Knee focal cartilage defect location heat map and local surface morphology characterisation: Insights for focal knee resurfacing implant design

Abstract

Purpose: The objective of this work was to characterise the dimensions of knee focal cartilage defects, generate a heat map of cartilage defect locations, and characterise the native articular surface morphology at the most common defect sites to aid in the design of generic focal knee resurfacing implants (FKRIs).

Methods: Healthy femoral cartilage was segmented from knee magnetic resonance imaging scans of 70 patients eligible for FKRI surgery. Cartilage defects were manually reconstructed to create three-dimensional (3D) native cartilage and cartilage defect models. A 3D statistical shape model of the native cartilage was created and 200 artificial models with varying shapes were generated. A heat map showing the frequency of defect occurrence location was created, and the size and aspect ratio of the defects were determined. Local radii of curvature were calculated at five locations that exhibit high defect occurrence for the medial and lateral condyles and in the trochlear groove.

Results: The median defect size and aspect ratio were 2.7 and 1.6, respectively. Cartilage defect location frequency heat maps were successfully generated for the medial and lateral condyles. The median mediolateral and anteroposterior radius of curvature at the medial condyle hotspot were 17.5 mm and 40.8 mm respectively, and 22.6 mm and 33.8 mm at the lateral condyle hotspot.

Conclusions: Characteristics of knee focal cartilage defect properties such as the size, their spatial distribution and local articular cartilage surface morphology offer valuable insights for the design of generic FKRIs.

Level of evidence: Level IV.

Keywords: cartilage defect location heat map; cartilage surface morphology; knee focal cartilage defects; lateral femoral condyle; medial femoral condyle; statistical shape modelling; trochlea.

Figure 1

The workflow for generating the Statistical Shape Model (SSM) and the heat map are schematically depicted. Native cartilage models are obtained after segmentation of healthy cartilage regions and reconstructed cartilage defects sites based on MRI scans obtained from patients eligible for FKRI surgery. Iterative closest point (ICP) and generalised procrustes analysis (GPA) algorithms are used for registration and alignment of native cartilage models. Principal component analysis (PCA) is then employed to construct the SSM. The transformation matrices that were used to align the native cartilage models are applied to align the defect models. The defect location frequency heat map is generated by colour‐coding the mean cartilage shape's surface based on the number of overlapping defects at each point.

Figure 2

A cylinder is fitted to the cartilage surface and its axis is employed to define the planes for measuring the anteroposterior and mediolateral radii of curvature (a). The major and minor axes of cartilage defects are illustrated, along with the locally measured anteroposterior and mediolateral radii of curvature within a circular elliptical region centred on a point of interest on the cartilage surface (b).

Figure 3

Distribution of defect size and aspect ratio for the medial condyle containing 49 defects (left) and lateral condyle containing 14 defects (right).

Figure 4

The (osteo)chondral defect location heat maps for the medial (a) and lateral (b) condyles.

Figure 5

Mediolateral and anteroposterior radii of curvatures for five points selected on the femoral condyles and trochlea of 200 cartilage models generated using the developed Statistical Shape Model (SSM). Points 1–5 are chosen over an area with a high frequency of defect occurrence for medial and lateral condyles, with Point 3 placed on the highest defect occurrence location. The box plots display the distribution of curvature values for each point, presenting the median value by the line inside of the box, the lower and upper quartiles by the top and bottom edges of each box, and the minimum and maximum values by the whiskers that extend above and below each box.

Figure 6

Results of the Statistical Shape Model (SSM) compactness, accuracy, and generalisation ability evaluation. (a) SSM compactness is presented as the cumulative variation of data included in the model based on the number of shape modes. The first 13 and 27 modes of shape variation account for 95% and 98% of the cartilage shape variation within the studied population, respectively. (b) Accuracy of the SSM based on the number of training samples; using the first 35 modes of variation, the accuracy of the SSM reached a value (0.48 mm) below the resolution of the magnetic resonance imaging (MRI) images (0.5 mm) for the first time, and (c) generalisation of the SSM based on the number of training samples. The generalisation ability of the SSM reached a value (0.30 mm) below the resolution of the MRI images (0.5 mm) for the first time when the first 55 modes of shape variation were used.

Figure 7

Distribution of the anteroposterior and mediolateral dimensions of the cartilage in the input data and the 200 generated cartilage samples. AP, anteroposterior, ML, mediolateral.