Meniscus sizing using three‐dimensional models of the ipsilateral tibia plateau based on CT scans – an experimental study of a new sizing approach

Silvan Beeler, Lazaros Vlachopoulos, Lukas Jud, Reto Sutter, Tobias Götschi, Philipp Fürnstahl, Sandro F. FucenteseJournal of Experimental Orthopaedics 2020 · DOI

TL;DR

This study introduces a novel, automated method for sizing meniscal allografts using 3D CT scans of the ipsilateral tibia plateau. This approach aims to improve the accuracy of allograft selection for knee surgery, moving beyond traditional 2D methods by leveraging the anatomical relationship between the tibia and meniscus. While promising, further refinement is needed to consistently identify the absolute best-fitting allograft.

What Problem Does This Paper Address?

Meniscus allograft transplantation is a crucial treatment for patients experiencing symptoms after meniscectomy. Achieving good clinical outcomes hinges on selecting an allograft that closely matches the size and shape of the native meniscus. However, directly measuring the 3D shape of a damaged or absent meniscus is challenging. Current methods rely on indirect sizing using conventional radiography (2D-RX) or MRI, which can be inaccurate and time-consuming, especially when dealing with a healthy contralateral meniscus as a reference. This paper tackles the problem of developing a more accurate, automated, and indirect sizing method that is independent of the meniscus's condition.

Methods

The researchers developed a 3D-CT sizing approach that utilizes 3D surface models of the proximal tibia plateau. The core idea is that a similar tibia plateau shape implies a similarly shaped meniscus.

  1. 3D Model Creation: High-resolution MRI and CT scans from 50 healthy knee joints were used to create detailed 3D surface models of both the menisci (from MRI) and the tibia plateau (from CT).
  2. Tibia Plateau Segmentation: Specific portions of the tibia plateau were segmented from the CT scans. Three configurations were tested: the entire plateau, half the plateau including the intercondylar area, and half the plateau excluding the intercondylar area. This exclusion was based on the hypothesis that the intercondylar region is less directly correlated with meniscus size.
  3. 3D-CT Sizing Algorithm: The study simulated a "tissue bank" of 49 menisci for each of the 50 original menisci. The 3D tibia plateau models were then used to find the best matching "allograft" tibia plateau based on surface distance. The assumption is that the best-fitting tibia plateau would house the best-fitting meniscus.
  4. Validation: The performance of the 3D-CT sizing method was compared against two established methods: 3D-MRI sizing (considered a gold standard for direct sizing) and conventional 2D-RX sizing (Pollard method). The comparison focused on mean and maximum surface distances, as well as differences in width, length, and height between the original and selected allografts.

Datasets

The study utilized a dataset of 50 unilateral knee joints from 50 patients. These joints were characterized by:

  • Mature skeletal age (closed growth plates).
  • Healthy menisci (no tears, degeneration, or extrusion).
  • No tibio-femoral osteoarthritis (Kellgren & Lawrence grade 0).
  • Available MRI (sagittal/coronal/axial, 3mm slice thickness).
  • Available CT (axial, ≤2mm slice thickness).
  • Available conventional radiography (AP/LAT).

The dataset comprised 18 left and 32 right knee joints, with a mean age of 27.6 years (range 16-46 years), including 35 females and 15 males. This dataset served as the basis for creating the 3D models and simulating the "tissue bank" for validation.

Key Insights and Ideas

  • Tibia Plateau as a Proxy: The fundamental insight is that the 3D shape of the tibia plateau can serve as a reliable indirect indicator of the corresponding meniscus's 3D size and shape. This is a significant departure from methods that rely on direct meniscus measurements or contralateral references.
  • Importance of Intercondylar Exclusion: Excluding the intercondylar area from the tibia plateau segmentation yielded the most accurate sizing results for both medial and lateral menisci. This suggests that the central intercondylar region is less critical for accurate meniscus sizing compared to the peripheral aspects of the plateau.
  • Automation is Key: The study emphasizes the feasibility of an automated, indirect sizing method. This reduces the potential for inter-observer variability inherent in manual measurements and speeds up the process.
  • 3D-CT Superiority over 2D-RX: The 3D-CT approach demonstrated superior accuracy compared to the traditional 2D-RX method, particularly for the medial meniscus, in terms of reducing outliers and improving overall fit.

Key Results

  • 3D-CT Sizing Performance: The "half tibia plateau without the intercondylar eminence" configuration provided the most reliable sizing results, showing significant differences in mean and maximum surface distances for both medial and lateral menisci compared to other configurations.
  • Comparison with 2D-RX: 3D-CT sizing was significantly better than 2D-RX for the medial meniscus. While not statistically significant for the lateral meniscus, the trend favored 3D-CT.
  • Outlier Reduction: The 3D-CT method significantly reduced outliers (mismatches) by 56% for the medial meniscus and 9% for the lateral meniscus compared to the 2D-RX method.
  • Limitations in Perfect Matching: Despite improvements, neither the 3D-CT nor the 2D-RX method could consistently select the absolute best fitting allograft (as determined by 3D-MRI) in every case. The mean surface distance differences, while small, indicate that the 3D shape wasn't perfectly reproduced.

Limitations and Future Work

  • Imaginary Tissue Bank: The study used a simulated "tissue bank" of 49 menisci. A larger, real-world tissue bank with a greater variety of allografts might yield different results, potentially improving accuracy.
  • Anatomical Mismatch: The study acknowledges the possibility of inter-individual anatomical differences between the tibia plateau and meniscus correlation, which could contribute to sizing inaccuracies.
  • Segmentation Resolution: The retrospective nature of the study meant some MRI scans had a slice thickness of up to 3mm, which could potentially affect the accuracy of the 3D surface models.
  • Knee Movement: While scans were performed with the knee extended, the natural deformation of the meniscus during knee movement is a potential limitation.
  • Future Work: The authors suggest that further improvements are possible, particularly through machine learning or individual adjustment of cutting planes for the tibia plateau. Developing a digital database of tibia plateau and corresponding meniscus models is also proposed. The ultimate goal is to achieve a method that can reliably identify the best possible 3D meniscus allograft, even from limited tissue banks.

Summary updated

Abstract

Automatized, indirect meniscus sizing using the 3D bone models of the tibia plateau is feasible and more precise than the previously described 2D-RX method.. However, further technical improvement is needed to select always the best available allograft.
BibTeX
@article{beeler2020meniscus,
  author = {Silvan Beeler and Lazaros Vlachopoulos and Lukas Jud and Reto Sutter and Tobias Götschi and Philipp Fürnstahl and Sandro F. Fucentese},
  title = {{Meniscus sizing using three‐dimensional models of the ipsilateral tibia plateau based on CT scans – an experimental study of a new sizing approach}},
  journal = {Journal of Experimental Orthopaedics},
  year = {2020},
  doi = {10.1186/s40634-020-00252-8},
}
Figure