Publikationen

Most magneto- and electroencephalography (M/EEG) based source estimation techniques derive their estimates sample wise, independently across time. However, neuronal assemblies are intricately interconnected, constraining the temporal evolution of neural activity that is detected by MEG and EEG; the observed neural currents must thus be highly context dependent. Here, we use a network of Long Short-Term Memory (LSTM) cells where the input is a sequence of past source estimates and the output is a prediction of the following estimate. This prediction is then used to correct the estimate. In this study, we applied this technique on noise-normalized minimum norm estimates (MNE). Because the correction is found by using past activity (context), we call this implementation Contextual MNE (CMNE), although this technique can be used in conjunction with any source estimation method. We test CMNE on simulated epileptiform activity and recorded auditory steady state response (ASSR) data, showing that the CMNE estimates exhibit a higher degree of spatial fidelity than the unfiltered estimates in the tested cases.

In this article, a just-in-time-learning (JITL)-aided canonical correlation analysis (CCA) is proposed for the monitoring and fault detection of multimode processes. A canonical correlation analysis (CCA)-based fault detection method has been applied to single-operating-mode processes. However, CCA has limitations in handling processes with multiple operating points. These limitations are illustrated by a numerical example. To reduce the time for searching relevant data, K-means is integrated into the JITL to build the local CCA model. Furthermore, the proposed method is compared with commonly used kernel-based methods in terms of computational complexity and interpretability of the results. Finally, the validity and efficacy of the proposed method are shown using an industrial benchmark process. Results show that the proposed method has better performance than conventional methods in terms of fault detection rate while still tracking changes in the system.

Background - Transcranial direct current stimulation (tDCS), a neuromodulatory non-invasive brain stimulation technique, has shown promising results in basic and clinical studies. The known interindividual variability of the effects, however, limits the efficacy of the technique. Recently we reported neurophysiological effects of tDCS applied over the primary motor cortex at the group level, based on data from twenty-nine participants who received 15min of either sham, 0.5, 1.0, 1.5 or 2.0 mA anodal, or cathodal tDCS. The neurophysiological effects were evaluated via changes in: 1) transcranial magnetic stimulation (TMS)-induced motor evoked potentials (MEP), and 2) cerebral blood flow (CBF) measured by functional magnetic resonance imaging (MRI) via arterial spin labeling (ASL). At the group level, dose-dependent effects of the intervention were obtained, which however displayed interindividual variability. - Method - In the present study, we investigated the cause of the observed inter-individual variability. To this end, for each participant, a MRI-based realistic head model was designed to 1) calculate anatomical factors and 2) simulate the tDCS- and TMS-induced electrical fields (EFs). We first investigated at the regional level which individual anatomical factors explained the simulated EFs (magnitude and normal component). Then, we explored which specific anatomical and/or EF factors predicted the neurophysiological outcomes of tDCS. - Results - The results highlight a significant negative correlation between regional electrode-to-cortex distance (rECD) as well as regional CSF (rCSF) thickness, and the individual EF characteristics. In addition, while both rCSF thickness and rECD anticorrelated with tDCS-induced physiological changes, EFs positively correlated with the effects. - Conclusion - These results provide novel insights into the dependency of the neuromodulatory effects of tDCS on individual physical factors.

Over the past thirty years, database management systems have been established as one of the most successful software concepts. In todays business environment they constitute the centerpiece of almost all critical IT systems. The reasons for this success are manyfold. On the one hand, such systems provide abstractions hiding the details of underlying hardware or operating systems layers. On the other hand, database management systems are ACID compliant, which enables them to represent an accurate picture of a real world scenario, and ensures correctness of the managed data.However, the currently used database concepts and systems are not well prepared to support emerging application domains such as eSciences, Industry 4.0, Internet of Things or Digital Humanities. Furthermore, volume, variety, veracity as well as velocity of data caused by ubiquitous sensors have to be mastered by massive scalability and online processing by providing traditional qualities of database systems like consistency, isolation and descriptive query languages. At the same time, current and future hardware trends provide new opportunities such as many-core CPUs, co-processors like GPU and FPGA, novel storage technologies like NVRAM and SSD as well as high-speed networks provide new opportunities.In this talk we present our research results for the use of modern hardware architectures for data management. We discuss the design of data structures for persistent memory and the use of accelerators like GPU and FPGA for database operations.