This study investigated the feasibility of deep learning-based segmentation in intra-abdominal kidney ultrasound registration, to enable image-guided robotic-assisted partial nephrectomy (RAPN). Two state-of-the-art models, DeepLabV3+ and SAMUS, were trained and evaluated using a novel intra-abdominal kidney ultrasound (IAKUS) dataset of 2,265 images from 15 RAPN patients. Moreover, a transfer-learning approach was adopted using the publicly available open kidney ultrasound (OKUS) dataset for pre-training. Results showed that SAMUS consistently outperformed DeepLabV3+ across all metrics, achieving an average Dice score of 88.0±2.0% and Hausdorff distance of 13.7 ± 3.8 mm, consistent with literature. SAMUS was pre-trained on ∼ 30k ultrasound images which enabled a zero-shot test, outperforming trained DeepLabV3+ configurations. Furthermore, no measurable difference was seen between OKUS and IAKUS training. Both findings suggest ultrasound specific features may be more important than organ specific features for training, and data diversity may be more important than strict anatomical similarity. The SAMUS model obtained a registration accuracy of 4.3 ± 2.8 mm and inference time of 4.35 fps, in line with literature reported clinical feasibility. The target registration was even improved by an average of 2.6 ± 4.2 mm in 11/13 patients compared with manual-based registration. This proves that deep learning-based registration is not only feasible in a clinical setting but exceeds manual-based registration.

Lower back pain (LBP) is the leading global cause of disability, with 619 million cases in 2020 projected to rise to 843 million by 2050. The societal and economic burden is substantial; in the UK, the National Health Service spends over £5 billion yearly on back-related care, while in the USA, costs from 2012 to 2014 for patients with a spine condition were $315 billion USD. Current treatments are limited, surgery is invasive and costly, while medication such as opioids offers only short-term relief with uncertain long-term benefit and considerable side effects. Spinal traction is widely used as a non-invasive treatment. It may alleviate pain through spinal decompression and improved
disc health, but existing devices are often expensive, uncomfortable, or require supervised use in a clinic.
This thesis describes the design, development, and evaluation of a novel traction device for the treatment of LBP, which is affordable, user-friendly, and suitable for unsupervised usage in physiotherapy practices or at home. A patient study of an early seated prototype (Prototype 2) with 20 patients revealed limitations in fixation and comfort: four patients could not be tested due to slipping through the belt fixation, and six of the remaining patients were unable to complete the 10-minute treatment. On average, patients reported only a slight reduction in pain (∆V AS of 0.47 ± 2.60), and a significant negative correlation was found between treatment-related discomfort and pain reduction (r = −0.57, p = 0.022).
Furthermore, market research at YES!Delft indicated that Prototype 2 would be too costly for the intended users. To optimize therapy, a literature review was conducted to determine how the therapist controlled parameters, specifically force, body position, and duration, to assess their potential influence on the treatment effect. The review suggested that traction is most effective at forces of < 20% body weight (increasing up to 40% if necessary), in prone or supine 90/90 positions, and with intermittent traction cycles lasting 10-20 minutes. Additional factors influencing outcomes included patient age, duration of symptoms, presence of sciatica, and timing of follow-up.
Based on these findings, a third prototype was developed and evaluated in a patient study comprising six treatments over a six-week period. The control group (n = 4) received manual traction in combination with mobilization and detonizing (which is standard physiotherapy treatment) and in the study group (n = 6) manual traction was replaced with treatment by Prototype 3. Both groups reported reductions in pain (VAS) and disability (ODI), with statistically significant improvement observed only in the study group (p = 0.0216). No significant differences between groups were detected, though effect sizes indicated individual improvement. These results demonstrate the viability of a patient-operated traction device and suggest therapeutic potential comparable to that of current physiotherapy treatments and reported literature. While further research with larger study groups and refined design parameters is needed, the device shows promise as a cost-effective solution for use in physiotherapy practices or home settings.

Complications following endovascular aneurysm repair (EVAR), such as Type IA endoleaks (TIAEL) and limb graft occlusions (LGO), remain a clinical concern. However, existing research aimed at resolving these complications is limited, mainly due to difficulties in visualizing and quantifying the deformation of stent fabric in standard postoperative images. This thesis presents a novel in vitro framework to visualize and quantify stent graft fabric folding in abdominal aortic aneurysm (AAA) phantoms. Silicone phantoms with representative AAA
anatomy were fabricated and validated using a combination of compression tests and a clinical evaluation of elasticity by experienced clinicians. Each phantom was implanted with an
Endurant II stent grafts under varying oversizing conditions. Micro-CT imaging was used to visualize the endograft fabric within the phantoms. A custom convolutional neural network-
based segmentation pipeline was developed to quantify fold severity as the proportion between the total phantom area and the stented lumen. Preliminary comparison with finite element simulations (PrediSurge) demonstrated the feasibility of the method. Therefore, this thesis establishes a validated methodology for fabric fold visualization and quantification, supporting future studies on the relationship between stent graft folding and post-EVAR complications.

This study compares the propulsion performance between hard- and soft-magnetic microrobots under rotating magnetic fields. Results show that hard-magnetic microrobots achieved step-out frequencies and maximum propulsion speeds 4.5 times higher than soft-magnetic microrobots. Below saturation magnetization, soft-magnetic microrobots demonstrated similar performance irrespective of magnetic susceptibility, highlighting that torque generation in these materials is purely geometry-dependent. Employing a tapered ribbon design increased propulsion speed by a factor of 3.5 compared to regular helical designs. These results provide a quantitative basis for selecting materials and designs, enabling designers to weigh the propulsion benefits of hard magnets against the biocompatibility of soft-magnetic microrobots.

Cervical cancer remains a major global health challenge and is one of the most frequently diagnosed cancers in women worldwide. Brachytherapy plays a central role in its treatment, particularly in locally advanced stages. Brachytherapy enables the precise delivery of high radiation doses directly to the tumour while limiting exposure to surrounding healthy tissues, making it a widely adopted and effective modality. To support the development and evaluation of brachytherapy procedures, anatomically realistic and mechanically representative pelvic phantoms are essential. These phantoms provide a safe and repeatable environment for testing needle insertion, assessing imaging compatibility, and exploring strategies for dose distribution, all without risk to patients. By replicating pelvic anatomy and enabling validation through CT and MRI guidance, such phantoms contribute to improved clinical accuracy and innovation in treatment planning.

This thesis presents the design and validation of a modular, anatomically accurate female pelvic phantom intended for research in gynaecological brachytherapy. The phantom was developed to support needle insertion and compatibility with computed tomography (CT), magnetic resonance imaging (MRI), and electromagnetic tracking (EMT).
The phantom includes soft-tissue components representing the vaginal canal, cervix, and surrounding support structures, cast from a silicone–cornstarch mixture and suspended within a rigid 3D-printed pelvic frame. An interchangeable tumour insert and surrounding tissue volume were incorporated to provide imaging contrast and enable clinical variation. Material testing guided the selection of suitable silicone mixtures for different tissue regions. To validate the usability of the phantom, ten interstitial needles were inserted and recorded using both CT and EMT. Custom registration using rigid and affine transformations was implemented to align needle trajectories from CT and EMT, enabling quantitative comparison of tip positions and insertion angles. This resulted in a median deviation in tip position and insertion angle of 11.06 mm and 8.04°, respectively. MRI scans further confirmed that the phantom produces sufficient contrast between different elements in the phantom.

In conclusion, the phantom was compatible with clinical tools and imaging modalities such as MRI, and enabled comparison of needle trajectories across CT and EMT. Although the validation was limited in scope, the results confirm that the phantom design is usable in the brachytherapy context. Opportunities for phantom improvement are discussed. With further development, the phantom can serve as a valuable platform for both validation and training in gynaecological brachytherapy.

Augmented Reality (AR) offers significant potential to enhance surgical procedures by overlaying critical digital information directly onto the surgeon's view, improving spatial awareness and precision. Accurate real-time tracking of surgical instruments is fundamental to the success of AR navigation systems. However, implementing established tracking modalities, such as infrared (IR) tracking using retro-reflective markers, on newer Head-Mounted Displays (HMDs) like the Magic Leap 2 presents unique challenges due to its specific sensor configuration, notably the lack of accessible IR intensity data from its Time-of-Flight (ToF) sensor and the presence of IR-cut filters on its World Cameras.

This thesis addresses these challenges through the development, implementation, and evaluation of ISITLeap, a novel system designed specifically for IR instrument tracking on the Magic Leap 2 platform. The system employs a stereo vision approach, utilizing the Magic Leap 2's built-in World Cameras configured with short exposure times, synergistically combined with a custom-designed external 760 nm IR illumination module to detect standard passive retro-reflective markers. A comprehensive software pipeline was developed, incorporating image undistortion adapted for the Magic Leap 2's equidistant projection model, marker detection involving binarization, connected components analysis, and area filtering, 3D marker reconstruction using stereo triangulation with empirically derived coordinate system corrections to reconcile Magic Leap 2 SDK data with geometric requirements, and finally, 6-DoF tool pose estimation via point-based registration. Crucially, the entire software pipeline, from image acquisition to pose estimation, operates directly on the Magic Leap 2 headset, requiring no external computing hardware.

Experimental validation successfully demonstrated the feasibility of the ISITLeap system for real-time tracking. Static evaluations characterized the system's performance, identifying an optimal working range of approximately 25-50 cm. Dynamic tracking experiments, evaluated against an NDI optical tracker using an ArUco marker as a common reference frame, yielded an overall Mean Absolute Error (MAE) of approximately 7 mm. Crucially, error analysis revealed that the measured accuracy was significantly influenced by limitations inherent in the ArUco marker-based evaluation methodology, likely masking the intrinsic performance capabilities of the ISITLeap core tracking system.

This work presents a complete, functional hardware and software solution tailored to overcome the specific sensor constraints of the Magic Leap 2, enabling IR marker tracking on this platform. It provides a crucial performance benchmark under the defined evaluation conditions and demonstrates a viable technical pathway for developing precise AR surgical navigation applications for the Magic Leap 2 headset. While the achieved referenced accuracy highlights the need for improved evaluation techniques and further system refinement (e.g., via dynamic filtering and optimized hardware), ISITLeap establishes a foundational system for future advancements in Magic Leap 2-based AR surgical guidance.

Surgical workflow analysis is crucial in optimising procedural efficiency, resource utilisation, and patient safety in catheterisation laboratories. Traditional manual work- flow analysis methods are labour-intensive and prone to inconsistencies, prompting the need for automated solutions that leverage machine learning and computer vision. This thesis presents an explainable two-stage model for surgical workflow analysis using ceiling-mounted cameras. The proposed approach integrates a YOLOv8 object detection model with a Gaussian Mixture Model - Hidden Markov Model (GMM-HMM) framework. The first stage detects key objects that serve as input to the second stage. The GMM-HMM component then infers surgical workflow phases by modelling spatial and temporal dynamics, enabling real-time phase classification. The model is validated on two datasets from different hospitals, achieving a classification accuracy of 95.2% for the RdGG dataset and 95.4% for the HH Tampere dataset, ensuring its generalisability across diverse clinical environments. Experimental results demonstrate that the model achieves high accuracy in detecting workflow phases, with an emphasis on explain- ability and robustness. The combination of YOLOv8’s efficient object detection with GMM-HMM’s structured temporal inference ensures that phase transitions are identified with minimal error. The model’s real-time feasibility and generalisation across hospitals highlight its potential for clinical implementation. This research advances automated surgical workflow analysis by addressing the dual challenges of interpretability and adaptability. Future work includes enhancing the model’s robustness to occlusions, integrating additional modalities such as audio data, and exploring its application in other surgical environments.

Background

The surgical management of Hepato-Pancreato-Biliary (HPB) cancer poses significant challenges, primarily due to the complexity of patients. However, the role of stereopsis (depth perception) in visualizing three-dimensional (3D) anatomical models remains relatively underexplored. Integrating stereoscopic technologies with 3D anatomical modelling holds promise for enhancing surgical planning and navigation, thereby addressing the inherent complexities of HPB surgeries.

Aim
This study aims to evaluate the effect of stereopsis on the preoperative surgical planning of colorectal liver metastases surgery.

Methods
A retrospective study was conducted with participants from the Department of Surgery, Leiden University Medical Centre (LUMC), to investigate the occurrence and severity of symptoms resulting from the use of a stereoscopic display. Subsequently, liver and transplant surgeons from the same department participated in another retrospective study comparing surgical plans for colorectal liver metastases performed stereoscopically with those performed monoscopically, within the same surgeon.

Results
14 out of 18 participants experienced (slight) symptoms from the use of a stereoscopic display, yet no one discontinued the study due to symptoms. In the subsequent study on the effect of stereopsis on preoperative surgical planning for colorectal liver metastases, 13 liver and transplant surgeons participated. Relative to a gold standard, there appears to be no significant difference between surgical plans executed monoscopically or stereoscopically. There is also no significant difference in the time taken to create these surgical plans (p=0.401). Despite the absence of significant difference between the plans, surgeons do express a (strong) preference for stereopsis in locating the tumor (61%), determining the surgical plan (61%), and assessing vascular involvement (69%).

Conclusion
It is evident that surgeons have a preference for stereopsis in visualizing 3D models, although our study found no discernible differences in outcomes between monoscopic and stereoscopic preoperative planning for colorectal liver metastases surgery. Future research is recommended to compare surgical plannings based on conventional two-dimensional imaging alone with conventional two-dimensional imaging supplemented by additional stereoscopic 3D models. This comparative analysis could offer further insights into the potential advantages of integrating stereoscopic technology into preoperative planning practices.

Tissue extraction in Minimally Invasive Surgery is often performed using a suction catheter, valued for its capability to continuously remove tissue from the body. However, suction catheters face limitations in further miniaturization and reliability due to energy losses and clogging. Kortman et al. (Kortman, 2023) proposed a solution: a soft tissue transporter that does not affect the surrounding tissue and enables miniaturization without sacrificing performance. The mechanism involves a soft silicone-based conveying toroid that transports tissue from one end of the tube to the other end of the tube. However, a shortcoming of the device is its inability to function as a continuous tissue transporter. Continuous transporters are often preferred as they shorten surgery times, reduce the amount of anesthesia needed and lower the overall operational costs compared to discrete transport mechanisms. Therefore, the goal of this study is to develop a continuous toroid tissue transporter, compatible with Minimal Invasive Surgery, for which the transport efficiency is independent of the tissue elasticity, shape and size. The final concept featured a toroid that was continuously everted while remaining stationary, enabling it to function as a conveyor-like mechanism. By integrating an internal skeleton with magnets at its edges, the toroid could be actuated externally using a gearbox. Additionally, an uniaxial mesh was incorporated into the toroid to allow flexibility in its width, allowing the shape grip to transport various objects, while preventing stretch in the length of the toroid, thereby improving eversion efficiency. A proof-of-principle prototype was developed to validate the working principle. Results demonstrated that the toroid efficiency was not limited by various tissue sizes, shapes or elasticities. Therefore, the study goal has predominantly been achieved.

Improvement of actuation efficiency can be achieved by further alignment between the actuation system and the toroid, as currently slip is causing a reduction in actuation efficiency. In conclusion, the Continuous Toroid Transporter demonstrates promise for surgical applications and other fields requiring precise object transportation and is recommended to be further developed.

The application of Augmented Reality (AR) technology in neurosurgery is becoming increas- ingly widespread, especially in External Ventricular Drainage (EVD) procedures, where achiev- ing efficient and precise registration remains a key challenge. Traditional marker-based registra- tion methods are often complex and time-consuming, making them unsuitable for emergency surgical situations. This paper presents a markerless automatic registration method based on HoloLens 2, using the Dlib library to automatically detect facial landmarks, and perform- ing point cloud registration based on these features. We designed four sets of experiments: first, to verify the feasibility of the triangle-based registration algorithm; second, to assess the accuracy and stability of Dlib’s feature point extraction on HoloLens; third, to validate the registration accuracy, including the impact of skin displacement; and finally, to evaluate the accuracy of the insertion path in a simulated EVD procedure. The experimental results show that this method simplifies the registration process and demonstrates advantages in terms of registration accuracy and speed, effectively meeting the real-time and precision requirements of emergency surgeries. This study provides an efficient and reliable AR navigation solution for EVD procedures, with promising prospects for further research.

Deep-learning-based object detectors, while offering exceptional performance, are data-dependent and can suffer from generalization issues. In this thesis, we investigated deep neural networks for detecting people and medical instruments in the vision-based workflow analysis system inside Catheterization Laboratories (Cath Labs). The central problem explored in this thesis is the fact that the performance of the detector can degrade drastically if it is trained and tested on data from different Cath Labs.

Our research aimed to investigate the underlying causes of this specific performance degradation and find solutions to mitigate this issue. We employed the YOLOv8 object detector and created datasets from clinical procedures recorded at Reinier de Graaf Hospital (RdGG) and Philips Best Campus, supplemented with publicly accessible images. An aggregated version of object detection metrics was created for multi-camera system evaluation. Through a series of experiments complemented by data visualization, we discovered that the performance degradation primarily stems from data distribution shifts in the feature space. Notably, the object detector trained on non-sensitive online images can generalize to unseen Cath Labs, outperforming the model trained on a procedure recording from a different Cath Lab. The detector trained on the online images achieved an mAP@0.5 of 0.517 on the RdGG dataset. Furthermore, by switching to the most suitable camera for each object, the multi-camera system can further improve detection performance significantly. An aggregated 1-camera mAP@0.5 of 0.679 is achieved for single-object classes on the RdGG dataset.

Introduction: High cardiovascular mortality primarily limits survival following kidney transplantation (KTx). The increased risk is intrinsically connected to cardiac structural and functional abnormalities. LV volumetric imaging measurements are used to quantify this, such as left ventricular ejection fraction (LVEF) and left ventricular end-diastolic volume index (LVEDVi). However, there is an increasing interest in the assessment of myocardial strain. It provides direct information about myocardial function by measuring the deformation of myocardial fibres during the cardiac cycle. An impaired LV diastolic strain is associated with adverse outcomes in end-stage kidney disease (ESKD) patients. Cardiac magnetic resonance imaging (MRI) has emerged as the golden standard for evaluating cardiac structure and function. However, studies investigating the prognostic value of myocardial strain are mainly based on echocardiography, and diastolic strain still needs to be examined in KTx recipients.
Aim: It was hypothesized that myocardial strain has the potential to detect early diastolic dysfunction more effectively when compared to LV volumetric measurements. The primary objective of this thesis was to determine whether it is possible to predict all-cause mortality of KTx recipients on cardiac MRI two weeks post-KTx by diastolic strain.
Methods: Data from patients included in the Amsterdam Leiden GROningen (ALEGRO) trial were used. Medis Suite MR was used to calculate LV volumetric outcomes and identify deformation of the LV during the cardiac cycle. A MATLAB script was developed to obtain diastolic strain rate outcomes from these data. Kaplan-Meier curves were used for survival analysis and Cox regression analysis was performed to identify independent predictors of survival.
Results: 67 patients were included with cardiac MRI at baseline, two weeks post-KTx. The mortality was 27\% (n=18) over a median follow-up of 10 years. Significant differences between survival distributions were found based on longitudinal peak early diastolic strain rate (PEDSR) at baseline. Using LV volumetric outcomes, including LVEF, LVEDVi and left ventricular end-systolic volume index (LVESVi), this was achievable starting from a follow-up of one year. Multivariable analysis showed that longitudinal PEDSR was the strongest independent predictor of mortality at baseline.
Conclusion: It is demonstrated that myocardial strain has the potential to detect early diastolic dysfunction more effectively when compared to LV volumetric outcomes. Longitudinal PEDSR measured two weeks post-KTx by cardiac MRI independently predicts all-cause mortality and provides incremental prognostic information beyond clinical parameters. Further research in cardiovascular therapeutics is an essential next step in improving KTx recipients' prognosis.

Background
An upper limb amputee currently has two choices for a body-powered prosthesis, a Voluntary Opening (VO) or Voluntary Closing (VC) hand prosthesis. Which type is best, depends on the task and the individual. Currently, there are no good options for VO and VC in one design.
Objectives
To design a proof-of-principle of a hand prosthesis and to validate the feasibility of the prosthesis. The hand prosthesis must have two different modes, the VO mode and the VC mode. Changing between these modes can be done without using the other hand..
Methods
First, a list of requirements and wishes was made, where cable forces are important because too much cable force can cause discomfort, fatigue or make the prosthesis difficult to control.
A new design was created and a prototype was built for user testing. SHAP was used as the outcome measure.
Results
The prototype allowed the user to change between VO and VC and SHAP showed promising Linear Index of Functionality (LIF) values. The calculated cable forces were mostly within the requirements. However, because the prototype of this design was produced with an FDM-printer, not all the tests could be done and the calculated forces could not be verified.
Conclusions
A prototype of a hybrid hand prosthesis with VO and VC modes showed promise, with positive test results. However, limitations in the prototype's construction hindered some tests. The calculated cable forces mostly met requirements, except for a slightly high pinch grip force. Further development is needed, and there are material and design recommendations. This concept demonstrates feasibility but requires more work to become a functional hand prosthesis.

Purpose: 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) is used in the diagnostic process and management of patients with metastatic colorectal cancer (mCRC). Also, 18F-FDG PET radiomic features have been found to hold prognostic value for clinical outcome in mCRC. However, no prognostic model has yet been developed to predict clinical outcome in mCRC using 18F-FDG PET images. Computer-aided pattern recognition can be helpful in this process but needs to be validated. The aim of this work was to develop and evaluate a medical-based explainable artificial intelligence (XAI) framework for discriminating between dichotomous progression free survival (PFS) in patients with mCRC undergoing anti-epidermal growth factor receptor (anti-EGFR) monoclonal antibody (mAb) treatment using pre-treatment 18F-FDG PET images.
Methods: We conducted an analysis of 18F-FDG PET images, expressed in standardized uptake values (SUV), obtained from 80 patients with mCRC who were eligible for third-line treatment with an anti-EGFR mAb as part of the IMPACT study. A coronal 2.5D Convolutional Neural Network (CNN) was built to capture features of the 18F-FDG PET images specific for the two patient groups and a medical-based XAI framework was developed to extract the 18F-FDG PET features used by the CNN. The images were randomly divided into a training and a validation set (10-fold cross-validation). Performance of the CNN was evaluated based on the average area under the curve (AUC), accuracy, sensitivity and specificity from the cross-validation. A statistical analysis was performed to assess the predictive value of the 18F-FDG PET features extracted by the XAI framework.
Results: The coronal 2.5D-CNN was able to discriminate between dichotomous PFS (median PFS: 152 days) in patients with mCRC undergoing anti-EGFR mAb treatment using pre-treatment 18F-FDG PET images, with an average AUC of 0.95  ± 0.11 (SD), accuracy of 94% ± 12, sensitivity of 91% ± 21 and specificity of 94% ± 21 %. The XAI framework showed that especially low 18F-FDG PET uptake volume features hold significant differences between the two patient groups.
Conclusion: The coronal 2.5D-CNN showed good performance to predict dichotomous PFS from pre-treatment 18F-FDG PET images in patients with mCRC undergoing anti-EGFR mAb treatment. Low 18F-FDG PET uptake volume features seem to have potential as IB in this patient cohort, but further validation is required.

Stiffness perception in laparoscopic surgery has an important role in achieving appropriate tissue manipulation. The need for haptic feedback to perceive tissue stiffness and the reduce in haptic feedback when using laparoscopic instruments suggest an inability to sufficiently experience stiffness perception during laparoscopic surgeries. Better understanding of the perception of tissue stiffness could contribute to the development of better laparoscopic instruments. A stiffness ranking task has been performed with 16 participants, to study the influence of force transmission ratio, contact area and grasping strategy on the haptic perception of variable tissue stiffness. Using a newly developed set of laparoscopic grasper handles varying in the chosen independent variables, being the force transmission ratio and contact area, the participants performed multiple grasps with each grasper handle on a randomized order of tissue samples with three different stiffness levels and ranked them according to the perceived level of stiffness. Force measurements were done during the task and a survey was conducted afterwards. The results indicate that force transmission ratio and contact area do not influence the haptic perception of variable tissue stiffness. In terms of grasping strategy, a lower force exertion is shown to be desirable for perceiving the variable tissue stiffness, while the amount of grasps per tissue sample have not shown to influence the perception. Future studies with larger sample sizes might indicate a difference between high and low force transmission ratios. It is concluded that future designs of laparoscopic grasper handles might not be restricted in terms of force transmission ratio and contact area requirements when a greater haptic perception is aimed at. More haptic perception of variable tissue stiffness is gained in the surgical domain of applying the appropriate grasping strategies.

This thesis presents a method for personnel activities observation, i.e., 3D human pose estimation and tracking, in a Catheterization Laboratory(Cath Lab). We mount five cameras from different angles in the Cath Lab, where surgeons and assistants are in similar clothes while doing surgery. Accurate 3D human pose estimation is the cornerstone of our method. Most previous 3D pose estimation methods train their models directly on a 3D pose dataset. However, these methods are not suitable for our task: 1) We do not have enough 3D pose data for training because of privacy issues and specificity 2) The model needs to be retrained in different operating rooms or the camera calibration changed. To solve these problems, we decompose the 3D human pose estimation task into two stages, avoiding the need for large amounts of 3D pose data and retraining. In the first stage, we apply YOLOX and HRNet for 2D human detection and 2D pose estimation. Simultaneously, the 2D object tracking network Bytetrack tracks person identities based on detection results. Then we use a matching algorithm to match the corresponding 2D poses from multiple views and reconstruct 3D poses. Given 3D poses and tracking identities, we at last introduce a hybrid method tracking algorithm. By feeding 2D tracking results into the matching and tracking algorithm, we increase the accuracy of the result in a scene where people are wearing similar clothes. We fine-tune and test our method with an operating room dataset. Finally, we validate the method on data from the Cath Lab.

Research on the possibilities to use Artificial Intelligence (AI) at the Operating Room (OR) is rising, although there is no implementation yet. AI-assisted monitoring could provide support to anesthesia assistants in their task of patient monitoring, by making predictions on the future occurrence of deviating events in the vital parameter values of the patient. As AI-assisted monitoring would come close the profession of anesthesia assistants, this explorative, qualitative research investigates the attitude of anesthesia assistants towards the use of AI in their task of patient monitoring. Furthermore, it investigates the motivational factors that are important for their motivation in their monitoring task as well as the expected impact of AI on their motivation. This design-based research subsequently describes how these results are incorporated into the development of an educational module about AI and motivation for anesthesia assistants. Gaining as well as giving insight in the motivation of anesthesia assistants is important, as their motivation is of impact on their professional behaviour while their decisions and actions are directly related to the safety of the patient.

Through literature research and semi-structured interviews, it was found that among anesthesia assistants there is little knowledge on the possibilities of AI at the OR, but that their attitude towards AI-assisted monitoring is positive as long as only suggestions are given. Only two out of eight anesthesia assistants are cautiously positive towards the possibility of AI to automate the monitoring process. Furthermore, it was found that the motivational factors of autonomy, relatedness, competence, self-efficacy, and self-actualisation are all important to very important in the motivation of anesthesia assistants. Overall, none of these factors stands out in being most or least important. The expected impact of AI on these motivational factors among anesthesia assistants varies from negative to positive and is highly dependent on the form of AI-assisted monitoring. As various forms of interrelatedness between all these motivational factors are identified by the anesthesia assistants, specific to their working practice and environment, the impact on one motivational factor might indirectly also impact others.

These results on attitude and motivation are incorporated in the design of an educational module for anesthesia assistants. The module consists of two parts. One part regarding AI-assisted monitoring and one part regarding the motivational factors that are important for their motivation and the impact of AI on motivation. Design requirements by the anesthesia assistants, as well as an expert validation by an anesthesia educator, are used to finalise the design of the module. This module can help anesthesia assistants to develop awareness and understanding of how AI-assisted monitoring, in the form of giving suggestions, can support them in their patient monitoring task. Furthermore, it helps them to develop awareness and understanding on their own motivation and the impact that the use of AI-assisted monitoring may have on their motivation, guiding them towards becoming deliberate professionals.

Pencil beam scanning is a technique used in the proton therapy. This technique makes use of a thin beam which can irradiate each spot at millimetre precision. It is a very promising technique, especially for more radiation sensitive areas, e.g. lung carcinogenesis, as it can spare more healthy surrounding tissue compared to conventional radiation therapy due to proton’s physical property of the Bragg peak: this is the explosively release of the particles’ energy at a certain point in the beam path; a minimal energy loss is obtained before this peak and no energy is deposited after this Bragg peak. In the PBS, interplay is a concern. The interplay effect is defined as the interference of the moving beam and breathing induced tumour motion. Hence an inhomogeneous dose distribution occurs which differs from the planned dose. In the interplay effect different parameters are involved, which can be divided into beam and patient parameters.

The HollandPTC developed a Quality Assurance tool for assessing the interplay effect in lung treatment plans in which a first order sinusoidal function is utilized. However, a sine function is not reflecting the daily variation in target motion and also beam delivery fluctuates over time. By implementing variations into our QA tool, a more realistic interplay assessment could be obtained. A previously done literature search provided us insufficient information regarding beam and patient parameters describing variations over time, hence we did our own research. In this thesis, an investigation was done to analyse the variation in beam and patient specific parameters.

For the machine parameters, the variation in Energy Layer Switching Time (ELST) and Beam On Time (BOT) was determined for the VARIAN ProBeam 4.0. Three clinical lung plans which contained in total nine beams, were irradiated on an array of ionization chambers multiple times over multiple days. The output files were analysed regarding the ELST and BOT. For each beam 15 measurements were done, a total of 135 measurements were executed over a period of two weeks. We found variations in ELST between energy layers, but also within an energy layer variations in ELST was observed. The within day and day-to-day variations were comparable. Regarding the BOT, this variable strongly depends on the planned dose. Looking at each individual beam: variations between energies are found; no variations within an energy layer was found. By dividing the measured dose of all output files within an energy layer by the corresponding BOT, the dose rate for each energy was obtained. The dose rate was low at low energies and gradually increased towards the higher energies.

Patient parameters were also investigated regarding the tumour motion pattern: the variation in amplitude, period and degree of asymmetry was investigated. In three lung patients, the breathing signal was recorded for 1.5 minutes utilizing the Anzai belt. Five measurements were done: two measurements for two patients and one measurement for one patient. The obtained signals were analysed in terms of amplitude, period and degree of asymmetry for each breathing cycle. Each cycle was fitted into the Lujan’s model, hence a value for each of the three variables were observed. For our group the amplitude variated between 6 and 12mm and the period variated between 2.5 and 4.5 seconds. The degree of asymmetry was most likely to be 1. We found significant interpatient variation, mainly for the amplitude and period.

In this analysis, we found that the ELST mainly depends on the step size of the degrader and slit movement. Unfortunately, no hard conclusion regarding the variation within an energy layer could be stated, probably the data transfer systems of the beam line are involved in this variation. Developing a prediction tool for assessing the ELST seems appealing, however more data in different energy ranges should be obtained. Regarding the BOT: this time was more or less dependent on the planned dose. Variations measured for this parameter were caused by inaccuracies of our measurement tool itself.

Regarding the patient parameters, significant differences are found between patients’ respiratory signal e.g. the tumour motion strongly depends on tumour characteristics. Hence, we advise to analyse the tumour motion pattern patient specifically.

In this thesis report the first insights towards the fluctuations in the relevant parameters were determined. The retrieved information can be used to optimize the interplay QA tool, to get a more realistic interplay assessment in the plans.

Augmented reality (AR) promises to be a valuable tool for anatomical education since any given anatomical site can be studied dynamically in three dimensions (3D). Yet, it has not been used with case-based learning in medical education despite the latter's demonstrated benefits.
This research, thus, investigates the potential benefit of 3D holographic animated surgeries on functional anatomical knowledge acquisition. We developed four educational AR applications comparing two cases with and without the presence of a surgical animation. Based on this, a randomized crossover trial was conducted among second year medicine students. All subjects (n = 10) underwent a spatial visualization assessment, followed by a learning session and an anatomical knowledge test for both patient cases. The presence of a surgical animation did not lead to a significant difference in anatomical test scores. In addition, no correlation (τ = -0.092, p = 0.78) was detected between the effect of surgical animations and spatial visualization abilities. However, these results are based on non-parametric statistics, as normality could not be assumed. Further research with samples representative for the population, conducting both qualitative and quantitative analysis using verified outcome measures and parametric statistics is required to draw accurate conclusions.

The large intestines, or colon, is the last part of the digestive tract before the rectum. Colorectal cancer is one of the main reasons that patients need a colon resection. A colon resection is a surgery in which the malignant part of the colon is removed. After removal of the tumor, the two ends of the colon need to be reattached, also called anastomosed. This is usually done using sutures or staples. One of the most frequently occurring and major complications is that the anastomosis does not heal correctly. This can cause an anastomotic leakage, which means that fecal matter and fluids with bacteria can flow into the otherwise sterile abdominal cavity. This is a serious complication that can even result in death. Therefore, a solution can be to insert a stent to cover the anastomosis from the inside, preventing feces and fluids from entering the abdominal cavity. Such a stent was developed at the department of Surgery of the Erasmus MC.

This thesis focuses on the redesign of this stent to cover an anastomosis. Two main problems with the stent were stent migration due to insufficient friction between the stent and the colonic wall, and intestinal blockage due to the presence of the stent.

Preliminary research and a literature study showed that the addition of barbs to the stent could increase friction with the colonic wall. Optimal barb dimensions were researched on fresh colon specimens in an experimental setup. Ideal barb parameters were a height of 1.5 mm, diameter of 1.0 mm, an angle of 30 degrees, with a distribution of 3 barbs per cm2. In addition, the ideal stent diameter was explored in an experiment with fresh colon specimens. The stent should expand the colon with 50-60%. To safely insert and extract a stent with barbs in the colon, a patent search was executed to explore the possibilities of barb and colon protection. A stent with a reducible diameter and a tubular protection structure were identified as the most promising methods to safely insert a barbed stent. The second problem of intestinal blockage was analyzed and causes of intestinal blockage were identified. Minimizing the frontal contact surface at the proximal end of the stent was assumed to be the most important factor to reduce intestinal blockage. A minimal frontal contact surface also reduces the force that feces exerts on the stent, thus reducing the chance of stent migration.

The previously mentioned experiments and analysis formed the basis of a list of design requirements for a new prototype stent. The requirements were divided into five categories; fixation, dimensions, materials, delivery method and performance. After analyzing different manufacturing processes and available materials, two concepts were designed and produced. One concept was chosen after careful considerations of both options.

The final prototype consisted of a stent with barbs that was 3D printed. This design was validated on the set of requirements that were set up earlier. The material of the stent is both flexible and stiff. For insertion and extraction of the stent, the diameter can be reduced by longitudinal inward folding, while the stent forms a rigid tube while in a deployed state. The barbs are stiff and strong and can withstand the force that would move the stent. The frontal contact surface of the stent was reduced by more than 18 times compared to the old prototype, to reduce the influence on intestinal blockage. The stent is surrounded by a sheet to protect the barbs and prevent tissue damage during insertion. A validation experiment showed that the stent was improved in fixation compared to the old prototype. Therefore, based on the research executed in this thesis, the final prototype should theoretically migrate less and cause less blockage. However, an ex-vivo experiment should validate the fixation force of the new prototype, and in-vivo experiments on pigs should determine if the prototype actually does not migrate and actually does not cause intestinal blockage.