In-situ assessment of collateral ligaments strain could be key to improving total knee arthroplasty outcomes by improving the ability of surgeons to properly balance the knee intraoperatively. Ultrasound (US) speckle tracking methods have shown promise in their capability to measure in-situ soft tissue strain in large tendons but prior work has also highlighted the challenges that arise when attempting to translate these approaches to the in-situ assessment of collateral ligaments strain. Therefore, the aim of this project was to develop and validate an US speckle tracking method to specifically assess in-situ strains of both the MCL and LCL. We hypothesize that coefficients of determination (R²) would be above 0.90 with absolute differences below 0.50% strain for the comparison between US-based and the reference strain, with better results expected for the LCL compared with the MCL.

Five cadaveric legs with total knee implants (NH019 2017-02-03) were submitted to a varus (LCL) and valgus (MCL) ramped loading (0 – 40N). Ultrasound radiofrequency (rf) data and reference surface strains data, obtained with 3D digital image correlation (DIC), were collected synchronously. Prior to processing, US data were qualitatively assessed and specimens displaying substantial imaging artefacts were discarded, leaving five LCL and three MCL specimens in the analysis. Ultrasound rf data were processed in Matlab (The MathWorks, Inc., Natick, MA) with a custom-built speckle tracking approach adapted from a method validated on larger tendons and based on normalized cross-correlation. Digital image correlation data were processed with commercial software VIC3D (Correlated Solutions, Inc., Columbia, SC). To optimize speckle tracking, several tracking parameters were tested: kernel and search window size, minimal correlation coefficient and simulated frame rate. Parameters were ranked according to three comparative measures between US- and DIC-based strains: R², mean absolute error and strains differences at 40N. Parameters with best average rank were considered as optimal.

To quantify the agreement between US- and DIC-based strain of each specimen, the considered metrics were: R², mean absolute error and strain differences at 40N. The LCL showed a good agreement with a high average R² (0.97), small average mean absolute difference (0.37%) and similar strains at 40N (DIC = 2.92 ± 0.10%; US = 2.99 ± 1.16%). The US-based speckle tracking method showed worse performance for the MCL with a lower average correlation (0.55). Such an effect has been observed previously and may relate to the difficulty in acquiring sufficient image quality for tracking the MCL compared to the LCL, which likely arises due to structural or mechanical differences; notably MCL is larger, thinner, more wrapped around the bone and stretches less. However, despite these challenges, the MCL tracking still showed small average mean absolute differences (0.44%) and similar strains at 40N (DIC = 1.48 ± 0.06%; US = 1.44 ± 1.89%).

We conclude that the ultrasound speckle tracking method developed is ready to be used as a tool to assess in-situ strains of LCL. Concerning the MCL strain assessment, despite some promising results in terms of strain differences, further work on acquisition could be beneficial to reach similar performance.

Introduction

Tendinopathies are among the most common musculoskeletal injuries. Nowadays, part of its diagnosis is established through subjective qualitative evaluation of 2D ultrasound (US). This enables limited diagnostic differentiation or therapeutic optimization and has limited added value to diagnosis in an earlier stage. It is generally accepted that extra diagnostic information can be obtained via strain evaluation. The accurate validation of strain estimation is challenging due to the lack of a ground-truth. Therefore we evaluate the repeatability of displacement and strain estimations in the longitudinal direction, using an easy, fast and interactive application to estimate local strain during dynamic loading of the tendon.