A compact and interpretable convolutional neural network for cross-subject driver drowsiness detection from single-channel EEG.

Driver drowsiness is one of the main factors leading to road fatalities and hazards in the transportation industry. Electroencephalography (EEG) has been considered as one of the best physiological signals to detect drivers' drowsy states, since it directly measures neurophysiological activities in the brain. However, designing a calibration-free system for driver drowsiness detection with EEG is still a challenging task, as EEG suffers from serious mental and physical drifts across different subjects. In this paper, we propose a compact and interpretable Convolutional Neural Network (CNN) to discover shared EEG features across different subjects for driver drowsiness detection. We incorporate the Global Average Pooling (GAP) layer in the model structure, allowing the Class Activation Map (CAM) method to be used for localizing regions of the input signal that contribute most for classification. Results show that the proposed model can achieve an average accuracy of 73.22% on 11 subjects for 2-class cross-subject EEG signal classification, which is higher than conventional machine learning methods and other state-of-art deep learning methods. It is revealed by the visualization technique that the model has learned biologically explainable features, e.g., Alpha spindles and Theta burst, as evidence for the drowsy state. It is also interesting to see that the model uses artifacts that usually dominate the wakeful EEG, e.g., muscle artifacts and sensor drifts, to recognize the alert state. The proposed model illustrates a potential direction to use CNN models as a powerful tool to discover shared features related to different mental states across different subjects from EEG signals.

EEG-Based Cross-Subject Driver Drowsiness Recognition With an Interpretable Convolutional Neural Network.

In the context of electroencephalogram (EEG)-based driver drowsiness recognition, it is still challenging to design a calibration-free system, since EEG signals vary significantly among different subjects and recording sessions. Many efforts have been made to use deep learning methods for mental state recognition from EEG signals. However, existing work mostly treats deep learning models as black-box classifiers, while what have been learned by the models and to which extent they are affected by the noise in EEG data are still underexplored. In this article, we develop a novel convolutional neural network combined with an interpretation technique that allows sample-wise analysis of important features for classification. The network has a compact structure and takes advantage of separable convolutions to process the EEG signals in a spatial-temporal sequence. Results show that the model achieves an average accuracy of 78.35% on 11 subjects for leave-one-out cross-subject drowsiness recognition, which is higher than the conventional baseline methods of 53.40%-72.68% and state-of-the-art deep learning methods of 71.75%-75.19%. Interpretation results indicate the model has learned to recognize biologically meaningful features from EEG signals, e.g., alpha spindles, as strong indicators of drowsiness across different subjects. In addition, we also explore reasons behind some wrongly classified samples with the interpretation technique and discuss potential ways to improve the recognition accuracy. Our work illustrates a promising direction on using interpretable deep learning models to discover meaningful patterns related to different mental states from complex EEG signals.

Intra-operative support via augmented reality techniques in endoscopic surgery.

Computer assisted operation planning systems are gaining increasing recognition in the field of surgery. These systems offer new possibilities for preparing an intervention, with the goal of reducing the amount of expensive operating-room time required for the intervention. The safest and most effective surgical approach should always be selected, but it is often difficult to transfer the output of the planning system to the intra-operative situation so that the planning results can be considered during the actual intervention. At the Fraunhofer Institute for Computer Graphics (IGD) in Darmstadt and the Centre for Advanced Media Technology (CAMTech) in Singapore methods are being developed to bridge the gap between the external planning session and the intra-operative case: Augmented Reality (AR) techniques are used to overlay preoperative scanned image data, as well as results of the planning session, on the operation field.

Nasal airflow diagnosis--comparison of experimental studies and computer simulations.

The lack of suited diagnostic tools providing insight into patient specific flow characteristics of the nasal airflow is one of the main problems in functional diagnosis. Diagnostic methods currently used do not provide the necessary information for flow analysis. But the flow distribution is essential for a physiological respiration, in particular for cleaning, moistening and tempering of the inhaled air as well as for the olfactory function of the nose. To overcome this current situation a cooperation project of the ENT surgeons and computer graphic engineers was established to develop the computer assisted planning system STAN (Simulation Tool for Airflow in the human Nose) combining Computer Fluid Dynamics (CFD) with advanced Computer Graphic Technology. The idea of the STAN system is to perform patient specific airflow simulations in the patient's nasal cavities. Therefore a geometrical model of the nasal airways is derived from the patient's tomography scans. A discretization of the surrounded flow volume is made by a computational grid. To establish the flow simulation Finite Element Methods are performed on the grid. A tailored visualization is offered to the surgeon that overlaps the flow pattern to the patient's tomography data shown in the coronal, sagittal and transversal plane. The surgeon can not only analyze the patient's current respiratory situation he has also the possibility to describe the planned surgical intervention. The goal is to simulate the flow distribution that can be expected after the surgical intervention and to offer a possibility to validate various surgical strategies. To verify the simulation results experimental investigations and measurements are made in nasal models. Silicon Models of patient's nose channels are made to analyze flow characteristics. The CT or MR scans of the same patients are used as input data for the simulation. The experimental outcome is compared to the simulation results to validate this diagnostic approach.

Parallel mutual information estimation for inferring gene regulatory networks on GPUs.

BACKGROUND: Mutual information is a measure of similarity between two variables. It has been widely used in various application domains including computational biology, machine learning, statistics, image processing, and financial computing. Previously used simple histogram based mutual information estimators lack the precision in quality compared to kernel based methods. The recently introduced B-spline function based mutual information estimation method is competitive to the kernel based methods in terms of quality but at a lower computational complexity. RESULTS: We present a new approach to accelerate the B-spline function based mutual information estimation algorithm with commodity graphics hardware. To derive an efficient mapping onto this type of architecture, we have used the Compute Unified Device Architecture (CUDA) programming model to design and implement a new parallel algorithm. Our implementation, called CUDA-MI, can achieve speedups of up to 82 using double precision on a single GPU compared to a multi-threaded implementation on a quad-core CPU for large microarray datasets. We have used the results obtained by CUDA-MI to infer gene regulatory networks (GRNs) from microarray data. The comparisons to existing methods including ARACNE and TINGe show that CUDA-MI produces GRNs of higher quality in less time. CONCLUSIONS: CUDA-MI is publicly available open-source software, written in CUDA and C++ programming languages. It obtains significant speedup over sequential multi-threaded implementation by fully exploiting the compute capability of commonly used CUDA-enabled low-cost GPUs.

CUDA-BLASTP: accelerating BLASTP on CUDA-enabled graphics hardware.

Scanning protein sequence database is an often repeated task in computational biology and bioinformatics. However, scanning large protein databases, such as GenBank, with popular tools such as BLASTP requires long runtimes on sequential architectures. Due to the continuing rapid growth of sequence databases, there is a high demand to accelerate this task. In this paper, we demonstrate how GPUs, powered by the Compute Unified Device Architecture (CUDA), can be used as an efficient computational platform to accelerate the BLASTP algorithm. In order to exploit the GPU's capabilities for accelerating BLASTP, we have used a compressed deterministic finite state automaton for hit detection as well as a hybrid parallelization scheme. Our implementation achieves speedups up to 10.0 on an NVIDIA GeForce GTX 295 GPU compared to the sequential NCBI BLASTP 2.2.22. CUDA-BLASTP source code which is available at https://sites.google.com/site/liuweiguohome/software.

A parallel algorithm for error correction in high-throughput short-read data on CUDA-enabled graphics hardware.

Emerging DNA sequencing technologies open up exciting new opportunities for genome sequencing by generating read data with a massive throughput. However, produced reads are significantly shorter and more error-prone compared to the traditional Sanger shotgun sequencing method. This poses challenges for de novo DNA fragment assembly algorithms in terms of both accuracy (to deal with short, error-prone reads) and scalability (to deal with very large input data sets). In this article, we present a scalable parallel algorithm for correcting sequencing errors in high-throughput short-read data so that error-free reads can be available before DNA fragment assembly, which is of high importance to many graph-based short-read assembly tools. The algorithm is based on spectral alignment and uses the Compute Unified Device Architecture (CUDA) programming model. To gain efficiency we are taking advantage of the CUDA texture memory using a space-efficient Bloom filter data structure for spectrum membership queries. We have tested the runtime and accuracy of our algorithm using real and simulated Illumina data for different read lengths, error rates, input sizes, and algorithmic parameters. Using a CUDA-enabled mass-produced GPU (available for less than US$400 at any local computer outlet), this results in speedups of 12-84 times for the parallelized error correction, and speedups of 3-63 times for both sequential preprocessing and parallelized error correction compared to the publicly available Euler-SR program. Our implementation is freely available for download from http://cuda-ec.sourceforge.net .