You are here
New publication: Characterization of Myocardial Motion Patterns by Unsupervised Multiple Kernel Learning
We propose an independent objective method to characterize different patterns of functional responses to stress in the heart failure with preserved ejection fraction (HFPEF) syndrome by combining multiple temporally-aligned myocardial velocity traces at rest and during exercise, together with temporal information on the occurrence of cardiac events (valves openings/closures and atrial activation). The method builds upon multiple kernel learning, a machine learning technique that allows the combination of data of different nature and the reduction of their dimensionality towards a meaningful representation (output space). The learning process is kept unsupervised, to study the variability of the input traces without being conditioned by data labels. To enhance the physiological interpretation of the output space, the variability that it encodes is analyzed in the space of input signals after reconstructing the velocity traces via multiscale kernel regression. The methodology was applied to 2D sequences from a stress echocardiography protocol from 55 subjects (22 healthy, 19 HFPEF and 14 breathless subjects). The results confirm that characterization of the myocardial functional response to stress in the HFPEF syndrome may be improved by the joint analysis of multiple relevant features.
Computational techniques in biomedical engineering are the main focus of our master. This comprises the quantitative analysis of biomedical images and signals as well as the modeling of living organisms and medical devices. Starting at the cellular level up to the level of whole organisms, the different analysis and modeling projects will cover many different scales of biological organization. Concrete examples include the analysis of electroencephalographic recordings from epilepsy patients, the design of surgical tools, or the self-organization of single cells and cellular populations in health and disease.
For more information and registration please visit the master's website.
Integration of electrical and structural information for scar characterization in the left ventricle (LV) is a crucial step to better guide radio-frequency ablation therapies, which are usually performed in complex ventricular tachycardia (VT) cases. This integration requires finding a common representation where to map the electrical information from the electro-anatomical map (EAM) surfaces and tissue viability information from delay-enhancement magnetic resonance images (DE-MRI). However, the development of a consistent integration method is still an open problem due to the lack of a proper evaluation framework to assess its accuracy. In this paper we present both: (i) an evaluation framework to assess the accuracy of EAM and imaging integration strategies with simulated EAM data and a set of global and local measures; and (ii) a new integration methodology based on a planar disk representation where the LV surface meshes are quasi-conformally mapped (QCM) by flattening, allowing for simultaneous visualization and joint analysis of the multi-modal data. The developed evaluation framework was applied to estimate the accuracy of the QCM-based integration strategy on a benchmark dataset of 128 synthetically generated ground-truth cases presenting different scar configurations and EAM characteristics. The obtained results demonstrate a significant reduction in global overlap errors (50-100%) with respect to state-of-the-art integration techniques, also better preserving the local topology of small structures such as conduction channels in scars. Data from seventeen VT patients were also used to study the feasibility of the QCM technique in a clinical setting, consistently outperforming the alternative integration techniques in the presence of sparse and noisy clinical data. The proposed evaluation framework has allowed a rigorous comparison of different EAM and imaging data integration strategies, providing useful information to better guide clinical practice in complex cardiac interventions.
We congratulate Dr. Daniel Romero García, who defended his dissertation on January 28th. Well done Dani ;).