Summary

Lateral extra-articular tenodesis augmentation of anterior cruciate ligament reconstruction did not affect the cartilage thickness of the lateral compartment within the tibiofemoral joint

Abstract

Introduction

Lateral extra-articular tenodesis (LET) augmentation of anterior cruciate ligament reconstruction (ACLR) has been reported to reduce failure rate and improve return to pre-injury level of sports participation (Getgood et al. 2022; Sonnery-Cottet et al. 2017). Notwithstanding, conflicting data exists on whether LET can result in over-constraint and degenerative changes of the lateral compartment within the tibiofemoral joint (Castoldi et al. 2020; Neri et al. 2021; Devitt et al. 2017).
Knee magnetic resonance imaging (MRI) but not knee radiographic data can objectively access the knee cartilage structure. However, clinical assessment of cartilage has been typically subjective because of the significant cost of segmenting the knee cartilage manually. Machine learning can automate this segmentation process and provide a quantitative assessment of knee cartilage. The hypothesis was that there was no difference in cartilage data on the lateral tibiofemoral compartment between ACLR with- and without- concomitant LET.

Methods

Two cohorts comprising 40 patients were matched for age, gender, height, weight, time from surgery to postoperative MRI, and graft diameter. Half of the patients had ACLR alone and the other half underwent concomitant LET with ACLR. All patients underwent a proton density-weighted MRI sequence of their operated knee joints. MONAI (Diaz-Pinto et al. 2024), 3D Slicer (Fedorov et al. 2012) and ITK-SNAP (Paul et al. 2006) was used to train perform deep-learning (DL), visualisation and manual segmentation of the knee cartilage, respectively. An experienced rater manually segmented the knee cartilages of all 40 patients. The dataset was divided into 2 sets: 80% for training and 20% for validation. Cartilage morphology was quantified using cartilage thickness data; each cartilage slice was flattened into a 2D structure (Monu et al. 2017).

Results

There was no statistical difference in post-operative clinical scores (IKDC, Lysholm and Tegner scores) and side-to-side knee laxity between the two groups. All DL predicted cartilage segmentation was visually comparable with the manually segmented ground truth. The Dice similarity coefficient for the ACLR with- and without- concomitant LET cohorts was 0.78±0.08 and 0.81±0.05, respectively. The thickness of anterior and posterior regions of the lateral femoral cartilage in the two cohorts: ACLR with LET = 1.81±0.26mm (anterior) and 2.12±0.31mm (posterior); ACLR without LET = 1.85±0.38mm (anterior) and 2.24±0.42mm (posterior); p>0.05 for both anterior and posterior regions. The thickness of anterior and posterior regions of the lateral tibia cartilage in the two cohorts: ACLR with LET = 2.20±0.42mm (anterior) and 2.77±0.35mm (posterior); ACLR without LET = 2.28±0.46mm (anterior) and 2.81±0.38mm (posterior); p>0.05 for both anterior and posterior regions.

Discussion

We determined that the described DL workflow of generating high Dice similarity coefficients against ground truths. The null hypothesis of no difference between ACLR with- and without- concomitant LET was accepted. Compared between two matched group of ACLR with- and without- concomitant LET cohorts, we found no clinical differences and cartilage thickness at different regions in the sagittal plane. ACLR with concomitant LET did not result in degenerative changes in lateral tibiofemoral compartment at 1-year post ACLR.