Agenda

CAS MSc Midterm Presentations

• Thursday, 21 October 2021
• 11:00-12:00
• Online [zoom]

Few-Shot Learning Using Speech and Physiological Signals for Emotion Recognition for Intelligent Voice Assistants

CAS MSc Kick-off Presentations

• Thursday, 21 October 2021
• 11:00-12:00
• Online [zoom]

Improving the Accuracy of the Image Search Engine for Digital History

MSc SS Thesis Presentation

• Friday, 22 October 2021
• 10:00-12:00
• TBA

Adaptive Graph Partition Methods for Structured Graphs

Graphs can be models for many real-world systems, where nodes indicate the entities and edges indicate the pairwise connections in between. In various cases, it is important to detect informative subsets of nodes such that the nodes within the subsets are 'closer' to each other. For example, in a cellular network, determining appropriate node subsets can reduce the operation costs. A subset is usually called a cluster. This leads to the graph clustering problem. Furthermore, plenty of systems in the real world are changing over time, and consequently, graphs as models vary with time as well. It is thus also important to update the clusters when the graph changes.

In this thesis work, we studied two problems from the cellular network background. We needed to partition graphs that have certain structures and cluster their nodes to minimize certain cost functions. In the first problem, we partitioned a bipartite graph by minimizing the so-called MinMaxCut cost function, while in the second problem, we partitioned a structured graph by minimizing the so-called Modified-MinMaxCut cost function. The structural property of the graph is incorporated in defining this new cost function. The solutions we proposed are under the framework of spectral clustering, where one relies on the eigenvectors of the graph matrices, e.g., the Laplacian matrix or the adjacency matrix, and any clustering algorithm, e.g., K-means, to partition nodes into disjoint clusters.

Furthermore, for the time-variant graph, we decomposed the problem into two steps. First, we transformed the variations in the graph topology into perturbations to the graph matrices. Then we transformed the update of the clusters into an update of the (generalized) eigenvectors of these graph matrices. We utilized matrix perturbation theory to update the generalized eigenvectors and then update the clusters. Our simulations showed that on synthetic data, the proposed method can efficiently track the eigenvectors and the clusters generated by the updated eigenvectors have almost the same cost function value as that of exact computation.

PhD Thesis Defence

• Wednesday, 27 October 2021
• 12:00
• Senaatszaal of Aula Building

Atrial Fibrillation Fingerprinting

Atrial fibrillation (AF) is a common age-related cardiac arrhythmia. AF is characterized by rapid and irregular electrical activity of the heart leading to a higher risk of stroke and heart failure. During AF, the upper chambers of the heart, called atria, experience chaotic electrical wave propagation. However, despite the various mechanisms introduced in the literature, there is still an ongoing debate on a precise and consistent mechanism underlying the initiation and perpetuation of AF. Some studies show that AF is rooted in impaired electrical conduction and structural damage of atrial tissue, known as electropathology. Atrial electrograms (EGMs) recorded directly from heart’s surface, provide an important diagnostic tool to localize and quantify the degree of electropathology in the tissue. However, the analysis of the electrograms is currently constrained by the lack of suitable methods that can reveal the hidden electrophysiological parameters of the tissue. These parameters can be used as local indication of electropathology in the tissue. We believe that understanding AF and improving AF therapy starts with developing a proper forward model that is accurate enough (from a physiological point of view) and simultaneously simple enough to allow for subsequent parameter estimation. Therefore, the main focus of this thesis is on developing a simplified forward model that can efficiently explain the observed EGM based on AF relevant tissue parameters. An initial step before performing any analysis on the data is to remove noise and artefacts. All atrial electrogram recordings suffer from strong far-field ventricular activities (VA). Therefore, as the first step, we propose a new framework for removal of VA from atrial electrograms, which is based on interpolation and subtraction followed by low-rank and sparse matrix decomposition. The proposed framework is of low complexity, does not require high resolution multi-channel recordings, or a calibration step for each individual patient. In the next step, we develop a simplified electrogram model. We represent the model in a compact matrix form and show its linear dependence on the conductivity vector, enabling the estimation of this parameter from the recorded electrograms. The results show that despite the low resolution and all simplifying assumptions, the model can efficiently estimate the conductivity map and regenerate realistic electrograms, especially during sinus rhythm. In the next contribution of this dissertation, we propose a new approach for a better estimation of local activation times for atrial mapping by reducing the spatial blurring effect that is inherent to electrogram recordings using deconvolution. Employing sparsity based regularization and first-order time derivatives in formulating the deconvolution problem, improved performance of transmembrane current estimation is obtained. In the final part, we focus on translating our findings from research to clinical application. Therefore, we studied the effect of electrode size on electrogram properties including the length of the block line observed on the resulting activation map, percentage of observed low voltage areas, percentage of electrograms with low maximum steepness, and the number of deflections in the recorded electrograms.