Image registration is the process of aligning images by finding the spatial relation between the images. Assuming two images called fixed and moving images are taken at different time, different... Show moreImage registration is the process of aligning images by finding the spatial relation between the images. Assuming two images called fixed and moving images are taken at different time, different spatial location, or via a different imaging technique, the aim of image registration is to find an optimal transformation that aligns the fixed and the moving images. Performing an automatic fast image registration with less manual finetuning can speed up numerous medical image processing procedures. In addition, an automatic quality assessment of registration can speed up this time-consuming task. In this thesis, we developed a fast learning-based image registration technique called RegNet.Predicting registration error can be useful for evaluation of registration procedures, which is important for the adoption of registration techniques in the clinic. In addition, quantitative error prediction can be helpful in improving the registration quality. In this thesis, we proposed two quality assessment mechanisms using random forests (RF) and convolutional long short term memory (ConvLSTM), in which the latter performs faster and more accurate. Show less
In this work, we attempt to answer the question: "How to learn robust and interpretable rule-based models from data for machine learning and data mining, and define their optimality?".Rules provide... Show moreIn this work, we attempt to answer the question: "How to learn robust and interpretable rule-based models from data for machine learning and data mining, and define their optimality?".Rules provide a simple form of storing and sharing information about the world. As humans, we use rules every day, such as the physician that diagnoses someone with flu, represented by "if a person has either a fever or sore throat (among others), then she has the flu.". Even though an individual rule can only describe simple events, several aggregated rules can represent more complex scenarios, such as the complete set of diagnostic rules employed by a physician.The use of rules spans many fields in computer science, and in this dissertation, we focus on rule-based models for machine learning and data mining. Machine learning focuses on learning the model that best predicts future (previously unseen) events from historical data. Data mining aims to find interesting patterns in the available data.To answer our question, we use the Minimum Description Length (MDL) principle, which allows us to define the statistical optimality of rule-based models. Furthermore, we empirically show that this formulation is highly competitive for real-world problems. Show less
Particles are omnipresent in biopharmaceutical products. In protein-based therapeutics such particles are generally associated with impurities, either derived from the drug product itself (e.g.... Show moreParticles are omnipresent in biopharmaceutical products. In protein-based therapeutics such particles are generally associated with impurities, either derived from the drug product itself (e.g. protein aggregates), or from extrinsic contaminations (e.g. cellulose fibers). These impurities can affect product stability, as well as cause adverse effects once introduced into the human body. Particulate impurities are present over a wide range of sizes (from nanometers to millimeters) making them difficult to characterize by using a single method.Novel drug products may also contain particles that act as the active pharmaceutical ingredient (e.g., living cells) or a drug delivery vehicle (e.g., lipid nanoparticles). Unwanted immunotoxicity and inconsistent in vivo functionality can result from particle instability and aggregate formation. Therefore, the efficacy and safety of these therapeutics is dependent on the particle composition, quantity and size distribution.Consequently, well-established methods are required to quantify and characterize particles in the submicron- and micron-size ranges. In this thesis, we developed new approaches which allow for comprehensive characterization of the particle populations present in biopharmaceutical products, both as impurities or as API. Furthermore, the performed work focused on comparing different particle characterization techniques to allow a better understanding of the limitations and strengths of each method applied. Show less
The ongoing increase in antimicrobial resistance combined with the low discovery of novel antibiotics is a serious threat to our health care. Genome mining has given new potential to the field of... Show moreThe ongoing increase in antimicrobial resistance combined with the low discovery of novel antibiotics is a serious threat to our health care. Genome mining has given new potential to the field of natural product discovery, as thousands of biosynthetic gene clusters (BGCs) are discovered for which the natural product is not known.Ribosomally synthesized and post-translationally modified peptides (RiPPs) represent a highly diverse class of natural products. The large number of different modifications that can be applied to a RiPP results in a large variety of chemical structures, but also stems from a large genetic variety in BGCs. As a result, no single method can effectively mine for all RiPP BGCs, making it an interesting source for new molecules.In this thesis, new methods are explored to mine genomes for the BGCs of novel RiPP variants, with a focus on discovering RiPPs that have new modifications. RRE-Finder is a new tool for the detection of RiPP Recognition Elements, domains that are often found in RiPP BGCs. DecRiPPter is another tool that employs machine learning models to discover new RiPP precursor genes encoded in the genomes. Both tools can be used to prioritize novel RiPP BGCs. Two candidate BGCs are characterized, one of which could be shown to specify a new RiPP, validating the approach. Show less