Ghoraani, Behnaz

Deep learning strategies combined with wearable sensors have advanced the capabilities of monitoring systems in biomedical applications, offering precise and efficient solutions for diagnosing and managing diseases. However, applying these systems faces several challenges. One of the challenges is the diminishing performance when these systems encounter new data with more complex patterns than those seen before. Another challenge is the limited availability of labeled data, on which deep learning-based systems depend highly. Additionally, obtaining high-quality labeled data to train deep learning models is often expensive, requiring significant time and resources. Another significant challenge is ensuring the practicality, accessibility, and convenience of the monitoring systems.
This dissertation proposes an innovative deep learning framework to overcome these challenges and improve system generalization performance in classification and regression tasks, specifically monitoring patients with neurological disorders like Parkinson’s.

Atrial Fibrillation (AF) is a debilitating heart rhythm disorder affecting over 2.7 million people in the US and over 30 million people worldwide annually. It has a high correlation with causing a stroke and several other risk factors, resulting in increased mortality and morbidity rate. Currently, the non-pharmocological therapy followed to control AF is catheter ablation, in which the tissue surrounding the pulmonary veins (PVs) is cauterized (called the PV isolation - PVI procedure) aims to block the ectopic triggers originating from the PVs from entering the atrium. However, the success rate of PVI with or without other anatomy-based lesions is only 50%-60%.
A major reason for the suboptimal success rate is the failure to eliminate patientspecific non-PV sources present in the left atrium (LA), namely reentry source (a.k.a. rotor source) and focal source (a.k.a. point source). It has been shown from several animal and human studies that locating and ablating these sources significantly improves the long-term success rate of the ablation procedure. However, current technologies to locate these sources posses limitations with resolution, additional/special hardware requirements, etc. In this dissertation, the goal is to develop an efficient algorithm to locate AF reentry and focal sources using electrograms recorded from a conventionally used high-resolution multi-pole diagnostic catheter.