Probability estimation with machine learning methods for dichotomous and multicategory outcome: applications.

Machine learning methods are applied to three different large datasets, all dealing with probability estimation problems for dichotomous or multicategory data. Specifically, we investigate k-nearest neighbors, bagged nearest neighbors, random forests for probability estimation trees, and support vector machines with the kernels of Bessel, linear, Laplacian, and radial basis type. Comparisons are made with logistic regression. The dataset from the German Stroke Study Collaboration with dichotomous and three-category outcome variables allows, in particular, for temporal and external validation. The other two datasets are freely available from the UCI learning repository and provide dichotomous outcome variables. One of them, the Cleveland Clinic Foundation Heart Disease dataset, uses data from one clinic for training and from three clinics for external validation, while the other, the thyroid disease dataset, allows for temporal validation by separating data into training and test data by date of recruitment into study. For dichotomous outcome variables, we use receiver operating characteristics, areas under the curve values with bootstrapped 95% confidence intervals, and Hosmer-Lemeshow-type figures as comparison criteria. For dichotomous and multicategory outcomes, we calculated bootstrap Brier scores with 95% confidence intervals and also compared them through bootstrapping. In a supplement, we provide R code for performing the analyses and for random forest analyses in Random Jungle, version 2.1.0. The learning machines show promising performance over all constructed models. They are simple to apply and serve as an alternative approach to logistic or multinomial logistic regression analysis.

Early and comprehensive management of atrial fibrillation: executive summary of the proceedings from the 2nd AFNET-EHRA consensus conference 'research perspectives in AF'.

Atrial fibrillation (AF) causes important mortality and morbidity on a population-level. So far, we do not have the means to prevent AF or AF-related complications adequately. Therefore, over 70 experts on atrial fibrillation convened for the 2nd AFNET/EHRA consensus conference to suggest directions for research to improve management of AF patients (Appendix 1). The group defined three main areas in need for research in AF: 1. better understanding of the mechanisms of AF; 2. Improving rhythm control monitoring and management; and 3. comprehensive cardiovascular risk management in AF patients. The group put forward the hypothesis that successful therapy of AF and its associated complications will require comprehensive therapy. This applies e.g. to the "old" debate of "rate versus rhythm control", since rhythm control is generally added to underlying (continued) rate control therapy, but also to the emerging debate of "antiarrhythmic drugs versus catheter ablation", of which both may be needed in most patients to maintain sinus rhythm, but also to therapy of conditions that predispose to AF and contribute to cardiovascular complications such as stroke, cognitive decline, heart failure, and acute coronary syndromes. We call for research initiatives aiming at a better understanding of the different causes of AF and its complications, and at development and validation of mechanism-based therapies. The future of AF therapy may require a combination of management of underlying and concomitant conditions, early and comprehensive rhythm control therapy, adequate control of ventricular rate and cardiac function, and continuous therapy to prevent AF-associated complications (e.g. antithrombotic therapy). The reasons for these suggestions are detailed in this paper.

Evaluation of brain injury after coronary artery bypass grafting. A prospective study using neuropsychological assessment and diffusion-weighted magnetic resonance imaging.

OBJECTIVE: Neurocognitive dysfunction is a common complication after cardiac surgery with cardiopulmonary bypass (CPB). Studies using magnetic resonance imaging (MRI) have demonstrated that new focal brain lesions can occur after coronary artery bypass grafting (CABG), even in patients without apparent neurological deficits. Diffusion-weighted MRI is superior to conventional MRI and allows for sensitive and early detection of ischemic brain lesions. We prospectively investigated cerebral injury early and 3 months after CABG using diffusion-weighted MRI and related the findings to clinical data and neurocognitive functions. METHODS: Twenty-nine patients [67.6+/-8.6 (52-85) years, 5 females] undergoing elective CABG with CPB were examined before surgery, at discharge and 3 months after surgery. A battery of standardized neuropsychological tests and questionnaires on depression and mood were administered. Conventional and diffusion-weighted MRI of the brain was performed and new lesions were analyzed. Clinical characteristics, neuropsychological test performance and radiographic data were collected and compared. RESULTS: There was no major neurological complication after CABG. Thirteen patients (45%) exhibited 32 new ischemic lesions on postoperative diffusion-weighted MRI. The lesions were small, rounded and equally dispersed in both hemispheres. Eight patients had at least two lesions. At discharge, significant deterioration of neuropsychological performance was observed in 6 of the 13 tests compared to baseline assessment. By 3 months postoperatively, 5 of the 6 tests returned to preoperative levels. Verbal learning ability, however, remained impaired. The presence of new focal brain lesions was not associated with impaired neuropsychological performance. There was also no correlation between clinical variables, intraoperative parameters and postoperative complications and MRI findings. CONCLUSIONS: Although neurocognitive decline after CABG is mostly transient, memory impairment can persist for months. New ischemic brain lesions on postoperative diffusion-weighted MRI do not appear to account for the persistent neurocognitive decline.