Localization of Origins of Premature Ventricular Contraction by Means of Convolutional Neural Network from 12-lead ECG
Screening for Cognitive Impairment by Model Assisted Cerebral Blood Flow Estimation
Changes in regional cerebral blood flow (CBF) have been linked with blood-brain barrier dysfunction in Alzheimer’s disease (AD). Screening for these CBF changes by out-of-clinic portable measurmenet devices has been proposed for improving the diagnosis of AD. We combine physiological modelling with blood pressure monitoring and carotid ultrasound imaging, where the computer model predicts 24-hour CBF profiles. These profiles are used in a machine learning classifier to improve dementia diagnosis.
Unambiguous Identification and Visualization of an Acoustically Active Catheter by Ultrasound Imaging in Real Time: Theory, Algorithm, and Phantom Experiments
To enable better intracardiac guidance and visualization, a prototype of an acoustically active catheter with a piezoelectric element on its distal end is used. The unique symmetric Doppler shifts from the piezoelectric element are detected by the proposed algorithm and the distal end is visualized with a unique color (yellow) in real-time (22- 50 Hz). The algorithm is robust to interference from blood flow with a localization range of few millimeters.
Computerized Lung Sound Screening for Pediatric Auscultation in Noisy Field Environments
While widely ubiquitous in clinical practice, stethoscopes are imperfect tools riddled with many shortcomings including inter-listener variability, subjectivity, and vulnerability to noise. In the present work, we explore the use of computer-aided screening of lung sounds un order to automate detection of abnormal breathing patterns indicative of respiratory pathologies. A proposed scheme combines an extended noise-suppression with a multiresolution analysis in order to screen abnormal breathing patterns from auscultation in >1000 children recorded in noisy clinical settings.
Computationally Efficient Algorithms for Sparse, Dynamic Solutions to the EEG Source Localization Problem
Estimating the currents that underlie the field potentials captured by electroencephalography (EEG), i.e., EEG source localization, is an ill-conditioned inverse problem. Existing solutions consider spatial continuity constraints, dynamic modeling, or sparsity constraints. The computational cost of combining these approaches, however, poses a challenge for practical applications. We propose a new computationally efficient EEG source localization method that employs spatial covariance estimation, state-space modeling, and sparsity-enforcing priors. We validate the performance of our method using both simulated and experimentally recorded EEG data. Our approach provides substantial performance improvements over existing methods and thereby facilitates practical applications in both neuroscience and medicine.
Micromagnetic Stimulation of the Mouse Auditory Cortex In Vivo Using an Implantable Solenoid System
Micromagnetic stimulation (µMS), using a submillimeter-sized coil to stimulate nerves, has the possibility to overcome the limitations of conventional methods characterized by their large device size and poor focality. However, little is known about µMS effects on brain activity. To validate µMS effectiveness, we developed an implantable µMS interface, evaluated its physical characteristics, and observed µMS-driven responses in mouse brain.
A Functional-genetic Scheme for Seizure Forecasting in Canine Epilepsy
Seizure prediction is currently a major concern in the epilepsy research community. In this work, we have proposed a new strategy to achieve accurate seizure forecasting by combining effective connectivity measures and artificial intelligence techniques. Results show performance improvement compared to previous studies, achieving average sensitivity of 84.82% and time in warning of 0.1.