Hip osteoarthritis is an important cause of disability and pain in adults, for which hip morphology is a significant risk factor. Hip morphology can be quantified using morphology measurements on pelvic radiographs. However, deviations in patient positioning, such as pelvic rotation, can influence the accuracy of these measurements. As a first step towards correcting pelvic rotation, this research proposes a model that quantifies pelvic rotation on digitally reconstructed radiographs (DRRs) using the area ratio of the obturator foramina.
Segmentations of the pelvis and both femurs of 30 randomly selected participants of the Generation R study were included. The segmentations had been created based on Magnetic Resonance Imaging scans. For each segmentation, DRRs were created with pelvic rotations ranging from -10° to +10° at 1° intervals. Landmarks were used to approximate the areas of the obturator foramina. The right-to-left ratio of the obturator foramen areas was calculated for each radiograph. To find the relationship between this ratio and the angle of pelvic rotation, one function was fitted to all results, and another to only the results for small angles (within ±5°). Both functions were inverted to predict the angle of rotation for a given obturator foramen area ratio. The resulting all-angle and small-angle models were validated on 30 DRRs with weighted random pelvic rotation taken from 15 different participants of the Generation R study, with a 3:1 split between small and large (between ±6° and ±10°) angles of rotation.
An exponential function was found to best describe the relationship between the obturator foramen area ratio and the angle of pelvic rotation. The all-angle model had a median deviation of -0.5° [IQR: -2°, 1°] from the true angles of rotation, and estimated the angle of rotation within 2° in 73% of cases. The small-angle model had a median deviation of 0° [IQR: -2°, 2°] and was accurate within 2° in 67% of cases. When assessing accuracy only for small angles of rotation, the median deviations were -1° [IQR: -2°, 1°] and -0.5° [IQR: -2°, 1°], and the accuracy percentages were 86% and 73%, respectively.
The proposed models showed moderate performance in assessing pelvic rotation. The results of both models are quite similar, though the all-angle model was more significantly influenced by increased variance at larger angles of rotation. While further improvement of the models is necessary, these performances show that quantification of pelvic rotation based on the obturator foramen area ratio on radiographs is possible. As such, this research is a promising first step towards correcting pelvic rotation and improving hip morphology measurements.

Accurate quantification of in vivo knee kinematics is essential for understanding joint health and disorders such as osteoarthritis. Optical motion capture (MoCap) is widely used in gait analysis, but its accuracy is compromised by soft tissue artefacts, limiting bone-level precision. Fluoroscopy provides direct skeletal visualization, yet it faces several limitations, including the high sensitivity of 2D–3D registration to initial pose estimation and the reliance on manual bone segmentation. Moreover, the potential of combining MoCap with fluoroscopy, particularly the role of MoCap in initializing registration for fluoroscopic knee tracking during walking—remains underexplored. The aim of this study is to develop and evaluate a multimodal framework that integrates MoCap with single-plane fluoroscopy to improve automatic 2D–3D bone registration during dynamic gait analysis. Specifically, the framework addresses three challenges: (1) temporal synchronization and spatial calibration of the multimodal system, (2) MoCap-based initial pose estimation, and (3) deep learning–based automated femur segmentation. A synchronized acquisition setup was implemented, simultaneously recording 100 Hz MoCap data and 15 Hz fluoroscopic images during treadmill walking. Sub-frame temporal alignment was verified through a pendulum experiment. Spatial calibration was achieved using a rigid marker box and solved via the Perspective-n-Point algorithm, while a dynamic correction procedure was introduced to update camera pose across trials using rigid markers attached to the X-ray source. For initialization, anatomical reference cubes were constructed from both MoCap markers and segmented MRI data, and rigid registration between these cubes enabled transformation into the fluoroscopy frame. The resulting initial pose was refined using a two-stage optimization algorithm and evaluated against a reference initialization across four trials. MoCap-based initialization produced anatomically plausible in-plane alignment (mean error < 8 mm, < 16°), but consistently showed a systematic depth offset of 30–40 mm. To reduce reliance on manual annotation, a 2D nnU-Net segmentation model was trained on six manually annotated fluoroscopic images. Despite the limited dataset, the model demonstrated good anatomical plausibility, confirming its potential for automated workflows. In conclusion, this thesis establishes a multimodal framework that combines MoCap-based initialization, dynamic camera calibration, and deep learning–based segmentation. The evaluation demonstrates both feasibility and limitations, providing a reproducible basis for quantitative analysis of dynamic knee kinematics and opening new avenues for more automated and personalized assessment of joint disorders.

Introduction
Corrective forearm osteotomies are complex procedures aimed at restoring anatomical alignment and function in patients with deformities caused by trauma or congenital conditions. While computer-assisted planning improves precision, existing workflows are labour-intensive and limit exploration of optimal solutions. This thesis presents the 3D-SurgiGen Add-on, a semi-automated, open-source planning tool designed to enhance efficiency, consistency, and clinical applicability.

Methods
The 3D-SurgiGen Add-on was developed as a modular plugin for Blender using open-source technologies and a multi-objective optimization framework. Development involved defining requirements, analyzing and restructuring an existing codebase, designing a new architecture, incremental integration, and iterative refinement based on testing and clinical feedback. The tool enables import of 3D forearm models, annotation of anatomical landmarks, and planning based on the mirrored contralateral bone. It supports both opening and closing osteotomies via an innovative optimisation strategy that enables multiple osteotomies while providing explicit control over distal alignment, critical for joint congruency. A key innovation is the user-in-the-loop workflow, which enables clinicians to interactively explore osteotomy alignment options and select a preferred solution from multiple optimised alternatives. The performance and usability of the tool were subsequently evaluated to assess its potential as a clinically relevant semi-automated planning solution.

Results
The tool was prospectively evaluated in five clinical cases involving deformities of both the radius and ulna. Semi-automated planning yielded a comparable number of clinically acceptable plans to manual planning, with similar failure rates and no reduction in planning time (median difference: 7 minutes, range: 4–17). Limitations include the small sample size and the exclusive evaluation of single-shaft osteotomies. Usability feedback further highlighted current limitations in visualisation, parameter interaction, and overall robustness of the Add-on.

Conclusion
The 3D-SurgiGen Add-on demonstrates the feasibility of semi-automated, optimisation-based forearm osteotomy planning. Future work should expand validation, refine clinical objectives, and improve usability. With continued refinement, the 3D-SurgiGen Add-on could serve as an efficient complementary tool for clinical decision-making, providing insight into surgical options and enabling the customisation of preoperative corrective osteotomy plans efficiently.