Search for: All records

Epilepsy is a brain disorder that causes seizures, affecting nearly half a million children in the US alone. In this study, we aimed to use a nonlinear driven method to characterize scalp EEG recordings of pediatric epilepsy patients (PE: n=7 ) compared to pediatric control subjects (PC: n=7 ) in a clinical environment. A time-varying approach was used to construct functional connectivity networks (FCNs) of all subjects. Next, the FCNs are mapped into the form of undirected graphs that are subjected to the extraction of graph theory-based features. An unsupervised clustering technique based on K-mean is used to delineate the PE from the PC group. Our findings show a statistically significant difference in the mean FCNs between PC and PE groups (t(340)=−15.9899,p<<0.0001) . Performance results showed an accuracy of 92.5% with a sensitivity of 90% and a specificity of 95.3%. This approach can help improve and validate the early diagnosis of PE by applying non-invasive scalp EEG signals. 

Analyzing the hippocampus in the brain through magnetic resonance imaging (MRI) plays a crucial role in diagnosing and making treatment decisions for several neurological diseases. Hippocampus atrophy is among the most informative early diagnostic biomarkers of Alzheimer's disease (AD), yet its automatic segmentation is extremely difficult given the anatomical structure of the brain and the lack of any contrast in between its different regions. The gold standard remains manual segmentation and the use of brain atlases. In this study, we use a well-known image segmentation model, UNet++, and introduce an attention mechanism called the Convolutional Block Attention Module (CBAM) to the UNet++ model. This integrated model improves the feature weights of our region of interest, and hence increases the accuracy in segmenting the hippocampus. Results show averages of 0.8715, 0.8107, 0.8872, and 0.9039 for the metrics of Dice, Jaccard, Precision, and Recall, respectively. 

The Standard Uptake Value (SUV) is conventionally calculated using the ratio of the injected PET radiotracer dose and subject body weight (Binj) . SUVs are used to obtain SUV ratios (SUVr), an important metric in many Alzheimer's Disease (AD) neuroimaging studies. However, SUVr can be obtained using only neuroimaging data, bypassing the need for Binj . This paper proposes the SUVr-LightWeight (SUVr-LW) algorithm which is not reliant on clinical data and instead focuses on PET intensity values. The SUVr-LW was evaluated using the Centiloid Project Florebetaben (FBB) subject cohort and reached a linear regression slope of 0.98, while the healthy control subjects produced a slope of 0.87. 

Modeling time-series data with asynchronous, multi-cardinal, and uneven patterns presents several unique challenges that may impede convergence of supervised machine learning algorithms, or significantly increase resource requirements, thus rendering modeling efforts infeasible in resource-constrained environments. The authors propose two approaches to multi-class classification of asynchronous time-series data. In the first approach, they create a baseline by reducing the time-series data using a statistical approach and training a model based on gradient boosted trees. In the second approach, they implement a fully convolutional network (FCN) and train it on asynchronous data without any special feature engineering. Evaluation of results shows that FCN performs as well as the gradient boosting based on mean F1-score without computationally complex time-series feature engineering. This work has been applied in the prediction of customer attrition at a large retail automotive finance company.