A fundamental problem in biomedical research is the low number of observations available, mostly due to a lack of available biosamples, prohibitive costs, or ethical reasons. Augmenting few real observations with generated in silico samples could lead to more robust analysis results and a higher reproducibility rate. Here, we propose the use of conditional single-cell generative adversarial neural networks (cscGAN) for the realistic generation of single-cell RNA-seq data. cscGAN learns non-linear gene-gene dependencies from complex, multiple cell type samples and uses this information to generate realistic cells of defined types. Augmenting sparse cell populations with cscGAN generated cells improves downstream analyses such as the detection of marker genes, the robustness and reliability of classifiers, the assessment of novel analysis algorithms, and might reduce the number of animal experiments and costs in consequence. cscGAN outperforms existing methods for single-cell RNA-seq data generation in quality and hold great promise for the realistic generation and augmentation of other biomedical data types.

We present Scaden, a deep neural network for cell deconvolution that uses gene expression information to infer the cellular composition of tissues. Scaden is trained on single-cell RNA sequencing (RNA-seq) data to engineer discriminative features that confer robustness to bias and noise, making complex data preprocessing and feature selection unnecessary. We demonstrate that Scaden outperforms existing deconvolution algorithms in both precision and robustness. A single trained network reliably deconvolves bulk RNA-seq and microarray, human and mouse tissue expression data and leverages the combined information of multiple datasets. Because of this stability and flexibility, we surmise that deep learning will become an algorithmic mainstay for cell deconvolution of various data types. Scaden’s software package and web application are easy to use on new as well as diverse existing expression datasets available in public resources, deepening the molecular and cellular understanding of developmental and disease processes.

Abstract-The major benefits of driving vehicles in controlled close formations such as platoons are that of increasing traffic fluidity and reducing air pollution. While V2V communications is requisite for platooning stability, the existing radio communications technologies (e.g., the IEEE 802.11p) suffer from poor performance in highly dense road scenarios, which are exactly to be created by platooning. This paper studies the applicability of visible light communications (VLC) system for information exchange between the platoon members. A complete VLC model is built enabling precise calculations of Bit-Error-Rate (BER) affected by inter-vehicle distance, background noise, incidence angle and receiver electrical bandwidth. Based on our analytical model, the optical parameters suiting platooning application are defined. Finally, a SIMULINK model is developed to study the performances of a platooning longitudinal and lateral control, where VLC is used for vehicle-to-vehicle information exchange. Our study demonstrates the feasibility of VLC-based platooning control even in the presence of optical noise at significant levels and up to certain degree of road curvature.

A fundamental problem in biomedical research is the low number of observations available, mostly due to a lack of available biosamples, prohibitive costs, or ethical reasons. Augmenting few real observations with generated in silico samples could lead to more robust analysis results and a higher reproducibility rate. Here we propose the use of conditional single cell Generative Adversarial Neural Networks (cscGANs) for the realistic generation of single cell RNA-seq data. cscGANs learn non-linear gene-gene dependencies from complex, multi cell type samples and use this information to generate realistic cells of defined types. Augmenting sparse cell populations with cscGAN generated cells improves downstream analyses such as the detection of marker genes, the robustness and reliability of classifiers, the assessment of novel analysis algorithms, and might reduce the number of animal experiments and costs in consequence. cscGANs outperform existing methods for single cell RNA-seq data generation in quality and hold

To cite this version:Mohamed Marouf, Yves Sorel. Scheduling non-preemptive hard real-time tasks with strict periods. ETFA '11 -16th

BackgroundThe Electrocardiogram ECG is one of the most important non-invasive tools for cardiac diseases diagnosis. Taking advantage of the developed telecommunication infrastructure, several approaches that address the development of telemetry cardiac devices were introduced recently. Telemetry ECG devices allow easy and fast ECG monitoring of patients with suspected cardiac issues. Choosing the right device with the desired working mode, signal quality, and the device cost are still the main obstacles to massive usage of these devices.MethodsIn this paper, we introduce design, implementation, and validation of a multi-purpose telemetry system for recording, transmission, and interpretation of ECG signals in different recording modes. The system consists of an ECG device, a cloud-based analysis pipeline, and accompanied mobile applications for physicians and patients. The proposed ECG device’s mechanical design allows laypersons to easily record post-event short-term ECG signals, using dry electrodes without any preparation. Moreover, patients can use the device to record long-term signals in loop and holter modes, using wet electrodes. In order to overcome the problem of signal quality fluctuation due to using different electrodes types and different placements on subject’s chest, customized ECG signal processing and interpretation pipeline is presented for each working mode.ResultsWe present the evaluation of the novel short-term recorder design. Recording of an ECG signal was performed for 391 patients using a standard 12-leads golden standard ECG and the proposed patient-activated short-term post-event recorder. In the validation phase, a sample of validation signals followed peer review process wherein two experts annotated the signals in terms of signal acceptability for diagnosis.We found that 96% of signals allow detecting arrhythmia and other signal’s abnormal changes. Additionally, we compared and presented the correlation coefficient and the automatic QRS delineation results of both short-term post-event recorder and 12-leads golden standard ECG recorder.ConclusionsThe proposed multi-purpose ECG device allows physicians to choose the working mode of the same device according to the patient status. The proposed device was designed to allow patients to manage the technical requirements of both working modes. Post-event short-term ECG recording using the proposed design provide physicians reliable three ECG leads with direct symptom-rhythm correlation.

No abstract

Computational methods for predicting ship speed profile in a complex ice field have traditionally relied on mechanistic simulations. However, such methods have difficulties capturing the entire complexity of shipice interaction process due to the incomplete understanding of the underlying physical phenomena. Therefore, data-driven approaches have recently gained increased attention in this context. Hence, this paper proposes a concept of a first machine learning-based simulator of ship speed profile in a complex ice field. The developed approach suggests using supervised machine learning to trace a function mapping several ship and ice parameters to the ship acceleration/deceleration between the two adjacent points along the route. The simulator is trained and tested on a dataset obtained from the full-scale tests of an icebreaking ship. The results show high accuracy of the developed method, with an average error of the simulated ship speed against the measured one ranging from 2.6% to 9.4%.