A concept for movement-based computerized segmentation of connective tissue in ultrasound imaging

Ramakrishnan, Ananth Hari, RAJAPPA, Muthaiah, Krithivasan, Kannan, Chatzistergos, Panagiotis E., CHOCKALINGAM, Nachiappan and Nalluri, Madhusudhana Rao (2022) A concept for movement-based computerized segmentation of connective tissue in ultrasound imaging. Multimedia Tools and Applications. ISSN 1380-7501

Text Final accepted.docx - AUTHOR'S ACCEPTED Version (default) Available under License All Rights Reserved (Under Embargo). Download (1MB)

Abstract or description

This study proposes a novel concept for the computerized segmentation of ultrasound images of connective tissue based on movement. Tendons and ligaments are capable of almost frictionless movement relative to their neighbouring tissues making them good candidates for movement-based segmentation. To demonstrate this concept, a central cross section of the patellar tendon was imaged in the axial plane while movement was generated by manually pulling and pushing the skin close to the imaging area. Maps of internal movement were created for four representative pairs of consecutive images using normalised cross corelation. Thresholding followed by a series of morphological operations (k-clustering, blob extraction, curve fitting) enabled the extraction of the superficial-most tendon boundary. Comparison against manually segmented outputs indicated good agreement against ground truth (average ± STDEV Bhattacharyya distance: 0.170 ± 0.039). In contrast to the more superficial parts of the tissue, the applied method for motion generation did not result in clearly visible movement in the tissue areas deeper in the imaging window. The segmentation of the entire tendon will require movement patterns that involve equally the entire tendon (e.g., generated by a contraction of the in-series muscle). The results of this study demonstrate for the first time that movement mapping can be used for the segmentation of connective tissue. Further research will be needed to identify the optimal way to use motion to complement existing segmentation approaches which are based on signal intensity, texture, and shape features.

Item Type:	Article
Uncontrolled Keywords:	Tendon, Ligament, Connective tissue Segmentation algorithm, Normalized cross correlation, Displacement map
Faculty:	School of Life Sciences and Education > Sport and Exercise
Depositing User:	Panagiotis CHATZISTERGOS
Date Deposited:	10 May 2022 11:12
Last Modified:	23 Apr 2023 01:38
URI:	https://eprints.staffs.ac.uk/id/eprint/7305

Actions (login required)

View Item