Effects of a comprehensive brain computed tomography deep learning model on radiologist detection accuracy.

OBJECTIVES: Non-contrast computed tomography of the brain (NCCTB) is commonly used to detect intracranial pathology but is subject to interpretation errors. Machine learning can augment clinical decision-making and improve NCCTB scan interpretation. This retrospective detection accuracy study assessed the performance of radiologists assisted by a deep learning model and compared the standalone performance of the model with that of unassisted radiologists. METHODS: A deep learning model was trained on 212,484 NCCTB scans drawn from a private radiology group in Australia. Scans from inpatient, outpatient, and emergency settings were included. Scan inclusion criteria were age ≥ 18 years and series slice thickness ≤ 1.5 mm. Thirty-two radiologists reviewed 2848 scans with and without the assistance of the deep learning system and rated their confidence in the presence of each finding using a 7-point scale. Differences in AUC and Matthews correlation coefficient (MCC) were calculated using a ground-truth gold standard. RESULTS: The model demonstrated an average area under the receiver operating characteristic curve (AUC) of 0.93 across 144 NCCTB findings and significantly improved radiologist interpretation performance. Assisted and unassisted radiologists demonstrated an average AUC of 0.79 and 0.73 across 22 grouped parent findings and 0.72 and 0.68 across 189 child findings, respectively. When assisted by the model, radiologist AUC was significantly improved for 91 findings (158 findings were non-inferior), and reading time was significantly reduced. CONCLUSIONS: The assistance of a comprehensive deep learning model significantly improved radiologist detection accuracy across a wide range of clinical findings and demonstrated the potential to improve NCCTB interpretation. CLINICAL RELEVANCE STATEMENT: This study evaluated a comprehensive CT brain deep learning model, which performed strongly, improved the performance of radiologists, and reduced interpretation time. The model may reduce errors, improve efficiency, facilitate triage, and better enable the delivery of timely patient care. KEY POINTS: • This study demonstrated that the use of a comprehensive deep learning system assisted radiologists in the detection of a wide range of abnormalities on non-contrast brain computed tomography scans. • The deep learning model demonstrated an average area under the receiver operating characteristic curve of 0.93 across 144 findings and significantly improved radiologist interpretation performance. • The assistance of the comprehensive deep learning model significantly reduced the time required for radiologists to interpret computed tomography scans of the brain.

Machine Learning Augmented Interpretation of Chest X-rays: A Systematic Review.

Limitations of the chest X-ray (CXR) have resulted in attempts to create machine learning systems to assist clinicians and improve interpretation accuracy. An understanding of the capabilities and limitations of modern machine learning systems is necessary for clinicians as these tools begin to permeate practice. This systematic review aimed to provide an overview of machine learning applications designed to facilitate CXR interpretation. A systematic search strategy was executed to identify research into machine learning algorithms capable of detecting >2 radiographic findings on CXRs published between January 2020 and September 2022. Model details and study characteristics, including risk of bias and quality, were summarized. Initially, 2248 articles were retrieved, with 46 included in the final review. Published models demonstrated strong standalone performance and were typically as accurate, or more accurate, than radiologists or non-radiologist clinicians. Multiple studies demonstrated an improvement in the clinical finding classification performance of clinicians when models acted as a diagnostic assistance device. Device performance was compared with that of clinicians in 30% of studies, while effects on clinical perception and diagnosis were evaluated in 19%. Only one study was prospectively run. On average, 128,662 images were used to train and validate models. Most classified less than eight clinical findings, while the three most comprehensive models classified 54, 72, and 124 findings. This review suggests that machine learning devices designed to facilitate CXR interpretation perform strongly, improve the detection performance of clinicians, and improve the efficiency of radiology workflow. Several limitations were identified, and clinician involvement and expertise will be key to driving the safe implementation of quality CXR machine learning systems.

Persistent knee extension deficits are common after anterior cruciate ligament reconstruction: a systematic review and meta-analysis of randomised controlled trials.

PURPOSE: Knee extension deficits complicate recovery from ACL injury and reconstruction; however, the incidence of knee extension loss is not well defined. The aim of this review was to identify the incidence of loss of extension (LOE) following ACL rupture and reconstruction, explore the definitions of knee extension deficits reported and identify prognostic factors affecting LOE incidence. METHODS: A systematic search was conducted in Medline, Cochrane Library and PEDro for studies in publication up to November 2021, with no restrictions on publication year. References were screened and assessed for inclusion using predetermined eligibility criteria. Randomised controlled trials (RCTs) that quantified knee angle, loss of extension or incidence of extension deficit were included for quality assessment and data extraction. Statistical summaries were generated and meta-analyses performed in two parts to examine: (i) the probability of a datapoint being zero incidence compared to a nonzero incidence and (ii) the relationship between the predictors and nonzero LOE incidence. RESULTS: A sample of 15,494 studies were retrieved using the search criteria, with 53 studies meeting eligibility criteria. The pooled results from 4991 participants were included for analysis, with 4891 participants who had undergone ACLR. The proportion of included studies judged at an overall low risk of bias was small (7.8%). The observed group and study were the most important predictors for whether a datapoint reported an incidence of extension deficit. Time to follow-up (P < 0.001) and graft type (P = 0.02) were found to have a significant influence on nonzero LOE incidence (%). Covariate adjusted estimates of average LOE indicated 1 in 3 patients presenting with LOE at 12 month follow-up, reducing to 1 in 4 at 2 years. CONCLUSIONS: This review examined the definitions for the measurement and interpretation of postoperative knee extension and established the trajectory of knee extension deficit after ACL injury and reconstruction. While factors associated with loss of extension were identified, the exact trajectory of knee extension deficits was difficult to infer due to discrepancies in measurement techniques and patient variation. On average, 1 in 3 patients may present with loss of extension of at least 3 degrees at 12-month follow-up, decreasing to 1 in 4 at 2 years. These results may be used by clinicians as an upper threshold for acceptable complication rates following ACLR. Future work should focus on LOE as a clinically relevant complication of ACL injury and treatment with appropriate attention to standardisation of definitions, measurements and better understanding of natural history. PROSPERO REGISTRATION NUMBER: CRD42018092295. LEVEL OF EVIDENCE: Level I.

Early Clinical Outcomes and Complications of Transfibular Total Ankle Arthroplasty: The Australian Experience.

BACKGROUND: We report the short-term outcomes of a retrospective cohort of Trabecular Metal total ankle arthroplasties (TAAs) via a transfibular approach from a single, nondesigner surgeon in Australia. METHODS: This was a retrospective cohort study. Demographic, clinical, and patient-reported outcome measures (PROMs) were collected. The primary outcome was the Foot and Ankle Outcome Score (FAOS), and secondary outcomes included patient satisfaction, the EuroQol-5 Dimension (EQ5D), and complications including revision rates. RESULTS: Between 2016 and 2019, 84 trabecular metal prostheses were implanted in 84 patients. Mean age (SD) at time of surgery was 68 (7.8) years, and 46 (55%) were male. Mean follow-up (SD) was 26.1 (13) months. There were significant (P < .001) improvements in the FAOS in the subscales of pain (47.9 to 79.1), activities of daily living (59.5 to 83.7), and quality of life (25.5 to 60.2) and EQ-5D (0.55 to 0.75) (P < .001), and overall satisfaction was 69.6%. The commonest complications were wound infection or breakdown (11.9%, n=10), fibular nonunion (3.5%), plate irritation (3.5%), and tibial nerve neuropathy (3.5%). There were no thromboembolic complications. Implant survivorship was 100%, with Trabecular Metal components retained in all patients. Two patients developed deep infection, with 1 requiring debridement and polyethylene exchange. No patients experienced implant loosening. CONCLUSION: Our results demonstrate that the trabecular metal survival rates are comparable with other total ankle implants in the Australian Orthopaedic Association National Joint Replacement Registry and as published in other international literature. Overall patient satisfaction was high, as were PROMs. However, the data highlight potential complications uniquely associated with this implant. The authors believe that these figures support TAA via a transfibular approach as a viable option in the treatment of ankle arthritis. LEVEL OF EVIDENCE: Level IV, retrospective cohort study.

The effect of osteotomy technique (flat-cut vs wedge-cut Weil) on pain relief and complication incidence following surgical treatment for metatarsalgia in a private metropolitan clinic: protocol for a randomised controlled trial.

BACKGROUND: Weil osteotomies are performed to surgically treat metatarsalgia, by shortening the metatarsal via either a single distal oblique cut with translation of the metatarsal head (flat-cut) or through the removal of a slice of bone (wedge-cut). The wedge-cut technique purportedly has functional and mechanical advantages over the flat-cut procedure; however, in vivo data and quality of evidence are currently lacking. This study aims to investigate whether wedge-cut Weil osteotomy compared to traditional flat-cut Weil is associated with increased pain relief and fewer complications up to 12 months postoperatively. METHODS: Patient, surgical and clinical data will be collected for 80 consecutive consenting patients electing to undergo surgical treatment of propulsive metatarsalgia in a randomised control trial, embedded within a clinical registry. The primary outcome is patient-reported pain as assessed by the Foot and Ankle Outcome Score (FAOS) - Pain subscale, and the secondary outcome is the incidence of procedure-specific complications at up to 12 months postoperatively. The groups will be randomised using a central computer-based simple randomisation system, with a 1:1 allocation without blocking and allocation concealment. A mixed-effects analysis of covariance will be used to assess the primary outcome, with confounders factored into the model. A binary logistic regression will be used to assess the secondary outcome in a multivariable model containing the same confounders. DISCUSSION: To the best of the authors' knowledge, the trial will be the first to examine the clinical efficacy of the wedge-cut Weil osteotomy compared to the flat-cut technique with a prospective, randomised control design. TRIAL REGISTRATION: Australian New Zealand Clinical Trials Registry ACTRN12620001251910. Registered on 23 November 2020.

Validity of intraoperative imageless navigation (Naviswiss) for component positioning accuracy in primary total hip arthroplasty: protocol for a prospective observational cohort study in a single-surgeon practice.

INTRODUCTION: Optimal outcomes in total hip arthroplasty (THA) are dependent on appropriate placement of femoral and acetabular components, with technological advances providing a platform for guiding component placement to reduce the risk of malpositioned components during surgery. This study will validate the intraoperative data captured using a handheld imageless THA navigation system against postoperative measurements of acetabular inclination, acetabular version, leg length and femoral offset on CT radiographs. METHODS AND ANALYSIS: This is a prospective observational cohort study conducted within a single-centre, single-surgeon private practice. Data will be collected for 35 consecutive patients (>18 years) undergoing elective THA surgery, from the research registry established at the surgeon's practice. The primary outcome is the agreement between intraoperative component positioning data captured by the navigation system compared with postoperative measurements using CT. A total of ten CT scans will be reassessed for interobserver and intraobserver reliability. The influence of patient and surgical factors on the accuracy of component position will also be examined with multivariable linear regression. ETHICS AND DISSEMINATION: Ethics approval for this study was provided through a certified ethics committee (Bellberry HREC approval number 2017-07-499). The results of this study will be disseminated through peer-reviewed journals and conference presentations. TRIAL REGISTRATION: Australian and New Zealand Clinical Trials Registry (ANZCTR) Trial ID: ACTRN12620000089932.

Implementation and quality assessment of a clinical orthopaedic registry in a public hospital department.

BACKGROUND: The aim of this study was to demonstrate a novel method of assessing data quality for an orthopaedic registry and its effects on data quality metrics. METHODS: A quality controlled clinical patient registry was implemented, comprising six observational cohorts of shoulder and knee pathologies. Data collection procedures were co-developed with clinicians and administrative staff in accordance with the relevant dataset and organised into the registry database software. Quality metrics included completeness, consistency and validity. Data were extracted at scheduled intervals (3 months) and quality metrics reported to stakeholders of the registry. RESULTS: The first patient was enrolled in July 2017 and the data extracted for analysis over 4 quarters, with the last audit in August 2018 (N = 189). Auditing revealed registry completeness was 100% after registry deficiencies were addressed. However, cohort completeness was less accurate, ranging from 12 to 13% for height & weight to 90-100% for operative variables such as operating surgeon, consulting surgeon and hospital. Consistency and internal validation improved to 100% after issues in registry processes were rectified. CONCLUSIONS: A novel method to assess data quality in a clinical orthopaedic registry identified process shortfalls and improved data quality over time. Real-time communication, a comprehensive data framework and an integrated feedback loop were necessary to ensure adequate quality assurance. This model can be replicated in other registries and serve as a useful quality control tool to improve registry quality and ensure applicability of the data to aid clinical decisions, especially in newly implemented registries. TRIAL REGISTRATION: ACTRN12617001161314; registration date 8/08/2017. Retrospectively registered.

Efficacy of a Second-Generation Rotating Bearing Tibial Platform in Total Knee Arthroplasty: A Prospective Observational Cohort Study with Registry Analysis.

There is a lack of clinical outcomes reported for the rotating bearing knee (RBK) total knee arthroplasty (TKA), which is a second-generation rotating platform knee, with purported benefits over earlier versions. The purpose of the study was to report the complications, short-term (minimum 1 year) patient-reported outcomes and long-term (up to 15 years) procedure survival in a consecutive series of patients receiving a rotating platform TKA (RBK) from an independent clinic. A retrospective analysis of a single-surgeon, private/public practice, with prospectively collected data in a subset of patients were performed. A total of 1,130 procedures (primary, revision from unicompartmental knee arthroplasty (UKA) to TKA) were crossmatched with manufacturer records. Clinical outcomes (complications, reoperations) were summarized and linked to patient-reported outcome measures (Eq. 5D, KSS-function, Oxford knee score [OKS]). OKS results were classified using minimally clinical important difference (MCID) and patient acceptable symptom state (PASS). PROMs were summarized and regression models used to determine relationships between patient factors and outcomes in this cohort. Cumulative percent revision was reported by the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) and compared between the senior author and national data using Kaplan-Meier survival analysis. We report a complication rate of 19.7% with the majority (> 60%) being thromboembolic events and complaints of stiffness. Significant improvements were observed in general health, knee pain, and function with > 89% exceeding the MCID for the OKS and > 65% exceeding the PASS for the OKS at an average follow-up of 3.2 years. We report a cumulative revision rate of 4.3% at 5 years and 4.8% at 14 years, with significantly lower revision rates in females and patients aged 55 to 64 years compared with AOANJRR data for fixed bearing designs. The RBK rotating platform TKA provides good functional outcomes, with relatively low revision and complications rates at up to 14 years follow-up. This design in conjunction with a gap balancing technique may be advantageous in certain patient subgroups.

Recovery of knee extension and incidence of extension deficits following anterior cruciate ligament injury and treatment: a systematic review protocol.

INTRODUCTION: Knee extension deficit or loss of extension (LOE) is a potential complication following ACL reconstruction (ACLR); however, the change in postoperative knee extension during rehabilitation is not well defined. The aim of this review is to establish the trajectory of knee extension recovery and incidence of knee extension deficit during rehabilitation after ACL rupture. METHODS AND ANALYSIS: A systematic search will be conducted in MEDLINE, Embase, Cochrane Library, Scopus, SPORTDiscus, and relevant trials databases of English language papers in publication as of May 2018, with no restrictions on publication year applied. References will be screened and assessed for eligibility by two independent reviewers as per the PRISMA guidelines. Cohort, cross-sectional or case-controlled studies will be included for the analysis. Data extraction will be conducted using a predefined template and quality of evidence assessed. Statistical summaries and meta-analyses will be performed as necessary. ETHICS AND DISSEMINATION: This review will provide clearer definitions for the measurement and interpretation of postoperative knee extension and establish its natural history after ACL reconstruction. Evidence of the incidence and factors associated with loss of extension will be identified. The findings of this systematic review will be disseminated in peer-reviewed journals and presented at national/international conferences. TRIAL REGISTRATION: The protocol was registered on the PROSPERO international prospective register of systematic reviews prior to commencement (registration number CRD42018092295 ).

Exercise responses during functional electrical stimulation cycling in individuals with spinal cord injury.

PURPOSE: This study compared acute exercise responses during arm cranking, functional electrical stimulation (FES)-assisted leg cycling, and combined arm and leg ("hybrid") cycling in individuals with spinal cord injury during maximal and submaximal exercise. METHODS: Nine male subjects with long-standing neurological lesions from C7 to T12 were recruited. All subjects performed arm crank ergometry (ACE), FES leg cycle exercise (FES-LCE), combined ACE + FES-LCE, and cycling on a hybrid FES tricycle (HYBRID). They were assessed for their peak exercise responses in all four modalities. Subsequently, their submaximal heart rates (HR), cardiac outputs (Q), stroke volumes (SV), and arteriovenous oxygen extractions (Ca-Cv)O2 were measured at 40%, 60%, and 80% of mode-specific V˙O2peak. RESULTS: Arm exercise alone and arm + leg exercise resulted in significantly higher V˙O2peak and HRpeak compared with FES-LCE (P < 0.05). Submaximal V˙O2 during FES-LCE was significantly lower than all other modalities across the range of exercise intensities (P < 0.05). ACE elicited 70%-94% higher steady-state V˙O2, and HYBRID evoked 99%-148% higher V˙O2 compared with FES-LCE. Steady-state FES-LCE also produced significantly lower Q, HR, and (Ca-Cv)O2. ACE evoked 31%-36% higher Q and 19%-47% greater HR than did FES-LCE. HYBRID elicited 31%-49% greater Q and 23%-56% higher HR than FES-LCE. CONCLUSIONS: Combined arm and leg exercise can develop a higher oxygen uptake and greater cardiovascular demand compared with ACE or FES-LCE alone. These findings suggested that combined arm + leg FES training at submaximal exercise intensities may lead to greater gains of aerobic fitness than would arm exercise alone. These data also proffered that FES leg cycling exercise by itself may be insufficient to promote aerobic fitness in the spinal cord injury population.