A new image processing methodology to analyze local electromechanical properties of cardiomyocytes in 2D cultures

Cardiomyocytes in 2D culture are extensively used to address various basic and translational research questions in cardiology. The origin of such cultures can be from various sources. Among the most important parameters to address in all such cultures are the electromechanical properties of the cells. However, as opposed to adult cardiomyocytes, the cultured cells have disorganized sarcomeres and often exhibit unsynchronized beating patterns. Thus, accurate quantification of contraction properties is challenging. Recently, global motion estimation methodologies were developed to address this issue. Yet, such methodologies lack sensitivity to local differences in electromechanical properties in each culture.

In this study we aim to address the challenge of contraction analysis in 2D culture utilizing image-processing and machine learning using the Matlab working environment.

Initially, optical flow motion estimation and subsequently 2D convolution was applied to image sequences resulting in a spatio-temporal vector field map. The beating variation of each pixel location was calculated as the Euclidian norm of the vector field at each time step. The spatial locations were then clustered into distinct groups with respect to the correlation of beating variation using the Fuzzy C-Means algorithm. The optimal number of clusters was automatically selected using an entropy based cluster validity measure. Using this methodology, we found that unorganized colonies in 2D can be spatially separated by their beating pattern in an efficient and accurate fashion.

Our results indicate that this novel methodology is a promising way to accurately analyze the local contractile properties and the overall electromechanical synchrony in CMs in 2D culture. Further validation and quantitate estimation of the new methodology in comparison with existing ones in the literature is currently ongoing.