gms | German Medical Science

Objectives: Effectively planning and achieving appropriate prosthetic fit is a challenge for cementless off-the-shelf stems, owing to the complex anatomy of the intramedullary canal. Therefore, the fit of these stem designs into the twisted and bowed canal is poorly understood.

We aimed to better understand the shape variations that characterises the femoral canal.

Our primary objective was to use Principal Component Analysis (PCA) to identify the main modes of variation.

Methods: This was a retrospective cohort study using 60 preoperative pelvic CT scans of patients who underwent total hip arthroplasty (THA) with cementless stems. 4 patients had bilateral THA. 3D models of the patients' femoral canal were generated from each CT scan, using a 3D image processing software.

The 64 segmented canals used to train the statistical model were made consistent for length and partially defeatured to standardise the morphology. A fixed coordinate system was used to align the models. To investigate variability, a point mapping tool was used to map each training case to a mean shape.

PCA was applied onto the point distribution models, to reduce dimensionality and observe shape variability, by extracting the eigenvectors (directions of variation) and eigenvalues (the extent of change) from the covariance matrix of the data.

The outcome measures were:

1.

The principal modes of variation;

2.

The variance of each mode.

Results and conclusion: The 5 main modes of variation identified were:

1.

Canal size;

2.

Varus/valgus orientation of the canal;

3.

Variation in the morphology of the greater trochanter (GT);

4.

Axial rotation (femoral version);

5.

Distal femoral twist/rotation.

These modes accounted for 85% of the total cumulative variance.

1.

Size accounted for 36% of the total variance (eigenvalue=3.8x10⁵);

2.

Varus/valgus orientation accounted for 19% of the variance (eigenvalue=1.9x10⁵);

3.

Variation in the GT accounted for 15% of the variance (eigenvalue=1.5x10⁵);

4.

Femoral version accounted for 9% of the variance (eigenvalue=8.9x10⁴);

5.

Distal femoral rotation accounted for 6% of the variance (eigenvalue=6.1x10⁴).

The fixation type and femoral component selection in primary THA has previously relied on the Dorr's classification system, however, it does not inform 3D planning of implant position and orientation.

The surgeon performing the tight press-fit fixation has very little intra-operative control of the position and orientation of the prosthesis once mechanical stability is achieved.

The modes of variations identified in this study better describe the variability in the shape of the internal geometry of the femur. This model may be used to plan and deliver the anteversion of a cementless stem more effectively.

Figure 1 [Fig. 1]