| 1. |
- Andersson, Jennifer, et al.
(författare)
-
Anomaly Detection for the Centralised Elasticsearch Service at CERN
- 2021
-
Ingår i: Frontiers in Big Data. - : Frontiers Media S.A.. - 2624-909X. ; 4
-
Tidskriftsartikel (refereegranskat)abstract
- For several years CERN has been offering a centralised service for Elasticsearch, a popular distributed system for search and analytics of user provided data. The service offered by CERN IT is better described as a service of services, delivering centrally managed and maintained Elasticsearch instances to CERN users who have a justified need for it. This dynamic infrastructure currently consists of about 30 distinct and independent Elasticsearch installations, in the following referred to as Elasticsearch clusters, some of which are shared between different user communities. The service is used by several hundred users mainly for logs and service analytics. Due to its size and complexity, the installation produces a huge amount of internal monitoring data which can be difficult to process in real time with limited available person power. Early on, an idea was therefore born to process this data automatically, aiming to extract anomalies and possible issues building up in real time, allowing the experts to address them before they start to cause an issue for the users of the service. Both deep learning and traditional methods have been applied to analyse the data in order to achieve this goal. This resulted in the current deployment of an anomaly detection system based on a one layer multi dimensional LSTM neural network, coupled with applying a simple moving average to the data to validate the results. This paper will describe which methods were investigated and give an overview of the current system, including data retrieval, data pre-processing and analysis. In addition, reports on experiences gained when applying the system to actual data will be provided. Finally, weaknesses of the current system will be briefly discussed, and ideas for future system improvements will be sketched out.
|
|
| 2. |
- Archetti, D, et al.
(författare)
-
Inter-Cohort Validation of SuStaIn Model for Alzheimer's Disease
- 2021
-
Ingår i: Frontiers in big data. - : Frontiers Media SA. - 2624-909X. ; 4, s. 661110-
-
Tidskriftsartikel (refereegranskat)abstract
- Alzheimer’s disease (AD) is a neurodegenerative disorder which spans several years from preclinical manifestations to dementia. In recent years, interest in the application of machine learning (ML) algorithms to personalized medicine has grown considerably, and a major challenge that such models face is the transferability from the research settings to clinical practice. The objective of this work was to demonstrate the transferability of the Subtype and Stage Inference (SuStaIn) model from well-characterized research data set, employed as training set, to independent less-structured and heterogeneous test sets representative of the clinical setting. The training set was composed of MRI data of 1043 subjects from the Alzheimer’s disease Neuroimaging Initiative (ADNI), and the test set was composed of data from 767 subjects from OASIS, Pharma-Cog, and ViTA clinical datasets. Both sets included subjects covering the entire spectrum of AD, and for both sets volumes of relevant brain regions were derived from T1-3D MRI scans processed with Freesurfer v5.3 cross-sectional stream. In order to assess the predictive value of the model, subpopulations of subjects with stable mild cognitive impairment (MCI) and MCIs that progressed to AD dementia (pMCI) were identified in both sets. SuStaIn identified three disease subtypes, of which the most prevalent corresponded to the typical atrophy pattern of AD. The other SuStaIn subtypes exhibited similarities with the previously defined hippocampal sparing and limbic predominant atrophy patterns of AD. Subject subtyping proved to be consistent in time for all cohorts and the staging provided by the model was correlated with cognitive performance. Classification of subjects on the basis of a combination of SuStaIn subtype and stage, mini mental state examination and amyloid-β1-42 cerebrospinal fluid concentration was proven to predict conversion from MCI to AD dementia on par with other novel statistical algorithms, with ROC curves that were not statistically different for the training and test sets and with area under curve respectively equal to 0.77 and 0.76. This study proves the transferability of a SuStaIn model for AD from research data to less-structured clinical cohorts, and indicates transferability to the clinical setting.
|
|
| 3. |
- Froese, L, et al.
(författare)
-
Computer Vision for Continuous Bedside Pharmacological Data Extraction: A Novel Application of Artificial Intelligence for Clinical Data Recording and Biomedical Research
- 2021
-
Ingår i: Frontiers in big data. - : Frontiers Media SA. - 2624-909X. ; 4, s. 689358-
-
Tidskriftsartikel (refereegranskat)abstract
- Introduction: As real time data processing is integrated with medical care for traumatic brain injury (TBI) patients, there is a requirement for devices to have digital output. However, there are still many devices that fail to have the required hardware to export real time data into an acceptable digital format or in a continuously updating manner. This is particularly the case for many intravenous pumps and older technological systems. Such accurate and digital real time data integration within TBI care and other fields is critical as we move towards digitizing healthcare information and integrating clinical data streams to improve bedside care. We propose to address this gap in technology by building a system that employs Optical Character Recognition through computer vision, using real time images from a pump monitor to extract the desired real time information.Methods: Using freely available software and readily available technology, we built a script that extracts real time images from a medication pump and then processes them using Optical Character Recognition to create digital text from the image. This text was then transferred to an ICM + real-time monitoring software in parallel with other retrieved physiological data.Results: The prototype that was built works effectively for our device, with source code openly available to interested end-users. However, future work is required for a more universal application of such a system.Conclusion: Advances here can improve medical information collection in the clinical environment, eliminating human error with bedside charting, and aid in data integration for biomedical research where many complex data sets can be seamlessly integrated digitally. Our design demonstrates a simple adaptation of current technology to help with this integration.
|
|
| 4. |
- Markidis, Stefano
(författare)
-
The Old and the New : Can Physics-Informed Deep-Learning Replace Traditional Linear Solvers?
- 2021
-
Ingår i: Frontiers In Big Data. - : Frontiers Media SA. - 2624-909X. ; 4
-
Tidskriftsartikel (refereegranskat)abstract
- Physics-Informed Neural Networks (PINN) are neural networks encoding the problem governing equations, such as Partial Differential Equations (PDE), as a part of the neural network. PINNs have emerged as a new essential tool to solve various challenging problems, including computing linear systems arising from PDEs, a task for which several traditional methods exist. In this work, we focus first on evaluating the potential of PINNs as linear solvers in the case of the Poisson equation, an omnipresent equation in scientific computing. We characterize PINN linear solvers in terms of accuracy and performance under different network configurations (depth, activation functions, input data set distribution). We highlight the critical role of transfer learning. Our results show that low-frequency components of the solution converge quickly as an effect of the F-principle. In contrast, an accurate solution of the high frequencies requires an exceedingly long time. To address this limitation, we propose integrating PINNs into traditional linear solvers. We show that this integration leads to the development of new solvers whose performance is on par with other high-performance solvers, such as PETSc conjugate gradient linear solvers, in terms of performance and accuracy. Overall, while the accuracy and computational performance are still a limiting factor for the direct use of PINN linear solvers, hybrid strategies combining old traditional linear solver approaches with new emerging deep-learning techniques are among the most promising methods for developing a new class of linear solvers.
|
|
