Campbell, B. C.V. et al. (2019) Penumbral imaging and functional outcome in patients with anterior circulation ischaemic stroke treated with endovascular thrombectomy versus medical therapy: a meta-analysis of individual patient-level data.ABSTRACT Background: CT-perfusion (CTP) and MRI may assist patient selection for endovascular thrombectomy. We aimed to establish whether imaging assessments of ischaemic core and penumbra volumes were associated with functional outcomes and treatment effect.

Campbell, B. C. V. et al. (2018) Effect of general anaesthesia on functional outcome in patients with anterior circulation ischaemic stroke having endovascular thrombectomy versus standard care: a meta-analysis of individual patient data. Lancet Neurology, 17(1), pp. 47-53. (doi:10.1016/S1474-4422(17)30407-6) This is the author's final accepted version.There may be differences between this version and the published version. You are advised to consult the publisher's version if you wish to cite from it.http://eprints.gla.ac.uk/149670/ variables. An alternative approach using propensity-score stratification was also used. To account for between-trial variance we used mixed-effects modeling with a random effect for trial incorporated in all models. Bias was assessed using the Cochrane tool.Findings: Of 1764 patients in 7 trials, 871 were allocated to endovascular thrombectomy. After exclusion of 74 patients (72 who did not undergo the procedure and 2 with missing data on anaesthetic strategy), 236/797 (30%) of endovascular patients were treated under GA. At baseline, GA patients were younger and had shorter time to randomisation but similar pre-treatment clinical severity compared to non-GA. Endovascular thrombectomy improved functional outcome at 3 months versus standard care in both GA (adjusted common odds ratio (cOR) 1·52, 95%CI 1·09-2·11, p=0·014) and non-GA (adjusted cOR 2·33, 95%CI 1·75-3·10, p<0·001) patients. However, outcomes were significantly better for those treated under non-GA versus GA (covariate-adjusted cOR 1·53, 95%CI 1·14-2·04, p=0·004; propensitystratified cOR 1·44 95%CI 1·08-1·92, p=0·012). The risk of bias and variability among studies was assessed to be low.Interpretation: Worse outcomes after endovascular thrombectomy were associated with GA, after adjustment for baseline prognostic variables. These data support avoidance of GA whenever possible. The procedure did, however, remain effective versus standard care in patients treated under GA, indicating that treatment should not be withheld in those who require anaesthesia for medical reasons. Funding:The HERMES collaboration was funded by an unrestricted grant from Medtronic to the University of Calgary. Research in contextEvidence before this study between abolition of the thrombectomy treatment effect in MR CLEAN and no effect in THRACE. Three single-centre randomised trials of general anaesthesia versus conscious sedation found either no difference in functional outcome between groups or a slight benefit of general anaesthesia. Added value of this studyThese data from contemporary, high quality randomised trials form the largest study to date of the association between general anesthesia and the benefit of endovascular thrombectomy versus standard care. We used two different approaches to adjust for baseline imbalances (multivariable logistic regression and propensity-score stratification). We found that GA for endovascular thrombectomy, as practiced in contemporary clinical care across a wide range of expert centres during the rand...

Introduction: To evaluate the accuracy of deep convolutional neural networks (DCNNs) for detecting neck of femur (NoF) fractures on radiographs, in comparison with perceptual training in medically-na€ ıve individuals. Methods: This study extends a previous study that conducted perceptual training in medically-na€ ıve individuals for the detection of NoF fractures on a variety of dataset sizes. The same anteroposterior hip radiograph dataset was used to train two DCNNs (AlexNet and GoogLeNet) to detect NoF fractures. For direct comparison with perceptual training results, deep learning was completed across a variety of dataset sizes (200, 320 and 640 images) with images split into training (80%) and validation (20%). An additional 160 images were used as the final test set. Multiple pre-processing and augmentation techniques were utilised. Results: AlexNet and GoogLeNet DCNNs NoF fracture detection accuracy increased with larger training dataset sizes and mildly with augmentation. Accuracy increased from 81.9% and 88.1% to 89.4% and 94.4% for AlexNet and GoogLeNet respectively. Similarly, the test accuracy for the perceptual training in top-performing medically-na€ ıve individuals increased from 87.6% to 90.5% when trained on 640 images compared with 200 images. Conclusions: Single detection tasks in radiology are commonly used in DCNN research with their results often used to make broader claims about machine learning being able to perform as well as subspecialty radiologists. This study suggests that as impressive as recognising fractures is for a DCNN, similar learning can be achieved by top-performing medically-na€ ıve humans with less than 1 hour of perceptual training.

Background Lung cancer is a major health problem. CT lung screening can reduce lung cancer mortality through early diagnosis by at least 20%. Screening high-risk individuals is most effective. Retrospective analyses suggest that identifying individuals for screening by accurate prediction models is more efficient than using categorical agesmoking criteria, such as the US Preventive Services Task Force (USPSTF) criteria. This study prospectively compared the effectiveness of the USPSTF2013 and PLCOm2012 model eligibility criteria. MethodsIn this prospective cohort study, participants from the International Lung Screening Trial (ILST), aged 55-80 years, who were current or former smokers (ie, had ≥30 pack-years smoking history or ≤15 quit-years since last permanently quitting), and who met USPSTF2013 criteria or a PLCOm2012 risk threshold of at least 1•51% within 6 years of screening, were recruited from nine screening sites in Canada, Australia, Hong Kong, and the UK. After enrolment, patients were assessed with the USPSTF2013 criteria and the PLCOm2012 risk model with a threshold of at least 1•70% at 6 years. Data were collected locally and centralised. Main outcomes were the comparison of lung cancer detection rates and cumulative life expectancies in patients with lung cancer between USPSTF2013 criteria and the PLCOm2012 model. In this Article, we present data from an interim analysis. To estimate the incidence of lung cancers in individuals who were USPSTF2013-negative and had PLCOm2012 of less than 1•51% at 6 years, ever-smokers in the Prostate Lung Colorectal and Ovarian Cancer Screening Trial (PLCO) who met these criteria and their lung cancer incidence were applied to the ILST sample size for the mean follow-up occurring in the ILST. This trial is registered at ClinicalTrials.gov, NCT02871856. Study enrolment is almost complete.