Javad Birjandtalab

Multi-Biosignal Analysis for Epileptic Seizure Monitoring

Persons who suffer from intractable seizures are safer if attended when seizures strike. Consequently, there is a need for wearable devices capable of detecting both convulsive and nonconvulsive seizures in everyday life. We have developed a three-stage seizure detection methodology based on 339 h of data (26 seizures) collected from 10 patients in an epilepsy monitoring unit. Our intent is to develop a wearable system that will detect seizures, alert a caregiver and record the time of seizure in an electronic diary for the patients physician. Stage I looks for concurrent activity in heart rate, arterial oxygenation and electrodermal activity, all of which can be monitored by a wrist-worn device and which in combination produce a very low false positive rate. Stage II looks for a specific pattern created by these three biosignals. For the patients whose seizures cannot be detected by Stage II, Stage III detects seizures using limited-channel electroencephalogram (EEG) monitoring with at most three electrodes. Out of 10 patients, Stage I recognized all 11 seizures from seven patients, Stage II detected all 10 seizures from six patients and Stage III detected all of the seizures of two out of the three patients it analyzed.

Automated seizure detection using limited-channel EEG and non-linear dimension reduction

Electroencephalography (EEG) is an essential component in evaluation of epilepsy. However, full-channel EEG signals recorded from 18 to 23 electrodes on the scalp is neither wearable nor computationally effective. This paper presents advantages of both channel selection and nonlinear dimension reduction for accurate automatic seizure detection. We first extract the frequency domain features from the full-channel EEG signals. Then, we use a random forest algorithm to determine which channels contribute the most in discriminating seizure from non-seizure events. Next, we apply a non-linear dimension reduction technique to capture the relationship among data elements and map them in low dimension. Finally, we apply a KNN classifier technique to discriminate between seizure and non-seizure events. The experimental results for 23 patients show that our proposed approach outperforms other techniques in terms of accuracy. It also visualizes long-term data in 2D to enhance physician cognition of occurrence and disease progression.

Nonlinear dimension reduction for EEG-based epileptic seizure detection

Approximately 0.1 percent of epileptic patients die from unexpected deaths. In general, for intractable seizures, it is crucial to have an algorithm to accurately and automatically detect the seizures and notify care-givers to assist patients. EEG signals are known as definitive diagnosis of seizure events. In this work, we utilize the frequency domain features (normalized in-band power spectral density) for the EEG channels. We applied a nonlinear data-embedding technique based on stochastic neighbor distance metric to capture the relationships among data elements in high dimension and improve the accuracy of seizure detection. This proposed data embedding technique not only makes it possible to visualize data in two or three dimensions, but also tackles the inherent difficulties regarding high dimensional data classification such as time complexity and memory requirement. We also applied a patient specific KNN classification to detect seizure and non-seizure events. The results indicate that our nonlinear technique provides significantly better visualization and classification efficiency (F-measure greater than 87%) compared to conventional dimension reduction approaches.

A two-stage clustering technique for automatic biaxial gating of flow cytometry data

Measurement of various markers of single cells using flow cytometry has several biological applications. These applications include improving our understanding of behavior of cellular systems, identifying rare cell populations and personalized medication. A common critical issue in the existing methods is approximation of the number of cellular populations which heavily affects the accuracy of results. In this work, we propose a novel technique to estimate the number of dominant subtypes and identify them in flow cytometry datasets. Our experimentation on 42 flow cytometry datasets indicates high performance and accurate clustering (F-measure > 91%) in identifying the main cellular populations.

A case study on minimum energy operation for dynamic time warping signal processing in wearable computers

Miniaturization and form factor reduction in wearable computers leads to enhanced wearability. Power optimization typically translates to form factor reduction, hence of paramount importance. This paper demonstrates power consumption analysis obtained for various operating modes in circuits suitable for wearable computers which are typically equipped with sensors that provide time series data (e.g., acceleration, ECG). Dynamic time warping (DTW) is considered a suitable signal processing technique for wearable computers, particularly due to its lower computational complexity requirement and the robustness to speed variations (acceleration and de-acceleration) in time series data. Wearable computers usually have very low computational performance requirements, which is explored in this work to minimize the system level energy consumption. We provide a comparison among three modes of operations, namely minimum energy operating point (MEOP), minimum voltage operation point (MVOP) and nominal voltage operating point (NVOP) all leveraging sleep transistors when circuits are inactive. The results show that the MVOP, in conjunction with sleep transistors, provides the least energy budget and leads to a reduction in energy consumption compared to the MEO, which is known as a suitable operating mode for ultra-low power circuits.

An adaptive deep learning approach for PPG-based identification

Wearable biosensors have become increasingly popular in healthcare due to their capabilities for low cost and long term biosignal monitoring. This paper presents a novel two-stage technique to offer biometric identification using these biosensors through Deep Belief Networks and Restricted Boltzman Machines. Our identification approach improves robustness in current monitoring procedures within clinical, e-health and fitness environments using Photoplethysmography (PPG) signals through deep learning classification models. The approach is tested on TROIKA dataset using 10-fold cross validation and achieved an accuracy of 96.1%.Wearable biosensors have become increasingly popular in healthcare due to their capabilities for low cost and long term biosignal monitoring. This paper presents a novel two-stage technique to offer biometric identification using these biosensors through Deep Belief Networks and Restricted Boltzman Machines. Our identification approach improves robustness in current monitoring procedures within clinical, e-health and fitness environments using Photoplethysmography (PPG) signals through deep learning classification models. The approach is tested on TROIKA dataset using 10-fold cross validation and achieved an accuracy of 96.1%.

Unsupervised EEG analysis for automated epileptic seizure detection

Epilepsy is a neurological disorder which can, if not controlled, potentially cause unexpected death. It is extremely crucial to have accurate automatic pattern recognition and data mining techniques to detect the onset of seizures and inform care-givers to help the patients. EEG signals are the preferred biosignals for diagnosis of epileptic patients. Most of the existing pattern recognition techniques used in EEG analysis leverage the notion of supervised machine learning algorithms. Since seizure data are heavily under-represented, such techniques are not always practical particularly when the labeled data is not sufficiently available or when disease progression is rapid and the corresponding EEG footprint pattern will not be robust. Furthermore, EEG pattern change is highly individual dependent and requires experienced specialists to annotate the seizure and non-seizure events. In this work, we present an unsupervised technique to discriminate seizures and non-seizures events. We employ power spectral density of EEG signals in different frequency bands that are informative features to accurately cluster seizure and non-seizure events. The experimental results tried so far indicate achieving more than 90% accuracy in clustering seizure and non-seizure events without having any prior knowledge on patients history.

Automatic limb identification and sleeping parameters assessment for pressure ulcer prevention

Pressure ulcers (PUs) are common among vulnerable patients such as elderly, bedridden and diabetic. PUs are very painful for patients and costly for hospitals and nursing homes. Assessment of sleeping parameters on at-risk limbs is critical for ulcer prevention. An effective assessment depends on automatic identification and tracking of at-risk limbs. An accurate limb identification can be used to analyze the pressure distribution and assess risk for each limb. In this paper, we propose a graph-based clustering approach to extract the body limbs from the pressure data collected by a commercial pressure map system. A robust signature-based technique is employed to automatically label each limb. Finally, an assessment technique is applied to evaluate the experienced stress by each limb over time. The experimental results indicate high performance and more than 94% average accuracy of the proposed approach.

Single and multi-subject clustering of flow cytometry data for cell-type identification and anomaly detection

BackgroundMeasurement of various markers of single cells using flow cytometry has several biological applications. These applications include improving our understanding of behavior of cellular systems, identifying rare cell populations and personalized medication. A common critical issue in the existing methods is identification of the number of cellular populations which heavily affects the accuracy of results. Furthermore, anomaly detection is crucial in flow cytometry experiments. In this work, we propose a two-stage clustering technique for cell type identification in single subject flow cytometry data and extend it for anomaly detection among multiple subjects.ResultsOur experimentation on 42 flow cytometry datasets indicates high performance and accurate clustering (F-measure > 91 %) in identifying main cellular populations. Furthermore, our anomaly detection technique evaluated on Acute Myeloid Leukemia dataset results in only <2 % false positives.

Gaits analysis using pressure image for subject identification

In this paper, a method for human identification using footprints obtained by a pressure sensor pad is proposed. The proposed method registers the sequence of pressure images in time and spatial domain and uses principle component analysis to produce a compact feature vector. The method is tested on footprints collected from 35 subjects and an accuracy of 97% is observed based on a 10-fold cross validation using SVM classifier.