@article{MalschLimanWiedmannetal.2018,
  author    = {Malsch, Carolin and Liman, Thomas and Wiedmann, Silke and Siegerink, Bob and Georgakis, Marios K. and Tiedt, Steffen and Endres, Matthias and Heuschmann, Peter U.},
  title     = {Outcome after stroke attributable to baseline factors—the PROSpective Cohort with Incident Stroke (PROSCIS)},
  series = {PLoS ONE},
  volume    = {13},
  journal   = {PLoS ONE},
  number    = {9},
  doi       = {10.1371/journal.pone.0204285},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-177342},
  pages     = {e0204285},
  year      = {2018},
  abstract  = {Background The impact of risk factors on poor outcome after ischemic stroke is well known, but estimating the amount of poor outcome attributable to single factors is challenging in presence of multimorbidity. We aim to compare population attributable risk estimates obtained from different statistical approaches regarding their consistency. We use a real-life data set from the PROSCIS study to identify predictors for mortality and functional impairment one year after first-ever ischemic stroke and quantify their contribution to poor outcome using population attributable risks. Methods The PROSpective Cohort with Incident Stroke (PROSCIS) is a prospective observational hospital-based cohort study of patients after first-ever stroke conducted independently in Berlin (PROSCIS-B) and Munich (PROSCIS-M). The association of baseline factors with poor outcome one year after stroke in PROSCIS-B was analysed using multiple logistic regression analysis and population attributable risks were calculated, which were estimated using sequential population attributable risk based on a multiple generalized additive regression model, doubly robust estimation, as well as using average sequential population attributable risk. Findings were reproduced in an independent validation sample from PROSCIS-M. Results Out of 507 patients with available outcome information after 12 months in PROSCIS-B, 20.5\% suffered from poor outcome. Factors associated with poor outcome were age, pre-stroke physical disability, stroke severity (NIHSS), education, and diabetes mellitus. The order of risk factors ranked by magnitudes of population attributable risk was almost similar for all methods, but population attributable risk estimates varied markedly between the methods. In PROSCIS-M, incidence of poor outcome and distribution of baseline parameters were comparable. The multiple logistic regression model could be reproduced for all predictors, except pre-stroke physical disability. Similar to PROSCIS-B, the order of risk factors ranked by magnitudes of population attributable risk was almost similar for all methods, but magnitudes of population attributable risk differed markedly between the methods. Conclusions Ranking of risk factors by population impact is not affected by the different statistical approaches. Thus, for a rational decision on which risk factor to target in disease interventions, population attributable risk is a supportive tool. However, population attributable risk estimates are difficult to interpret and are not comparable when they origin from studies applying different methodology. The predictors for poor outcome identified in PROSCIS-B have a relevant impact on mortality and functional impairment one year after first-ever ischemic stroke.},
  language  = {en}
}
@article{PfitznerMayNuechter2018,
  author    = {Pfitzner, Christian and May, Stefan and N{\"u}chter, Andreas},
  title     = {Body weight estimation for dose-finding and health monitoring of lying, standing and walking patients based on RGB-D data},
  series = {Sensors},
  volume    = {18},
  journal   = {Sensors},
  number    = {5},
  doi       = {10.3390/s18051311},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-176642},
  pages     = {1311},
  year      = {2018},
  abstract  = {This paper describes the estimation of the body weight of a person in front of an RGB-D camera. A survey of different methods for body weight estimation based on depth sensors is given. First, an estimation of people standing in front of a camera is presented. Second, an approach based on a stream of depth images is used to obtain the body weight of a person walking towards a sensor. The algorithm first extracts features from a point cloud and forwards them to an artificial neural network (ANN) to obtain an estimation of body weight. Besides the algorithm for the estimation, this paper further presents an open-access dataset based on measurements from a trauma room in a hospital as well as data from visitors of a public event. In total, the dataset contains 439 measurements. The article illustrates the efficiency of the approach with experiments with persons lying down in a hospital, standing persons, and walking persons. Applicable scenarios for the presented algorithm are body weight-related dosing of emergency patients.},
  language  = {en}
}
@article{KazuhinoWernerToriumietal.2018,
  author    = {Kazuhino, Koshino and Werner, Rudolf A. and Toriumi, Fuijo and Javadi, Mehrbod S. and Pomper, Martin G. and Solnes, Lilja B. and Verde, Franco and Higuchi, Takahiro and Rowe, Steven P.},
  title     = {Generative Adversarial Networks for the Creation of Realistic Artificial Brain Magnetic Resonance Images},
  series = {Tomography},
  volume    = {4},
  journal   = {Tomography},
  number    = {4},
  doi       = {10.18383/j.tom.2018.00042},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-172185},
  pages     = {159-163},
  year      = {2018},
  abstract  = {Even as medical data sets become more publicly accessible, most are restricted to specific medical conditions. Thus, data collection for machine learning approaches remains challenging, and synthetic data augmentation, such as generative adversarial networks (GAN), may overcome this hurdle. In the present quality control study, deep convolutional GAN (DCGAN)-based human brain magnetic resonance (MR) images were validated by blinded radiologists. In total, 96 T1-weighted brain images from 30 healthy individuals and 33 patients with cerebrovascular accident were included. A training data set was generated from the T1-weighted images and DCGAN was applied to generate additional artificial brain images. The likelihood that images were DCGAN-created versus acquired was evaluated by 5 radiologists (2 neuroradiologists [NRs], vs 3 non-neuroradiologists [NNRs]) in a binary fashion to identify real vs created images. Images were selected randomly from the data set (variation of created images, 40\%-60\%). None of the investigated images was rated as unknown. Of the created images, the NRs rated 45\% and 71\% as real magnetic resonance imaging images (NNRs, 24\%, 40\%, and 44\%). In contradistinction, 44\% and 70\% of the real images were rated as generated images by NRs (NNRs, 10\%, 17\%, and 27\%). The accuracy for the NRs was 0.55 and 0.30 (NNRs, 0.83, 0.72, and 0.64). DCGAN-created brain MR images are similar enough to acquired MR images so as to be indistinguishable in some cases. Such an artificial intelligence algorithm may contribute to synthetic data augmentation for "data-hungry" technologies, such as supervised machine learning approaches, in various clinical applications.},
  subject      = {Magnetresonanztomografie},
  language  = {en}
}