Development of an automated, machine learning-based methodology for yield point identification from tensile testing data of soft tissues

Biological soft tissue makes up many parts of the human body including blood vessels, skin, muscle, and nerves. Disorders that occur in soft tissue are wide-ranging in type and degree and require biomechanical inspection for treatment. One method of inspection is uniaxial extension testing, where a soft tissue specimen is clamped down to a machine and stretched along a single axis. Data recorded during extension can be interpreted as a curve that can be analyzed to determine biomechanical properties.

This study aims to investigate the yield point, which is a specific biomechanical property found in curves from extension testing data. When a soft tissue specimen is stretched, there exists a certain point where further extension will cause plastic deformation, or irreversible damage to the tissue; this is referred to as the yield point. The gold standard method of identifying this point is through visual inspection, which is unreliable and prone to subjectivity. An objective methodology for determining the yield point can enable further findings relating to patterns in soft tissue behavior.

The method used to identify the yield point in this study involves machine learning, which encompasses computer algorithms that improve automatically through experience. Machine learning is regarded as an effective method in distinguishing patterns in large datasets that may not be evident in standard analysis. Additionally, this study investigates the viability of two previously developed methods of identifying the yield point and evaluates the collective methodologies as tools for soft tissue examination.