Ultra-Low Power Analog Folded Neural Network for Cardiovascular Health Monitoring.

Wearable sensors are increasingly used for continuous health monitoring, but their small size limits battery capacity, affecting user experience and monitoring capabilities. To overcome this, we introduce an ultra-low power analog Folded Neural Network (FNN) for physiological signal processing in a batteryless fashion. Our proposed FNN, by serializing computation, provides several benefits over traditional analog implementations, such as lower space, lower power consumption, and lower peak-to-average power ratio. We evaluate our method extensively using a dataset designed for ECG-based screening and diagnosis. Our analysis considers factors such as thermal noise, spatial requirements, and power consumption. Additionally, we evaluate detection performance, investigating various parameters of the proposed FNN. This evaluation provides insights into the optimal configuration for accurate anomaly detection. We observe a good trade-off for accuracy around 6 layers and a hidden size of 30 and further demonstrate that such architecture could be implemented in a wearable device and executed in a batteryless fashion.

Multimodal data analysis of epileptic EEG and rs-fMRI via deep learning and edge computing.

BACKGROUND AND OBJECTIVE: Multimodal data analysis and large-scale computational capability is entering medicine in an accelerative fashion and has begun to influence investigational work in a variety of disciplines. It is also informing us of therapeutic interventions that will come about with such development. Epilepsy is a chronic brain disorder in which functional changes may precede structural ones and which may be detectable using existing modalities. METHODS: Functional connectivity analysis using electroencephalography (EEG) and resting state-functional magnetic resonance imaging (rs-fMRI) has provided such meaningful input in cases of epilepsy. By leveraging the potential of autonomic edge computing in epilepsy, we develop and deploy both noninvasive and invasive methods for monitoring, evaluation, and regulation of the epileptic brain. First, an autonomic edge computing framework is proposed for the processing of big data as part of a decision support system for surgical candidacy. Second, a multimodal data analysis using independently acquired EEG and rs-fMRI is presented for estimation and prediction of the epileptogenic network. Third, an unsupervised feature extraction model is developed for EEG analysis and seizure prediction based on a Convolutional deep learning (CNN) structure for distinguishing preictal (pre-seizure) state from non-preictal periods by support vector machine (SVM) classifier. RESULTS: Experimental and simulation results from actual patient data validate the effectiveness of the proposed methods. CONCLUSIONS: The combination of rs-fMRI and EEG/iEEG can reveal more information about dynamic functional connectivity. However, simultaneous fMRI and EEG data acquisition present challenges. We have proposed system models for leveraging and processing independently acquired fMRI and EEG data.

Toward Wireless Health Monitoring via an Analog Signal Compression-Based Biosensing Platform.

Wireless all-analog biosensor design for the concurrent microfluidic and physiological signal monitoring is presented in this paper. The key component is an all-analog circuit capable of compressing two analog sources into one analog signal by the analog joint source-channel coding (AJSCC). Two circuit designs are discussed, including the stacked-voltage-controlled voltage source (VCVS) design with the fixed number of levels, and an improved design, which supports a flexible number of AJSCC levels. Experimental results are presented on the wireless biosensor prototype, composed of printed circuit board realizations of the stacked-VCVS design. Furthermore, circuit simulation and wireless link simulation results are presented on the improved design. Results indicate that the proposed wireless biosensor is well suited for sensing two biological signals simultaneously with high accuracy, and can be applied to a wide variety of low-power and low-cost wireless continuous health monitoring applications.

Random ensemble learning for EEG classification.

Real-time detection of seizure activity in epilepsy patients is critical in averting seizure activity and improving patients' quality of life. Accurate evaluation, presurgical assessment, seizure prevention, and emergency alerts all depend on the rapid detection of seizure onset. A new method of feature selection and classification for rapid and precise seizure detection is discussed wherein informative components of electroencephalogram (EEG)-derived data are extracted and an automatic method is presented using infinite independent component analysis (I-ICA) to select independent features. The feature space is divided into subspaces via random selection and multichannel support vector machines (SVMs) are used to classify these subspaces. The result of each classifier is then combined by majority voting to establish the final output. In addition, a random subspace ensemble using a combination of SVM, multilayer perceptron (MLP) neural network and an extended k-nearest neighbors (k-NN), called extended nearest neighbor (ENN), is developed for the EEG and electrocorticography (ECoG) big data problem. To evaluate the solution, a benchmark ECoG of eight patients with temporal and extratemporal epilepsy was implemented in a distributed computing framework as a multitier cloud-computing architecture. Using leave-one-out cross-validation, the accuracy, sensitivity, specificity, and both false positive and false negative ratios of the proposed method were found to be 0.97, 0.98, 0.96, 0.04, and 0.02, respectively. Application of the solution to cases under investigation with ECoG has also been effected to demonstrate its utility.

Comparative performance evaluation of automated segmentation methods of hippocampus from magnetic resonance images of temporal lobe epilepsy patients.

PURPOSE: Segmentation of the hippocampus from magnetic resonance (MR) images is a key task in the evaluation of mesial temporal lobe epilepsy (mTLE) patients. Several automated algorithms have been proposed although manual segmentation remains the benchmark. Choosing a reliable algorithm is problematic since structural definition pertaining to multiple edges, missing and fuzzy boundaries, and shape changes varies among mTLE subjects. Lack of statistical references and guidance for quantifying the reliability and reproducibility of automated techniques has further detracted from automated approaches. The purpose of this study was to develop a systematic and statistical approach using a large dataset for the evaluation of automated methods and establish a method that would achieve results better approximating those attained by manual tracing in the epileptogenic hippocampus. METHODS: A template database of 195 (81 males, 114 females; age range 32-67 yr, mean 49.16 yr) MR images of mTLE patients was used in this study. Hippocampal segmentation was accomplished manually and by two well-known tools (FreeSurfer and hammer) and two previously published methods developed at their institution [Automatic brain structure segmentation (ABSS) and LocalInfo]. To establish which method was better performing for mTLE cases, several voxel-based, distance-based, and volume-based performance metrics were considered. Statistical validations of the results using automated techniques were compared with the results of benchmark manual segmentation. Extracted metrics were analyzed to find the method that provided a more similar result relative to the benchmark. RESULTS: Among the four automated methods, ABSS generated the most accurate results. For this method, the Dice coefficient was 5.13%, 14.10%, and 16.67% higher, Hausdorff was 22.65%, 86.73%, and 69.58% lower, precision was 4.94%, -4.94%, and 12.35% higher, and the root mean square (RMS) was 19.05%, 61.90%, and 65.08% lower than LocalInfo, FreeSurfer, and hammer, respectively. The Bland-Altman similarity analysis revealed a low bias for the ABSS and LocalInfo techniques compared to the others. CONCLUSIONS: The ABSS method for automated hippocampal segmentation outperformed other methods, best approximating what could be achieved by manual tracing. This study also shows that four categories of input data can cause automated segmentation methods to fail. They include incomplete studies, artifact, low signal-to-noise ratio, and inhomogeneity. Different scanner platforms and pulse sequences were considered as means by which to improve reliability of the automated methods. Other modifications were specially devised to enhance a particular method assessed in this study.

A Distributed Computing Framework for Real-Time Detection of Stress and of Its Propagation in a Team.

Stress is one of the key factor that impacts the quality of our daily life: From the productivity and efficiency in the production processes to the ability of (civilian and military) individuals in making rational decisions. Also, stress can propagate from one individual to other working in a close proximity or toward a common goal, e.g., in a military operation or workforce. Real-time assessment of the stress of individuals alone is, however, not sufficient, as understanding its source and direction in which it propagates in a group of people is equally-if not more-important. A continuous near real-time in situ personal stress monitoring system to quantify level of stress of individuals and its direction of propagation in a team is envisioned. However, stress monitoring of an individual via his/her mobile device may not always be possible for extended periods of time due to limited battery capacity of these devices. To overcome this challenge a novel distributed mobile computing framework is proposed to organize the resources in the vicinity and form a mobile device cloud that enables offloading of computation tasks in stress detection algorithm from resource constrained devices (low residual battery, limited CPU cycles) to resource rich devices. Our framework also supports computing parallelization and workflows, defining how the data and tasks divided/assigned among the entities of the framework are designed. The direction of propagation and magnitude of influence of stress in a group of individuals are studied by applying real-time, in situ analysis of Granger Causality. Tangible benefits (in terms of energy expenditure and execution time) of the proposed framework in comparison to a centralized framework are presented via thorough simulations and real experiments.

Laboratory Validation of Inertial Body Sensors to Detect Cigarette Smoking Arm Movements.

Cigarette smoking remains the leading cause of preventable death in the United States. Traditional in-clinic cessation interventions may fail to intervene and interrupt the rapid progression to relapse that typically occurs following a quit attempt. The ability to detect actual smoking behavior in real-time is a measurement challenge for health behavior research and intervention. The successful detection of real-time smoking through mobile health (mHealth) methodology has substantial implications for developing highly efficacious treatment interventions. The current study was aimed at further developing and testing the ability of inertial sensors to detect cigarette smoking arm movements among smokers. The current study involved four smokers who smoked six cigarettes each in a laboratory-based assessment. Participants were outfitted with four inertial body movement sensors on the arms, which were used to detect smoking events at two levels: the puff level and the cigarette level. Two different algorithms (Support Vector Machines (SVM) and Edge-Detection based learning) were trained to detect the features of arm movement sequences transmitted by the sensors that corresponded with each level. The results showed that performance of the SVM algorithm at the cigarette level exceeded detection at the individual puff level, with low rates of false positive puff detection. The current study is the second in a line of programmatic research demonstrating the proof-of-concept for sensor-based tracking of smoking, based on movements of the arm and wrist. This study demonstrates efficacy in a real-world clinical inpatient setting and is the first to provide a detection rate against direct observation, enabling calculation of true and false positive rates. The study results indicate that the approach performs very well with some participants, whereas some challenges remain with participants who generate more frequent non-smoking movements near the face. Future work may allow for tracking smoking in real-world environments, which would facilitate developing more effective, just-in-time smoking cessation interventions.

Statistical validation of automatic methods for hippocampus segmentation in MR images of epileptic patients.

Hippocampus segmentation is a key step in the evaluation of mesial Temporal Lobe Epilepsy (mTLE) by MR images. Several automated segmentation methods have been introduced for medical image segmentation. Because of multiple edges, missing boundaries, and shape changing along its longitudinal axis, manual outlining still remains the benchmark for hippocampus segmentation, which however, is impractical for large datasets due to time constraints. In this study, four automatic methods, namely FreeSurfer, Hammer, Automatic Brain Structure Segmentation (ABSS), and LocalInfo segmentation, are evaluated to find the most accurate and applicable method that resembles the bench-mark of hippocampus. Results from these four methods are compared against those obtained using manual segmentation for T1-weighted images of 157 symptomatic mTLE patients. For performance evaluation of automatic segmentation, Dice coefficient, Hausdorff distance, Precision, and Root Mean Square (RMS) distance are extracted and compared. Among these four automated methods, ABSS generates the most accurate results and the reproducibility is more similar to expert manual outlining by statistical validation. By considering p-value<;0.05, the results of performance measurement for ABSS reveal that, Dice is 4%, 13%, and 17% higher, Hausdorff is 23%, 87%, and 70% lower, precision is 5%, -5%, and 12% higher, and RMS is 19%, 62%, and 65% lower compared to LocalInfo, FreeSurfer, and Hammer, respectively.