Summary Introduction The increasing use of cardiac catheterisation procedures has raised concerns about occupational radiation exposure and its associated health risks for clinicians. Radiation exposure can lead to deterministic and stochastic effects, including cataract and various cancers. To minimise these risks, the ALARA (As Low As Reasonably Achievable) principle must be followed in the clinical setting by applying the available measures. Radiation protection consists of passive and active measures. Passive protection includes architectural shielding, stationary shielding, and personal protective equipment, while active protection focuses on minimising radiation during interventions, employing adjustable lead screens, modifying imaging techniques and provide feedback to improve the usage of these measures. However, current methods for measuring and providing feedback on radiation exposure are insufficient. The primary goal of this research is to find a comprehensive and improved approach to guidance and implementation of protective measures, to ensure the safety and well-being of healthcare workers who face radiation exposure in their daily practice. An integral component of achieving this goal is to create a model that calculates radiation exposure for clinicians in the cathlab with the use of a lead screen. This research aims to determine the feasibility of using dose rate data from measurements without a lead screen to estimate the dose rate in an environment where a lead screen is employed. Moreover, potential methods for communicating dose rate data to clinicians during a procedure will be explored, aiming to increase their awareness of radiation exposure and encourage the adoption of appropriate protective measures without distracting them from the procedure itself. Method Data collection is necessary to construct a model that can predict the radiation scattering. Measurements in the catheterisation laboratory were performed by placing Philips DoseAware Detectors (PDDs) at various locations and heights. Combinations of fluoroscopy or acquisition mode, anteroposterior (AP) view or a 40° tilted view of the C- arm, and the presence or absence of a lead screen, resulted in eight different setups, with each 168 measurements. In the programming process careful consideration has been given to the ease of refining and improving the model afterwards. Three potential source options for the radiation were considered: the centre of the phantom, a point where 50% of the scattering has occurred, and a 3D representation of the entire phantom. The modelled lead screen together with the source was used to create an attenuation grid, which will be used for the estimation of the effect of the lead screen. The adequacy of the estimated dose rates with the lead screen was evaluated using boxplots and counting the number of correctly estimated points. Then, the data was interpolated using Kriging. The interpolated data was visualised, and the radiation doses for the assistant and cardiologist were calculated at the chest and head levels. The exposure calculation was performed for three lead screen positions and a scenario without the lead screen. The placements considered were close to the clinician, close to the phantom, and at the edge of the phantom. To explore radiation feedback preferences, a literature review and a questionnaire for interventional cardiologists were conducted. Results Measurements were taken for eight different setups, each with 168 measurements, including the combinations of fluoroscopy or acquisition mode, anteroposterior (AP) view or a 40° tilted view of the C-arm, and the presence or absence of a lead screen. The interpolated data revealed that the dose rate is higher when the acquisition mode is used, and radiation levels are higher when the C-arm is rotated. Three different source types were incorporated into the measurements, and the accuracy of the estimations was evaluated based on whether they were within a deviation of 0.1 mSv/h from the actual measurements. The 3D source was found to be the most accurate representation of the source in the model. Visualisation of inaccurately estimated values showed that the model overestimated the amount of radiation blocked in certain cases. The radiation exposure for clinicians increased as the lead shield was positioned closer to the radiation source. A questionnaire was conducted with five interventional cardiologists to investigate their radiation safety practices. The results indicated that radiation safety is deemed important, with an average score of 8/10. The lead screen is an essential shielding component, but its placement varies, and there is uncertainty regarding its optimal positioning. Monthly reports on total radiation dose exposure were provided. The clinicians hardly paid attention to this information, as they believed that any potential harm had already occurred. The highest-rated techniques were advice on optimal lead screen position (displayed on the lead screen) and a visualisation using augmented reality that displays the scatter pattern on the monitor. Discussion and Conclusion This study addresses radiation safety in interventional cardiology by exploring protective measures. A preliminary model was developed to predict radiation exposure based on the location of the lead screen enabling the computation of the radiation dose received by clinicians during a procedure. This dose rate model helps to visualise the impact of lead screen positioning on radiation exposure, which can aid in real-time decision-making during procedures. The model also revealed that placing the lead screen closer to the clinician, rather than closer to the source, could potentially enhance radiation attenuation by a factor of 6.2. This results in a possible dose reduction up to 32.5 times. The most preferred option by the interventional cardiologists was to display the suggestion of the optimal position of the lead screen directly on the shield. A potential addition to the feedback system could be an augmented reality (AR) visualisation of the radiation effect on the monitor, clearly showing safe areas for personnel. The precise feedback mechanism should be piloted and refined to ensure seamless integration into clinical practice without distracting the clinician from the procedure itself. Understanding the limitations of the measurements and the model is crucial for optimising its application in practice and the improve the reliability. The accuracy of measurements obtained with the DoseAware badges, while sufficient for this project, could be improved by using more accurate instruments. The phantom used in the study does not accurately represent real patients, and future research should involve patient- specific inputs and more realistic phantoms. One of the key future objectives include collecting more data in different settings and positions of the C-arm, as well as investigating different positions and orientations of the ceiling-mounted lead screen. The model itself has limitations in its representation of the scattering source and interpolation methods. To improve the model, researchers should account for other sources of scatter, and assess alternative interpolation techniques using more powerful computing resources. Moreover, to calculate the received radiation dose, the model should consider the entire body, accounting for the protection provided by a lead apron and the concept of effective radiation dose. Accomplishing these objectives will allow the model to play an important role in enhancing cathlab radiation safety, benefiting the health and safety of clinicians.

Coronary artery disease (CAD) is one of the leading causes of death and disability worldwide. In CAD, the coronary arteries, that supply the myocardium with oxygen, are narrowed or even blocked by a process called atherosclerosis. Invasive coronary angiography (ICA) is the gold standard for the diagnosis of CAD, as well as for intraprocedural guidance of percutaneous coronary interventions. At this moment, stenosis severity is determined via visual inspection by a cardiologist. This method has several important drawbacks: a significant inter- and intra-rater variability and a high positive prediction bias. Currently available additional assessment techniques improve results, however, they come with prolonged procedural time, complication risks and increased costs. Additionally, a relevant number of syndromes exists that cannot be diagnosed sufficiently with these techniques. Automated software that uses all information within the ICA images and relates it to the context of complaints and outcomes, could be a valuable tool for improvement. Since machine learning (ML) can find relations between patient groups based on images, this could be a possible solution. However, despite promising results, clinical implementation of ML is still limited. Limited clinical applicability of developed algorithms and a lack of large high-quality datasets are the cause of this. In this research it was investigated if ML can help in solving the diagnostic and therapeutic challenges encountered in interventional cardiology (IC) and what is needed to apply ML to ICA images and other sources of coupled medical data. This was done in several steps. First the expectations and perceived barriers by interventional cardiologists on ML-based algorithms in clinical practice were assessed. Next, the feasibility of health insurance code-based querying of electronic health records (EHRs) for the creation of ML datasets was assessed. Third, a proof-of-concept study on creating a dataset and a deep learning network for predicting lesion significance on ICA images was carried out. Last, a roadmap for the curation of data for the development of ML models in IC was created. Generally, interventional cardiologists have positive expectations for applying ML in their clinical practice. Furthermore, the willingness to collaborate in the development and clinical validation of ML algorithms is high. This is essential for translating ML models to clinical practice. Health insurance code-based querying of EHRs is not a feasible approach for creating datasets for the complex syndromes which currently pose diagnostic challenges in IC. However, the EHRs hold valuable data for the development of ML datasets. The same goes for the data collection strategy in the proof-of-concept study and it was shown that it is feasible to train deep learning networks on this data. ML shows promising results for solving the diagnostic and therapeutic challenges encountered in interventional cardiology and directions for further research were identified. For the creation of algorithms that can be applied in clinical practice, close collaboration between ML professionals and clinicians is needed. Besides this, further research is needed to develop scalable strategies for the creation of large datasets, containing adequately labelled patients that represent the real-life population.

With a 5-year survival rate of only 9%, pancreatic cancer is one of the deadliest types of cancer. Among other things, this is caused by the extreme difficulty of diagnosing recurrent pancreatic cancer in an early stage. One of the important next steps in discovering pancreatic cancer automatically on a CT scan is finding healthy pancreatic tissue. This research explores the feasibility of segmenting the pancreas on a CT scan using a deep learning approach, focusing on postoperative cases after pancre- atic resection, by combining state-of-the-art segmentation models for the preoperative pancreas with novel techniques, such as 3D Multi-Scale Convolutional Blocks (MCBs) and KNet architectures. Us- ing pretraining on preoperative data, a complete segmentation pipeline was designed to segment the pancreas in the postoperative state. The experimental results demonstrate that deep learning models can effectively segment the pancreas, despite the increased complexity and difficulty of the postoper- ative state. Notably, employing advanced techniques such as the KNet architecture and MCBs yields significant performance improvements. The newly designed architectures presented in this research, MKNet, MSKNet, and MAKNet, achieve state-of-the-art results for both preoperative and postoperative pancreas segmentation, greatly improving the Hausdorff Distance (HD) and 95th Percentile Hausdorff Distance (HD95) in particular. In the preoperative state, substantial performance improvements were observed compared to the previous state-of-the-art, with a 12.17% increase for HD and a 14.64% in- crease for HD95. Similarly, in the postoperative state, improvements of 11.99% for HD and 13.25% for HD95 were measured. Additionally, qualitative evaluations were conducted with seven experienced radiologists and radiotherapists to validate the algorithm’s performance in a clinical context. Remark- ably, in 83% of the cases, the algorithm was evaluated to accurately segment the pancreas, requiring minimal or no modifications according to the medical experts. This study not only contributes to the state-of-the-art in pancreas segmentation but also introduces comprehensive quantitative and qualitative evaluation methodologies. Furthermore, it addresses the segmentation of the postoperative pancreas, which has not yet been touched upon in previous re- search. Limitations are acknowledged, such as the availability of a limited dataset, and future research directions are proposed, including generalizability and enhancements to the proposed models. Despite the limited availability of data, which impacts the performance and generalization capabilities to some degree, the research in this thesis showcases the capability of the MKNet family architectures to ac- curately segment the postoperative pancreas, offering potential benefits for medical applications, data annotation acceleration, and future research in this domain.

Introduction The objective of this research is to assess whether there is a correlation between quantitatively assessed ablation margins and the occurrence of local tumor progression (LTP) using dedicated image processing software. Methods 28 patients with 45 de novo HCCs treated with percutaneous thermal ablation, e.g. radiofrequency ablation (RFA) and microwave ablation (MWA) between January 2014 and March 2019 were retrospectively included. Semi-automated segmentation of the liver and the ablation zone, manual segmentation of the tumor, and semi-automated registration of pre- and postprocedural contrast enhanced computed tomography (CECT) and magnetic resonance imaging (MRI) images was performed using in-house developed software deLIVERed. The image processing results were used for quantitative analysis of the minimal ablation margin (MAM) and the ablation margin surface area. The outcome of the quantitative analysis was compared to LTP occurence. Results Image processing of the scans of 39 of the 45 tumors was feasible. 5/39 tumors developed LTP. Based on quantitative analysis, thermal ablation was insufficient in 29/39 tumors. Of these 29 tumors, 4 developed LTP. The median MAM for the LTP group and no LTP group was -4.6 and -1.7 mm, respectively. There was no clear correlation between the MAM and LTP occurrence. The median tumor surface area exposed to insufficient margins was 28.9% and 7.7.% with and without development of LTP, respectively. Conclusion Quantitative analysis of postablation images can provide insight in ablation margins and the development of LTP. The ablation margin surface area provides additional information to the MAM. However, additional research is needed in order to further investigate the implications of ablation margin surface area.