Search

In this paper, we propose bag of adversarial features (BAFs) for identifying mild traumatic brain injury (MTBI) patients from their diffusion magnetic resonance images (MRIs) (obtained within one month of injury) by incorporating unsupervised feature learning techniques. MTBI is a growing public health problem with an estimated incidence of over 1.7 million people annually in USA. Diagnosis is based on clinical history and symptoms, and accurate, concrete measures of injury are lacking. Unlike most of the previous works, which use hand-crafted features extracted from different parts of brain for MTBI classification, we employ feature learning algorithms to learn more discriminative representation for this task. A major challenge in this field thus far is the relatively small number of subjects available for training. This makes it difficult to use an end-to-end convolutional neural network to directly classify a subject from MRIs. To overcome this challenge, we first apply an adversarial auto-encoder (with convolutional structure) to learn patch-level features, from overlapping image patches extracted from different brain regions. We then aggregate these features through a bag-of-words approach. We perform an extensive experimental study on a dataset of 227 subjects (including 109 MTBI patients, and 118 age and sex-matched healthy controls) and compare the bag-of-deep-features with several previous approaches. Our experimental results show that the BAF significantly outperforms earlier works relying on the mean values of MR metrics in selected brain regions. [ABSTRACT FROM AUTHOR]

Epilepsy is one of the most common neurological disorders affecting a significant portion of the world’s population and approximately 2.5 million people in the United States. Important biomedical research effort is focused on the development of energy efficient devices for the real-time detection of seizures. In this paper, we propose an Internet of Medical Things (IoMT)-based automated seizure detection system which will detect a seizure from electroencephalography (EEG) signals using a voltage level detector (VLD) and a signal rejection algorithm (SRA). The proposed system analyzes neural signals continuously and extracts the hyper-synchronous pulses for the detection of seizure onset. Within a time frame, if the number of pulses exceeds a predefined threshold value, a seizure is declared. The SRA reduces false detections, which in turn enhances the accuracy of the seizure detector. The design was validated using system-level simulations and consumer electronics proof of concept. The proposed seizure detector reports a sensitivity of 96.9% and specificity of 97.5%. The use of minimal circuitry can lead to reduction of power consumption compared to many contemporary approaches. The proposed approach can be generalized to other sensor modalities and the use of wearable or implantable solutions, or a combination of the two. [ABSTRACT FROM AUTHOR]

Epilepsy, which is characterized by recurrent spontaneous seizures, has a considerably negative impact on both the quality and the expectancy of life of the patient. Approximately 3.4 million individuals in the USA and up to 1% of the world population is afflicted by epilepsy. This necessitates the real-time detection of seizures which can be done by the use of an Internet of Things (IoT) framework for smart healthcare. In this paper, we propose an electroencephalogram (EEG)-based seizure detection system in the IoT framework which uses the discrete wavelet transform (DWT), Hjorth parameters (HPs), statistical features, and a machine learning classifier. Seizure detection is done in two stages. In the first stage, EEG signals are decomposed by the DWT into sub-bands and features (activity, signal complexity, and standard deviation) were extracted from each of these sub-bands. In the second stage, a deep neural network (DNN) classifier is used to classify the EEG data. A prototype of the proposed neuro-detect was implemented using the hardware-in-the-loop approach. The results demonstrate a significant difference in HP values between interictal and ictal EEG with ictal EEG being less complex than interictal EEG. In this approach, we report an accuracy of 100% for a classification of normal versus ictal EEG and 98.6% for normal and interictal versus ictal EEG. [ABSTRACT FROM AUTHOR]