Assessment of the contractile properties of engineered heart tissues by optical flow

Due to their physiological and behavioral resemblance to in-vivo cardiac muscles, Engineered Heart Tissues (EHTs) are a promising tool for cardiovascular drug discovery, cell therapy and disease modeling. The beating pattern of EHTs is a direct indicator of their health and function, hence various techniques have been devised to gauge their contractile behavior; nevertheless, the established methods rely on invasive, expensive and delicate procedures. This thesis presents a novel workflow for the extraction, characterization and assessment of the contractile properties of EHTs that relies on motion estimation by optical flow (OF) and is purely based on bright-field images. It returns information in the form of (1) motion maps, (2) contraction profiles and (3) Histogram of Oriented Optical Flow, then classifies the tissue as correctly- or incorrectly-behaving according to the extracted characteristics of motion. Additionally, this paper tests the proposed method with three OF algorithms (block-matching, FlowNet2 and Farnebäck) and compares their performance on the given dataset: block-matching was found unsuitable for this study; FlowNet2 was only capable of analyzing tissues undergoing significant displacement while Farnebäck was configured with an iterative refinement to reach notable versatility and robustness. The iterative Farnebäck algorithm resulted in a classification accuracy of 88.8%, proving the applicability and potential of OF for the assessment of EHTs’ contractile properties.