrease delayed and missed paediatric MSK injuries