Model-based validation of a tracking algorithm to quantify strain in muscles from two-dimensional ultrasound images

Model-based validation of a tracking algorithm to quantify strain in muscles from two-dimensional ultrasound images

de Graaf, Jurjan (TU Delft Mechanical, Maritime and Materials Engineering)

Schouten, Alfred (mentor)
Mugge, Winfred (graduation committee)
Kasbergen, Cor (graduation committee)
Maas, Huub (graduation committee)
Herbert, Rob (graduation committee)
Bolsterlee, Bart (mentor)

Delft University of Technology

Mechanical Engineering | Biomechanical Design - BioRobotics

2019-04-17

Fascicle shear is theoretically a mechanism by which skeletal muscles lengthen. Existing ultrasound based techniques allow measurement of muscle architecture parameters, but not the quantification of shear strain. In this study, an algorithm that tracks the shear strain from ultrasound images using a continuum representation of the muscle was developed. Errors in strain tracking may arise due to misalignment of the imaging plane, movement artifacts and non-uniform strain within the muscle. The goal of this study was to develop and validate the newly developed strain tracking algorithm. Various potential sources of error are investigated. A computer model was created consisting of a three-dimensional (3D) synthetic volume, representing the muscle. Virtual ultrasound images were then sampled from the 3D volume by intersecting the synthetic volume under a known angle of the imaging plane. The measurement error was defined as the difference between the known strain that was imposed virtually to the 3D muscle volume and the strain calculated with the tracking algorithm. The measurement error was determined for conditions of combined axial and shear strain, plane misalignment, plane rotation and non-uniform strain. Conditions were simulated between 30 and 100 times, each time with a different synthetic muscle volume.
The developed strain tracking algorithm provided strain measurement of sub-pixel accuracy and precision when the imaging plane was aligned with the fascicles. Rotation of the ultrasound transducer relative to the muscle resulted in invalid measurements. Axial strain was overestimated when the muscle exhibited a non-uniform axial strain pattern. Largest errors (underestimation of strain by up to 65%) were caused by misalignment of the imaging plane with the fascicles. The large effect of misalignment emphasizes the need for careful transducer placement that requires anatomical information about the muscle structure. Strain tracking methods based on three-dimensional avoid the need for alignment, potentially allowing more accurate measurement of strain.

tracking
muscle
ultrasound imaging
speckle
digital image
Correlation
continuum mechanics

http://resolver.tudelft.nl/uuid:e4097a80-9ed5-4faa-8ef3-d36bab4fe0b9

Student theses

master thesis

© 2019 Jurjan de Graaf