This thesis explores the impact of 3D femoral rotational effects on 2D X-ray femoral shapes using Statistical Shape Modeling (SSM). The study is driven by the clinical relevance of accurately diagnosing hip disorders from 2D radiographs, which are inherently influenced by the fem
...
This thesis explores the impact of 3D femoral rotational effects on 2D X-ray femoral shapes using Statistical Shape Modeling (SSM). The study is driven by the clinical relevance of accurately diagnosing hip disorders from 2D radiographs, which are inherently influenced by the femur's position and rotation. The primary objective is to understand how these rotations affect the femur's apparent shape on radiographs and quantify the influence of rotation on shape variations observed in 2D SSMs.
The research was structured around two main experiments. The first experiment involves the creation of a 3D SSM of the whole femur using CT scans. This model was then used to explore how projections of the proximal femur change with different rotations. The second experiment focuses on the proximal femur and constructs two 2D SSMs. These were derived from Digitally Reconstructed Radiographs (DRRs) of the proximal femur, which are obtained by rotating the CTs (±30° in 5° increments) around the shaft axis of the femur. One 2D SSM is based solely on DRR data, while the other combines DRR data with the WorldCOACH dataset of hip X-rays. This approach aims to provide an understanding of how rotation influences the shape modes of the femur as described by the 2D SSMs. For the first five shape modes the resulting b-values presented in a boxplot. Lastly, Welch's Analysis of Variance (ANOVA) is performed on these b-values for both 2D SSMs. The ANOVA analysis is performed on the following independent values: the femur, which represents the natural bone shape of that specific femur, and the angle, which represents the angle of rotation at which the DRR was created.
Findings show that sensitivity to angle changes varies across SSM modes, with shape changes due to external rotation being more evident than those caused by internal rotation. The combined results from the plots and Welch's ANOVA analyses of the two models (DRR-only and combined WorldCOACH and DRR) indicate that modes 1 to 4 are more heavily influenced by rotation, while mode 5 is more influenced by natural shape differences. This suggests that higher modes, which describe more specific and less variable shapes, are less influenced by rotation and more by intrinsic anatomical variations.