Sten Vercammen

• 11/04/2023
• 5 p.m.
• Venue: Campus Middelheim, G.010
• Supervisors: Serge Demeyer, Görel Hedin & Markus Borg
• Department of Computer Science

Abstract

The growing reliance on automated software tests raises a fundamental question: How trustworthy are these automated tests? Today, mutation testing is acknowledged within academic circles as the most promising technique for assessing the fault-detection capability of a test suite. The technique deliberately injects faults (called mutants) into the production code and counts how many of them are caught by the test suite.

Mutation testing shines in systems with high statement coverage because uncaught mutants reveal weaknesses in code which is supposedly covered by tests. Safety-critical systems –where safety standards dictate high statement coverage– are therefore a prime candidate for mutation testing. In safety-critical software, C and C++ dominate the technology stack. Yet this is not represented in the mutation testing community: a systematic literature review on mutation testing from 2019 analysed 502 papers and reported that from the 190 empirical studies, 62 targeted the C language family and out of the 76 mutation testing tools, only 15 targeted the C language family. Despite the apparent potential, mutation testing is difficult to adopt in industrial settings, because the technique —in its basic form— requires a tremendous amount of computing power. Without optimisations, the entire code base must be compiled and tested separately for each injected mutant. Hence for medium to large test suites, mutation testing without optimisations becomes prohibitively expensive. ​ ​

To make mutation testing effective in an industrial setting, we set three objectives: (1) generate fewer mutants, (2) process them smarter and (3) execute them faster. To meet our objectives, we investigate the most promising techniques from the current state-of-the-art. This ranges from leveraging cloud technology to compiler integrated techniques using the Clang front-end. These optimisation strategies allow to eliminate the compilation and execution overhead in order to to support efficient mutation testing for the C language family. ​ ​

As a final step, we perform an empirical study on the perception of mutation testing in industry. The aim is to investigate whether the advances are sufficient to allow industrial adoption and to identify any remaining barriers preventing industrial adoption. ​ ​

In this Ph.D. thesis we show that a combination of mutation testing optimisation techniques from the do fewer, do faster, and do smarter are needed to perform mutation testing in a continuous integration setting. Furthermore, the industrial perception of mutation testing is evolving as additional organisations recognise its potential.

Senne Van Alphen

• 17/03/2023
  
• 11 a.m.
• Venue: Campus Drie Eiken, Building O, O.01
• ​Online PhD defence
• Supervisors: Annemie Bogaerts & Rony Snyders
• Department of Chemistry

Abstract

200 years ago, humanity started the industrial revolution by discovering fossil fuels, which lead to unprecedented technological advancements. However it has become alarmingly clear that the major environmental concerns associated with fossil fuels require a short-term transition from a carbon-based energy economy to a sustainable one based on green electricity. A key step concerning this transition exists in developing electricity-driven alternatives for chemical processes that rely on fossil fuels as a raw material. A technology that is gaining increasing interest to achieve this, is plasma technology.

Using plasmas to induce chemical reactions by selectively heating electrons in a gas has already delivered promising results for gas conversion applications like CO2 conversion and N2 fixation, but plasma reactors still require optimization to be considered industrially competitive to existing fossil fuel-based processes and emerging other electricity-based technologies. In this thesis I develop computational models to describe plasma reactors and identify key mechanisms in three different plasma reactors for three different gas conversion applications, i.e. N2 fixation, combined CO2-CH4 conversion and CO2 splitting.

I first developed models to describe a new rotating gliding arc (GA) reactor operating in two arc modes, which, as revealed by my model, are characterized by distinct plasma chemistry pathways. Subsequently, my colleague and I study the quenching effect of an effusion nozzle to this rotating GA reactor, reaching the best results to date for N2 fixation into NOx at atmospheric pressure, i.e., NOx concentrations up to 5.9%, at an energy cost down to 2.1 MJ/mol.

Afterwards, I investigate the possible improvement of N2 admixtures in plasma-based CO2 and CH4 conversion, as significant amounts of N2 are often found in industrial CO2 waste streams, and gas separations are financially costly. Through combining my models with the experiment from a fellow PhD student, we reveal that moderate amounts of N2 (i.e. around 20%) increase both the electron density and the gas temperature to yield an overall energy cost reduction of 21%.

Finally, I model quenching nozzles for plasma-based CO2 conversion in a microwave reactor, to explain the enhancements in CO2 conversion that were demonstrated in experiments. Through computational modelling I reveal that the nozzle introduces fast gas quenching resulting in the suppression of recombination reactions, which have more impact at low flow rates, where recombination is the most limiting factor in the conversion process.

Reyes Salas

• 24/02/2023
  
• 4 p.m.
• Venue: Campus Drie Eiken, Q.002
• ​Online PhD defence
• Supervisors: Wendt Müller, Luc Lens & Jan Mees
• Department of Biology

Abstract

The world is witnessing unprecedented rates of habitat degradation due to anthropogenic activities, especially urbanisation. Yet, some species are commonly believed to have successfully adapted to breed in urban areas. However, we have still a poor understanding of the actual fitness consequences. The fact that animals are attracted to an urban environment might conceal that urban landscapes can act as ecological traps, since even highly opportunistic species might have difficulties to keep up with the high rate of environmental change. This dissertation tackles this question by exploring the capacities to breed in changing environments along with in depth studies on the drivers of territoriality and on the role of the early life social environment for the offspring in order to deduce potential consequences of reproducing in urban landscapes. To this end, a colonial breeding seabird species, the lesser black-backed gull (Larus fuscus), which is thought to thrive in highly anthropogenic environments, is used as model species.

First, I explored whether nesting site relocations, as frequently occurring in rapidly changing urban environments, impact on reproductive success, and I could show that individuals that lost their breeding site due to anthropogenic activities laid smaller eggs and that the likelihood of skipping a breeding season increased. In a next step and by using GPS tracking devices, I then showed that investing time in territoriality imposes a carry-over effect on reproductive investment. While not measured explicitly, it can be assumed that time costs increase after a relocation and might be the cause of the observed negative reproductive investment in relocated birds.

Moreover, we currently lack a profound understanding of the importance of a territory for the chicks, even though it is known that the social early-life environment can shape an individual’s (behavioural) phenotype. This is particularly relevant in lesser black-backed gulls, because here chicks experience high levels of aggression when crossing into a neighbouring territory. I indeed found that chicks raised in dense areas where territories are closer together showed the lowest exploration activity. In a final step, I deployed a novel tracking technology to study the movement behaviour of the chicks in the colony, and I could show that the social environment also affected the movement activity, territory size and social associations among chicks from neighbouring nests.

Kasra Rafiezadeh Shahi

• 24/02/2023
  
• 4 p.m.
• Venue: Stadscampus, Building C, C.002
• ​Online PhD defence
• Supervisors: Paul Scheunders, Pedra Ghamisi & Richard Gloaguen
• Department of Physics

Abstract

In recent years, there has been an explosive growth in remotely sensed (RS) data usage in geoscience and Earth observation applications. The recent technical advancements allow users to acquire various types of rich information. A quintessential example of RS data is hyperspectral imagery. A hyperspectral image (HSI) provides rich spectral data over a wide range of the electromagnetic spectrum. Such information enables users to identify, track and distinguish different materials and objects. Another RS data example is light detection and ranging (LiDAR). LiDAR yields information on the altitude of the observed objects. Hence, it allows distinguishing objects that might share common spectral characteristics but have different altitudes (e.g., tree species). ​ 

Although RS data have high potential for material characterization, the processing of such data poses challenges, for example, on the high dimensional nature of the datasets or an efficient fusion of multiple data sources. Machine learning approaches march as the pioneer solutions to the aforementioned issues. ​ ​

Among machine learning approaches, supervised learning approaches perform accurately in various tasks (e.g., classification and regression). Nevertheless, such approaches demand an immense number of training samples during their process. Specifically, in geoscience and Earth observation applications, acquiring training samples is a labor-intensive, time-consuming, and expensive task. Moreover, in some cases, it is not possible to generate training samples due to limited accessibility. Therefore, supervised approaches constitute a shortcoming with respect to the availability of training samples. ​ ​

On the contrary, unsupervised learning approaches accomplish different tasks (e.g., feature extraction and clustering) by merely analyzing the data itself. To be more specific, the clustering problem refers to grouping similar pixels into clusters. In general, clustering approaches can be split into two categories: (1) Conventional shallow learning (CSL) and (2) Deep learning (DL)-based approaches. This study is devoted to the development of unsupervised CSL and DL approaches for single- and multi-sensor remote sensing data clustering.

Jinxin Wang

• 09/02/2023
  
• 2 p.m.
• Venue: Campus Drie Eiken, O.03
• Supervisors: Vera Meynen & Annemie Bogaerts
• Department of Chemistry

Abstract

The plasma-based dry reforming in a dielectric barrier discharge (DBD) reactor is important to achieve sustainable goals, but many challenges remain. For example, the conversion and energy yield of DBD reactors are relatively low, and the catalysts or packing materials used in existing studies cannot improve them, possibly due to the unsuitable properties and structures of catalysts or packing materials for plasma processes.

In order to study the effect of catalyst structure on plasma-based dry reforming, a controllable synthesis of the catalyst supports or templates was explored. In Chapter 2, an initially immiscible synthesis method was proposed to synthesize uniform silica spheres, which can replace the organic solvent-based Stöber method to successfully synthesize silica particles with the same size ranges as the original Stöber process without addition of organic solvents. Using the silica spheres as templates, 3D porous Cu and CuO catalysts with different pore sizes were synthesized in Chapter 3 to study the effect of catalyst pore size on the plasma-catalytic dry reforming. In most cases, the smaller the pore size, the higher the conversion of CH4 and CO2 due to the reaction of radicals and ions formed in the plasma. An exception are the samples synthesized from 1 μm silica, which show better performance due to the electric field enhancement for pore sizes close to the Debye length. Besides the pore size, the particle diameter of the catalyst or packing is also one of the important factors affecting the interaction between plasma and catalyst. In Chapter 4, SiO2 spheres (with or without supported metal) were used to study the effect of different support particle sizes on plasma-based dry reforming. We found that a uniform SiO2 packing improves the conversion of plasma-based dry reforming. The conversion of plasma-based dry reforming first increases and then decreases with increasing particle size, due to the balance between the promoting and hindering effect of the particle filling on the plasma discharge. Chapter 5 is to improve the design of the DBD reactor itself, in order to try to increase its low energy yield. Some stainless steel rings were put over the inner electrode rod of the DBD reactor. The presence of rings increases the local electric field, the displaced charges and the discharge fraction, and also makes the discharge more stable and with more uniform intensity. The placement of the rings improves the performance of the reactor at 30 W supplied power. ​ ​

Brent van Bladel

• 02/02/2023
  
• 5 p.m.
• Venue: Campus Middelheim, G.010
• Supervisor: Serge Demeyer
• Department of Computer Science

Abstract

As software has become ever important in our lives, all that code needs to be of a high quality. A common way to achieve this is via software testing, where additional "test code" is written with the sole purpose of finding mistakes in the original code, or "production code". As test code has the large responsibility of ensuring qualitative software, it is critical that the test code itself is of high quality as well. However, while the quality of test code is often synonymous with its ability to find bugs, it is equally important to ensure readability and maintainability of test code to allow agile teams working incrementally to update, extend, and maintain the test code each iteration. ​ ​

The presence of code duplication, or "code clones", can affect the readability and maintainability of code. While code clones have already been extensively researched in production code, research on test code duplication is limited. And yet, duplicate tests are a common occurrence, as the quickest way for a developer to test a new feature is to copy, paste, and modify an existing test. In this thesis, we address this gap in the literature by answering two research questions. First, we investigate whether the structure of test code can be exploited to detect semantic code clones. Second, we investigate whether test code duplication should be considered independently of production code duplication. ​ 

In the end, we show that test code is a rich source for studying clones and that further investigation is warranted.

Sarah Ahannach

• 24/01/2023
• 5 p.m.
• Venue: Campus Drie Eiken, O.03
• ​Online PhD defence​
  
• Supervisors: Sarah Lebeer & Gilbert Donders
• Department of Bioscience Engineering

Abstract

Women’s health and safety is receiving increased global attention in the last decennia, which - some may say - is long overdue. Science is unfortunately still not filling society’s needs when it comes to women’s health and particularly vaginal health. For instance, in the last 40 years no significant breakthroughs have been made on managing bladder and vaginal infections, despite important implications for physical and mental health of women, their children and partners. Yet, the field is moving forward and women’s health is now subject to a growing interest in the global microbiome research field. In particular, the vaginal microbiome has already been suggested to be crucial for vaginal disease prevention, successful fertilization and healthy pregnancies. Nevertheless, despite the growing understanding of the female microbiome and its importance, large-scale studies on the healthy women and the link to various lifestyle factors are generally lacking. Such knowledge on factors affecting the microbiome is needed when microbiome analyses are explored for forensic applications, such as to provide trace evidence helping reconstruct a crime event. This PhD work aims to lay cornerstones that will ultimately help all women achieve the best possible health and safety through microbial management and cutting-edge microbiome analyses. To this end, three research objectives were formulated: (i) to benchmark the female microbiome in Belgium and the main associating factors on its composition using a citizen science approach; (ii) to come to a better mechanistic and ecological understanding of the benefits of vaginal lactobacilli to the host and their potential as pre- and probiotics; and (iii) to explore microbiome analysis as a tool to investigate trace evidence in forensic casework. To support and enable the completion of these research objectives a citizen science project on women’s health with scientific and societal objectives named Isala (https://isala.be/en/) was set up, and a daughter project studying the application of microbial fingerprinting in sexual assault investigations, named GeneDoe. Overall, this PhD thesis contributes to a new understanding of women’s microbiome and how citizen science facilitates microbiome research while breaking the taboo on sensitive topics. It opens new research directions into the inner workings of the vaginal ecosystem, vitamin-producing bacteria, and applying microbiome analysis to forensics. This work will improve knowledge of the female microbiome stability and dynamics by presenting novel findings for clinical trials to unravel underlying mechanisms; the development of novel biotherapeutics; and the design of novel tools for diagnostics and criminal investigations.

Jeroen Van Houtte

• 19/01/2023
• 3 p.m.
• Venue: Campus Middelheim, G.010
• ​Online PhD defence​
• Supervisors: Jan Sijbers & Toon Huysmans
• Department of Physics

Abstract

Over the past decade, digital data generation and collection has become increasingly important in biomedicine. Surgeons heavily rely on biomedical data for diagnoses, pre-operative planning, follow-up, etc. Learning from large collections of data, such as optical surface scans and images, can help us in automating diagnoses and reducing human subjectivity. The knowledge of shape variability in a large dataset can be a guide for product development for example. Letting a computer “understand” images based on past examples enables computer-assisted robotic surgeries. ​ ​

This thesis provides data-driven solutions for biomedical problems. On a first level, we present a framework to combine the shape information of different patients into one digital model. On a second level, these digital models serve as prior knowledge for computers to automatically “understand” new data. A crucial step on both, the modeling and the application level, is the anatomical alignment of data. This step, known as “registration”, involves the identification of corresponding points between different data which remains a challenging task to computers.

The manuscript is divided into three parts. Part I provides an introduction to geometry and image processing, X-ray imaging and deep-learning. ​ ​

Part II presents the contributions of this thesis to (statistical) shape modeling of articulating bodies. Articulating statistical shape models (SSM) describe any individual up to a certain accuracy, while maintaining the possibility to be articulated into different poses. Hence, the acquisition of person-specific 3D models is no longer required. Two different articulating SSM’s have been constructed: a SSM of the human hand for splint design based on low quality 3D-scans, and a SSM of a horse limb for veterinary applications. The ability of SSM’s to describe many individuals also allows it to generate virtual data to train deep-learning models on. ​ ​

Part III of the thesis focuses on solving a specific registration problem in X-ray imaging. X-ray imaging or radiography is the most common imaging procedure for many orthopedic interventions thanks to its ability to visualize internal structures with a relatively low radiation dose and low acquisition cost. However, interpretation from 2D radiographs can be hampered by overlapping structures, magnification effects and the patient’s positioning. To avoid the difficulties associated with 2D projections, we developed two deep-learning methods, with and without statistical prior, to register a 3D model to a pair of radiographs. The registered model enables a 3D-interpretation, while keeping the benefits of RX over CT, in terms of costs and radiation dose. ​ ​

Shahid Ullah Khan

• 19/01/2023
• 2 p.m.
• Venue: Campus Drie Eiken, Q.002
• ​Online PhD defence​
• Supervisors: Karolien De Wael & Sammy Verbruggen
• Department of Chemistry

Abstract

Singlet molecular oxygen (1O2) has continuously attracted researchers' interest because of its involvement in various processes, such as in photodynamic reactions in biological and chemical systems. 1O2 is an effective electrophile and potent oxidizing agent and can be easily generated by photosensitization via the illumination of organic dyes with visible light. As described in Chapter 1, 1O2 has gained prominence in various applications such as wastewater treatment, photodynamic therapy of cancer, organic synthesis, and recently developed 1O2-based photoelectrochemical (PEC) sensing of phenolic compounds. Phenolic compounds are a potential source of contaminants that originates from industrial effluents and waste products of chemical and pharmaceutical industries. These phenolic compounds pose severe threats to humans and aquatic life after reaching the environment. Therefore, it is imperative to develop photoactive materials that efficiently generate 1O2 and oxidize phenolic compounds and antibiotics. The existing 1O2 generating photosensitizers (PSs) include porphyrins, phthalocyanines (Pcs), subphthalocyanines (SubPcs), and other dyes such as derivatives of xanthene (e.g., Rose Bengal (RB)), and fluorinated boron-dipyrromethene (BODIPYs), and phenothiazinium dyes (e. g. Methylene Blue) which display long-lived triplet excited state and can be used in 1O2-based applications. This thesis focuses on preparing efficient hybrid materials based on newly synthesized Pcs, different surface area titanium dioxide (TiO2) and plasmonic gold nanoparticles (AuNPs) for their use in the PEC detection of phenolic compounds.

The first focus was on developing a fast amperometric method to test the photo-electrocatalytic activity of 1O2 producing PSs dissolved in MeOH based on the redox cycling of an electroactive phenolic compound, hydroquinone (HQ) (Chapter 2). This method of testing PSs does not require the accumulation of a reaction product since the amperometric signal develops near instantly when the light is on, which enables dynamic monitoring of a PSs activity at varying conditions in a single experiment. This method was crucial to measure high 1O2 quantum yield and low yield in the same experimental conditions. Moreover, the obtained results revealed a range of working parameters affecting the PEC activity of PSs. ​ ​

The next goal was to immobilize tert-butyl substituted aluminum Pc (t-BuPcAlCl) on the solid support, which showed a high 1O2 quantum yield. However, before immobilizing Pc on a solid support such as TiO2, it is essential to know the electronic energy level of Pcs for the possible electron transfers from Pcs to TiO2. Therefore, Chapter 3 explored the (spectro)electrochemical properties of t-BuPcAlCl Pc. Next, in Chapter 4, t-BuPcAlCl Pc and other tert-butyl substituted Pcs with Zn central metal, t-BuPcZn, and its metal-free derivative t-BuPcH2 were immobilized on different surface area TiO2. The PEC activity of immobilized Pcs on TiO2 toward different phenols and antibiotics was studied, and the action mechanism was revealed and compared with sterically hindered fluorinated Pc F64PcZn. ​ ​

In the final part of this thesis plasmonic AuNPs were introduced combined with trimethylsilane-protected acetylene functionalized ZnPc (TMSZnPc) to study the synergistic effect that boosts the overall activity toward the detection of phenols under visible light illumination (Chapter 5) . The TMSZnPc was coupled with AuNPs via a click chemistry approach. The 1O2 quantum yield of TMSZnPc improved significantly after conjugating with AuNPs, and, subsequently, the PEC activity for detecting HQ. The theoretical and experimental investigation demonstrated that the plasmonic enhancement of TMSZnPc is driven by the near-field mechanism. This shows the importance of plasmonic AuNPs with other photoactive species for their use in 1O2-based applications. ​ ​

The fundamental knowledge obtained in this doctoral study will ultimately deepen the understanding of developing 1O2-based PEC sensors for detecting phenolic compounds and pharmaceuticals in the wastewater stream, helping to choose efficient materials and, in the last instance, a more sustainable future especially access to clean water for everyone.

Joaquim Sanctorum

• 12/01/2023
  
• 4.30 p.m.
• Venue: Campus Groenenborger, U.025
• ​Online PhD defence
• Supervisor: Joris Dirckx
• Department of Physics

Abstract

Birth is the most critical event in the early ontogenesis of mammals, of which we humans are also a part. Unfortunately, for a considerable percentage of newborns, this event can occur prematurely. According to the World Health Organization (WHO), preterm birth (before 37 weeks of pregnancy) occurs in 5% to 18% of annual births across 184 countries, representing about 15 million babies per year.

Modern advances in medicine have drastically increased the survival rates of premature babies, yet, many survivors are prone to developing lifetime disabilities. It was estimated that motoric deficits might develop in 5% to 27% of premature infants, of which celebral palsy is the most severe. To this day, there are no established therapies to overcome these issues due to the lack of fundamental knowledge on the connection between growth retardation at birth and the further development of the infant. ​ ​

Fundamental insights can be gained by studying the development of the movement apparatus and the related motoric defects during a longer period of time (longitudinal study) using X-ray imaging techniques. Movement patterns can be analyzed using X-ray stereofluoroscopy (XRSF), which provides 3D information on the movements of bones and joints. X-ray computed tomography (CT), on the other hand, yields structural information on the bones and the muscles. Repeated application of these two X-ray techniques and a combination of the data types can provide essential information to study the movement apparatus over time. ​ ​

Due to their hazardous nature, these techniques are to be avoided in infant-related research. However, low-birth-weight piglets exhibit a high degree of physiological similarity to human babies and therefore are a clinically relevant model. Yet, piglets develop rather quickly (a matter of days or even hours), posing profound time constraints on the envisaged longitudinal research approach, as XRSF and CT are not available in the same facility. ​ ​

The work presented in this dissertation is aimed at combining XRSF and CT in the same set-up to make longitudinal studies on the motoric development in rapidly developing animals possible. The main research topics discussed in this manuscript are image distortion correction, geometry calibration of the modular set-up, extended field-of-view tomography and the use of multi-exposure fusion techniques. The dissertation thus covers a broad range of challenges that may arise when transforming a standard XRSF set-up to a high-speed tomograph, and may serve as a guide to face these.