Comparison of computational fluid dynamics with transcranial Doppler ultrasound in response to physiological stimuli

Cerebrovascular haemodynamics are sensitive to multiple physiological stimuli that require synergistic response to maintain adequate perfusion. Understanding haemodynamic changes within cerebral arteries is important to inform how the brain regulates perfusion; however, methods for direct measurement of cerebral haemodynamics in these environments are challenging. The aim of this study was to assess velocity waveform metrics obtained using transcranial Doppler (TCD) with flow-conserving subject-specific three-dimensional (3D) simulations using computational fluid dynamics (CFD). Twelve healthy participants underwent head and neck imaging with 3 T magnetic resonance angiography. Velocity waveforms in the middle cerebral artery were measured with TCD ultrasound, while diameter and velocity were measured using duplex ultrasound in the internal carotid and vertebral arteries to calculate incoming cerebral flow at rest, during hypercapnia and exercise. CFD simulations were developed for each condition, with velocity waveform metrics extracted in the same insonation region as TCD. Exposure to stimuli induced significant changes in cardiorespiratory measures across all participants. Measured absolute TCD velocities were significantly higher than those calculated from CFD (P range < 0.001-0.004), and these data were not correlated across conditions (r range 0.030-0.377, P range 0.227-0.925). However, relative changes in systolic and time-averaged velocity from resting levels exhibited significant positive correlations when the distinct techniques were compared (r range 0.577-0.770, P range 0.003-0.049). Our data indicate that while absolute measures of cerebral velocity differ between TCD and 3D CFD simulation, physiological changes from resting levels in systolic and time-averaged velocity are significantly correlated between techniques.

Paying attention to cardiac surgical risk: An interpretable machine learning approach using an uncertainty-aware attentive neural network

Machine learning (ML) is increasingly applied to predict adverse postoperative outcomes in cardiac surgery. Commonly used ML models fail to translate to clinical practice due to absent model explainability, limited uncertainty quantification, and no flexibility to missing data. We aimed to develop and benchmark a novel ML approach, the uncertainty-aware attention network (UAN), to overcome these common limitations. Two Bayesian uncertainty quantification methods were tested, generalized variational inference (GVI) or a posterior network (PN). The UAN models were compared with an ensemble of XGBoost models and a Bayesian logistic regression model (LR) with imputation. The derivation datasets consisted of 153,932 surgery events from the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZSCTS) Cardiac Surgery Database. An external validation consisted of 7343 surgery events which were extracted from the Medical Information Mart for Intensive Care (MIMIC) III critical care dataset. The highest performing model on the external validation dataset was a UAN-GVI with an area under the receiver operating characteristic curve (AUC) of 0.78 (0.01). Model performance improved on high confidence samples with an AUC of 0.81 (0.01). Confidence calibration for aleatoric uncertainty was excellent for all models. Calibration for epistemic uncertainty was more variable, with an ensemble of XGBoost models performing the best with an AUC of 0.84 (0.08). Epistemic uncertainty was improved using the PN approach, compared to GVI. UAN is able to use an interpretable and flexible deep learning approach to provide estimates of model uncertainty alongside state-of-the-art predictions. The model has been made freely available as an easy-to-use web application demonstrating that by designing uncertainty-aware models with innately explainable predictions deep learning may become more suitable for routine clinical use.

Risk of venous thromboembolism in knee, hip and hand osteoarthritis: a general population-based cohort study

OBJECTIVES

Osteoarthritis is a leading cause of immobility and joint replacement, two strong risk factors for venous thromboembolism (VTE). We aimed to examine the relation of knee, hip and hand osteoarthritis to the risk of VTE and investigate joint replacement as a potential mediator.

METHODS

We conducted three cohort studies using data from The Health Improvement Network. Up to five individuals without osteoarthritis were matched to each case of incident knee (n=20 696), hip (n=10 411) or hand (n=6329) osteoarthritis by age, sex, entry time and body mass index. We examined the relation of osteoarthritis to VTE (pulmonary embolism and deep vein thrombosis) using a multivariable Cox proportional hazard model.

RESULTS

VTE developed in 327 individuals with knee osteoarthritis and 951 individuals without osteoarthritis (2.7 vs 2.0 per 1000 person-years), with multivariable-adjusted HR being 1.38 (95% CI 1.23 to 1.56). The indirect effect (HR) of knee osteoarthritis on VTE through knee replacement was 1.07 (95% CI 1.01 to 1.15), explaining 24.8% of its total effect on VTE. Risk of VTE was higher in hip osteoarthritis than non-osteoarthritis (3.3 vs 1.8 per 1000 person-years; multivariable-adjusted HR=1.83, 95% CI 1.56 to 2.13). The indirect effect through hip replacement yielded an HR of 1.14 (95% CI 1.04 to 1.25), explaining 28.1% of the total effect. No statistically significant difference in VTE risk was observed between hand osteoarthritis and non-osteoarthritis (1.5 vs 1.6 per 1000 person-years; multivariable-adjusted HR=0.88, 95% CI 0.67 to 1.16).

CONCLUSION

Our large population-based cohort study provides the first evidence that knee or hip osteoarthritis, but not hand osteoarthritis, was associated with an increased risk of VTE, and such an association was partially mediated through knee or hip replacement.

Do intrapersonal factors mediate the association of social support with physical activity in young women living in socioeconomically disadvantaged neighbourhoods? A longitudinal mediation analysis

Background

Levels of physical activity (PA) decrease when transitioning from adolescence into young adulthood. Evidence suggests that social support and intrapersonal factors (self-efficacy, outcome expectations, PA enjoyment) are associated with PA. The aim of the present study was to explore whether cross-sectional and longitudinal associations of social support from family and friends with leisure-time PA (LTPA) among young women living in disadvantaged areas were mediated by intrapersonal factors (PA enjoyment, outcome expectations, self-efficacy).

Methods

Survey data were collected from 18–30 year-old women living in disadvantaged suburbs of Victoria, Australia as part of the READI study in 2007–2008 (T0, N = 1197), with follow-up data collected in 2010–2011 (T1, N = 357) and 2012–2013 (T2, N = 271). A series of single-mediator models were tested using baseline (T0) and longitudinal data from all three time points with residual change scores for changes between measurements.

Results

Cross-sectional analyses showed that social support was associated with LTPA both directly and indirectly, mediated by intrapersonal factors. Each intrapersonal factor explained between 5.9–37.5% of the associations. None of the intrapersonal factors were significant mediators in the longitudinal analyses.

Conclusions

Results from the cross-sectional analyses suggest that the associations of social support from family and from friends with LTPA are mediated by intrapersonal factors (PA enjoyment, outcome expectations and self-efficacy). However, longitudinal analyses did not confirm these findings.