Characterization of myocardial motion patterns by unsupervised multiple kernel learning

• Published in: Medical image analysis Vol. 35; pp. 70 - 82
• Main Authors: Sanchez-Martinez, Sergio, Duchateau, Nicolas, Erdei, Tamas, Fraser, Alan G., Bijnens, Bart H., Piella, Gemma
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.01.2017; Elsevier BV; Elsevier
• Subjects: Case-Control Studies; Computer Science; Conditioning; Echocardiography; Exercise - physiology; Exercise Test; Heart; Heart - diagnostic imaging; Heart - physiology; Heart diseases; Heart failure; Heart Failure - diagnostic imaging; Heart Failure - physiopathology; Humans; Learning algorithms; Machine learning; Medical Imaging; Motion; Movement; Multiple kernel learning; Multiscale analysis; Myocardial motion; Pattern analysis; Regression analysis; Reproducibility of Results; Rest - physiology; Sensitivity and Specificity; Stress response; Stresses; Unsupervised Machine Learning; Variability; Velocity; Multiple kernel learning; Myocardial motion; Echocardiography; Pattern analysis; pattern analysis; multiple kernel learning; echocardiography
• ISSN: 1361-8415; 1361-8423
• DOI: 10.1016/j.media.2016.06.007

•Multiple myocardial velocity patterns from a stress protocol are jointly analyzed.•Unsupervised Multiple Kernel Learning is used to reduce the complexity of the data.•The variability analysis on the learnt space unravels healthy/diseased differences.•The joint analysis of multiple patterns notably improves the characterization. [Display omitted] 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.