Dresden 2020 – scientific programme
The DPG Spring Meeting in Dresden had to be cancelled! Read more ...
Parts | Days | Selection | Search | Updates | Downloads | Help
DY: Fachverband Dynamik und Statistische Physik
DY 47: Focus Session: Nonlinear Dynamics of the Heart I (joint session DY/BP)
DY 47.10: Talk
Thursday, March 19, 2020, 12:30–12:45, ZEU 118
Real-time Processing of Optical Fluorescence Videos showing Contracting Hearts using Neural Networks — •Jan Lebert1,2,3 and Jan Christoph1,2,3 — 1University Medical Center Göttingen, Germany — 2Max Planck Institute for Dynamics and Self-Organization, Göttingen, Germany — 3German Center for Cardiovascular Research, Partnersite Göttingen, Germany
Optical mapping is an established fluorescence imaging technique for studying electrophysiological wave phenomena in isolated, intact hearts and cardiac cell cultures. Mechanical contraction of the cardiac tissue, however, can lead to severe motion artifacts in the recorded optical signals. Pharmacological electromechanical uncoupling agents have been used to compensate for these artifacts. However, recently numerical motion tracking and post-processing algorithms were developed to suppress motion artifacts and separate the recorded electrical waves from mechanical contraction.
Here, we present a deep convolutional neural network (CNN) approach for the real-time tracking of contracting and fluorescing hearts in optical mapping videos. Our approach provides a dramatic speed-up in the processing of optical mapping data and superior performance over conventional optical flow estimation algorithms, which are sensitive to noise and can be irritated by fluorescence-encoded wave patterns, as they assume brightness consistency. After training the network on various experimental and synthetically generated optical mapping data, we evaluated the network’s performance and found it to perform robustly under various conditions.