The brain and central nervous system handle waste transport differently from the rest of the body. The pathways through which waste in the brain is transported is still a debated topic. The glymphatic system describes a pathway from the perivascular spaces around veins and arteries to the subarachnoid space. This pathway is filled with Cerebrospinal Fluid (CSF) and exploring this fluid and the orientation of the flow relative to the blood vessels is essential to get new insight into the waste transport and related neurological issues.

We analyze this system using a non-invasive diffusion weighted MRI sequencing technique called CSF-STREAM, which generates a DTI-like tensor field called CSF-Mobility, allowing us to track the CSF. The tensor data can be used to analyze the flow of the CSF by applying eigen-decomposition on the tensor matrix. We derive a vector field from the tensors' principal eigenvectors that represents the general flow of CSF.

We introduce a workflow aimed at analysis, comparison and interpretation of the CSF data. It begins by selecting a region that includes a vessel and the surrounding CSF. The vessel is segmented from the region of interest (ROI) and its centerline is extracted to represent the vessel's orientation. The vector field then visualized using a hedgehog plot with a color mapping that indicates the relative orientation, as well as with a streamlines visualization. The streamlines are restricted to the vicinity of the vessel by generating seed points within a spherical radius surrounding the centerline. To make the visualizations easier to interpret, we apply two different transformations on the vector field. A straightening transformation is applied to the vector field by using straightened Curved Planar Reformation (CPR) on the vessel centerline. An additional unfolding transformation is introduced in which the vectors are rotated around the straightened centerline, creating a cross-sectional view of the vector field. We conduct a user study to evaluate the workflow and compare the visualization methods and to to see which methods work best for interpreting the relation between the flow orientation and the vessels. Results show that the workflow and visualizations are indeed suitable techniques and that the straightening of the vector field along the vessel makes it easier to interpret the data, while the unfolding transformation makes the context too complex to understand with limited time and explanation. Overall, the workflow and tool set have the capability to give researchers more insight into the waste transport in the brain.

Pixel art relies on carefully constructed color ramps to simulate shading and depth within limited palettes. However, editing these ramps remains a tedious and error-prone manual process. This research introduces a semi-automatic tool that supports the detection and modification of color ramps in pixel art. The system builds a graph to model relationships between image colors based on perceptual similarity and spatial adjacency, then extracts and validates color ramps using customizable criteria. Afterwards, if a user edits a color, changes propagate consistently along the defined ramps, preserving their visual structure. The method streamlines palette editing while respecting artistic intent, offering both automation and control.

Pixel art can suffer from perceptual artifacts, such as banding and pillow-shading, which result from poor pixel placement and weaken visual quality. Banding occurs when two adjacent pixel segments of different colors align their endpoints along a shared axis. Pillow-shading is a pronounced form of banding, characterized by concentric strips that perfectly follow the object's outline. Prior methods for fixing artifacts in pixel art rely heavily on auxiliary reference images and do not address banding or pillow-shading. We present a semi-automated method that corrects both artifacts using established artistic guidelines and without relying on references. It supports user annotations to resolve ambiguities when multiple corrections are possible and to help preserve the artistic intent. We tested our software tool, Pixel Fixer, on several pixel artworks affected by banding or pillow-shading. Using an error metric that counts pairs of banded pixel segments, we observed consistent reductions across all inputs.
Related dataset 4TU.ResearchData: https://doi.org/10.4121/11446578-c1f3-474c-9cc4-6978a79b675b

In pixel art, texturing is the process of adding detail to an object to make it resemble a real material. While texturing is crucial in creating high-quality pixel art, it is an arduous process that is time consuming for artists. We present an algorithm that automates texturing by arranging elements throughout an input image in a way that implies a feature of the reproduced object, outlines its tridimensional shape, and respects the original shading. We additionally present heuristics on how to determine the color of a feature drawn over existing pixel art, and a method to create vector fields from user brush strokes. Aside from easing artists' work, this algorithm is an important step towards the procedural generation of pixel art. Quality is demonstrated by comparing results to manually authored textures.

This paper investigates how standard 3D reconstruction techniques can be adapted to work with orthographic projections of pixel art images. Reconstructing 3D models from 2D images is typically done with real world objects. However, little work has explored this problem in the context of pixel art, which has a lower resolution, and uses stylistic colors and shading, making it a problem that requires different or adapted techniques. We implement three reconstruction algorithms based on some common practices in the field: silhouette-based intersection, spatial carving using color consistency, and gradient-based depth estimation. Results show that silhouette intersection is an effective method for simple models, but fails to capture concave regions. Spatial carving addresses this limitation, although it also fails in some use cases. An adapted depth estimation technique is also used, which most accurately captures these regions, although it does not fill in the model as well. We implement a custom Blender plugin to support user annotations and improve accuracy in some ambiguous areas. We conclude that a hybrid approach, where silhouette intersection is combined with depth estimation, gives the most accurate results, and we suggest future work that can be done on the topic, including better color merging principles, adding custom viewpoint orientations, adding support for multiple object reconstruction and using perspective projections.

Wave Function Collapse (WFC) can be described as a family of algorithms, meant for content generation through constraint solving. One variant is Hierarchical WFC, where a hierarchical structure is given to the tileset used in WFC. This variant has seen use in a mixed-initiative procedural music generation model, where pieces of music simulating a single instrument playing chords and a melody are generated. In this paper, we explore how this model can be adapted to generate (coherent) pieces of music simulating several instruments playing in parallel. To achieve this, three model properties had to be defined: a new canvas structure, a set of constraints that instruments impose on each other, and the manner in which cells of all canvasses are collapsed. A hierarchical structure has been defined of a single section canvas, of which every cell has an inheriting chord canvas, of which every cell has an inheriting melody canvas for each instrument. The section cells impose constraints on inheriting chord and melody cells, and the chord cells impose constraints on inheriting melody cells. Besides this, melody cells impose constraints on cells belonging to other melody canvases (i.e. other instruments). With this canvas structure and these sets of constraints, the cells could be collapsed, such that coherent pieces of music simulating several instruments playing in parallel were generated.

AI-generated music is a huge research field with many different approaches and models being the result of it. One such model is the ProceduraLiszt model, which utilizes the Wave Function Collapse algorithm, an algorithm similar to constraint programming, to generate its music. This research builds upon that model. It does so by trying to reverse engineer a given piece of music into a set of satisfied constraints that the model is compatible with. We present an approach that allows for the inference of constraints of a given music file that adheres to the MIDI format called MidiAnalyser. We run our model on a set of music files and analyze the inferred constraints. The constraints include aspects like key and note range.

Procedurally generating a coherent and emotionally resonant piece of music can be very challenging. The Wave Function Collapse (WFC) algorithm is very effective when it comes to generating randomized patterns and maps that resemble an input sample. A version of this algorithm using a hierarchy of sections, chords and melody was used to create a model capable of generating music. In this paper, we extend the capabilities of this model to improve its utility and help composers more effectively utilize this music generation method to create engaging pieces. Our model offers improvements over previous methods by allowing composers to retain desirable elements of the music output while regenerating others, thus streamlining the iterative nature of music composition. We consider and compare different music visualization techniques and explore various user interface (UI) interaction methods to facilitate the effective selection of elements from the output. We designed and implemented this model with the conclusion that it significantly enhances the user experience and allows for creating a much more sound and complete piece of music compared to the original.

The tumor composition of breast cancer determines how tumors behave. Yet, there is a limited understanding of the arrangement of tumor cells in relation to cells in the tumor microenvironment (TME). In this research, we have characterized distance relationships between 324 cell-type pairs in 749 tissue samples of the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) study using Weibull distribution estimations, summarizing comprehensive spatial relationships with two parameters. The research showcased the first application of the method to a dataset of this substantial size and a dataset acquired with imaging mass cytometry. We identified distinct spatial relationships among breast cancer subtypes, particularly for basal, HER2-enriched, and luminal A tumors. The spatial relationships indicate attractive and repulsive interactions between different cell types and define cellular arrangements regardless of cellular abundance.
Moreover, several spatial relationships had significant associations with survival outcomes. Both findings could improve patient stratification and prognosis and emphasize the wealth of information that spatial analyses can retrieve. The results also confirm that Weibull distribution estimations are a suitable and effective method to summarize distance distributions. The application to other cohorts could lead to new insights into the tumor composition of different cancer types. Finally, the spatial profiling method was used to characterize neighborhoods and revealed distinct spatial relationships consistent with neighborhood characteristics but also provided new hallmarks.

A cellular automaton for simulating territories is presented. In it, cells have a certain amount of markings of two different groups. The amount of markings for each group gets higher based on the amounts of that group in neighboring cells and the amount of markings of the opposite group in the current cell to simulate an avoidance tendency. The markings also decay at a certain rate. Depending on the parameters, the simulation can end up in a mixed state, where there are no clear territories, or a segregated state, where both groups have a large amount of connected cells where they are dominant. Small changes in these parameters can change the outcome significantly. Unless an unconsidered combination of parameters changes it, this model is not the most realistic. It could however have uses in more influence-based processes, such as the spread and boundaries of languages or religion.

This paper describes the process and evaluation methods by which we adapted a Reaction-Diffusion model and an order parameter to monitor its segregation state in a 2D Cellular Automaton model. The model simulates Turing pattern formations, whose behavior will be studied by an order parameter formulated for this project. This order parameter acts as an indicator of the mixture state of the world. The implementation process of such a cellular automaton and an order parameter is described in detail, as well as the evaluation of this model and its order parameter. Different evaluation methods will be performed on this order parameter to show it as a valid indicator of the segregation state of the world in this model, as well as showcase different interesting properties and limitations of the parameters. Such methods will involve plots like Evolution, Phase-Transition, or Linear Regression plots, and the paper will conclude by summing up all the work taken into this project to confirm that this order parameter, with its formulation in this paper, is a valid and interesting one to use.

In 1952, Alan M. Turing presented a reaction-diffusion model that described formation of skin patterns. The patterns he predicted have later been found in various natural phenomena, such as in skins of fish or even in vegetation around termite hills. His patterns have even been taken to the micro-level. In 1984, David A. Young proposed a discretisation of this model, which enabled computer simulation. Both Turing and Young had only looked at two-dimensional patterns, until Martin Skrodzki and Konrad Polthier took the patterns to the third dimension in 2017. In this paper, these 3D simulations are generalised to produce patterns with more than two substances. We want to see whether Turing-like patterns also emerge there. To increase simulation speeds, a Graphics Processing Unit (GPU) implementation is described for this multi-species extension. Furthermore, we begin to investigate to what extent an order parameter can be defined, to analyse the formation of 3D structures. We found that our multi-species extension produced Turing-like structures that look similar to the ones found by former models. Our GPU simulation provided a significant performance increase. We also found that our order parameter can distinguish between well-mixed, well-segregated, and fully dominated states. However, it is yet unclear whether it can also be used to classify shapes.

Many artificial intelligence (AI) systems are built using black-box machine learning (ML) algorithms. The lack of transparency and interpretability reduces their trustworthiness. In recent years, research into explainable AI (XAI) has increased. These systems are designed to tackle common ML issues such as trust, accountability, and transparency. However, research into the evaluation of XAI is still low. In this paper, common trends in the evaluation of state-of-the-art model-agnostic XAI models and any missing or undervalued evaluation methods are identified. First, a taxonomy is explored, and an overview of existing evaluation metrics found in literature is made. Then, using this overview, a thorough analysis and comparison of the evaluation methods of 5 state-of-the-art model-agnostic XAI models (LIME, SHAP, Anchors, PASTLE, and CASTLE) is done. It has been discovered that only a small subset of the found evaluation metrics is used in the evaluation of the state-of-the-art models. Metrics that are not often assessed in user-studies but deserve more attention are (appropriate) trust, task time length, and task performance. For synthetic experiments, only fidelity is commonly assessed. The models are also only assessed using proxy tasks, none of them are assessed using real-world tasks. In addition, each identified metric was found to have various different measurement methods and units of measurement, indicating a lack of standardization.

The ever increasing presence of Machine Learning (ML) algorithms and Artificial Intelligence (AI) agents in safety-critical and sensitive fields over the past few years has spurred massive amounts of research in Explainable Artificial Intelligence (XAI) techniques (models). This new frontier of AI research aims to resolve some of the fundamental issues that accompany the usage of ML algorithms in sensitive fields such as medicine or criminology. For ML algorithm's to be implemented and used within fields such as medicine, it is not simply enough that they are proficient and effective tool at solving their assigned task (such as classifying whether a patient has Covid-19 or not). These ML techniques lack the ability to allow their human counterparts the possibility of understanding why they have made such a prediction, therefore not allowing the human supervisor a peak into the black box. This black box problem is one of the underlying difficulties that currently prevent the widespread adoption of ML/AI algorithms into these safety-critical fields. XAI techniques aim to solve this very issue and in particular Model-Agnostic XAI techniques aim to generate explanations on the predictions of any ML or AI algorithm. In this paper we will be exploring and investigating the different Model-Agnostic XAI techniques and looking into their individual advantages and disadvantages. After we've analyzed each individual technique, we will take a more global view into the different characteristics and how the XAI implementations compare to each other using different metrics for comparison. Finally we will propose future improvements and extensions that can be made to the various investigated XAI techniques.

The spread of AI techniques has lead to its presence in critical situations, with increasing performance that can compromise on its understanding. Users with no prior AI knowledge rely on these techniques such as doctors or recruiters with a need for transparency and comprehensibility of the mechanisms. The advent of Explainable Artificial Intelligence techniques responds to such issues with a diversity that has lead to the construction of a taxonomy for this domain. Notably, there is a distinction between model-specific and model-agnostic techniques. Rightly operational XAI technique should go through an evaluation process. In this paper, we investigate the different available tools and metrics for the evaluation of XAI techniques to then assess the evaluation quality of five state-of-the-art model specific techniques: TCAV, SIDU, ACE, Net2Vec and Concept Analysis with ILP. It has been concluded that despite broad existing literature on evaluation methods, there is a lack of exhaustive assessment of criteria and a lack of standardization in regards of the evaluation of these model-specific- techniques.

The significant progress of Artificial Intelligence (AI) and Machine Learning (ML) techniques such as Deep Learning (DL) has seen success in their adoption in resolving a variety of problems. However, this success has been accompanied by increasing model complexity resulting in a lack of transparency and trustworthiness. Explainable Artificial Intelligence (XAI) has been proposed as a solution to the need for trustworthy AI/ML systems. A large number of studies about XAI are published in recent years, with a majority discussing the specifics of XAI. Hence this work aims to formalize existing XAI literature from a high-level approach in terms of (1) benefits, (2) requirements, (3) challenges and (4) the underlying building blocks involved. Additionally, this paper presents a case study of XAI within the medical image analysis domain followed by future works and research directions in the field and from a general perspective, all to serve as a foundation and reference point to make the topic more accessible to novices.

There are many experiments conducted with Automatic Speech Recognition (ASR) systems, but many either focus on specific speaker categories or on a language in general. Therefore, bias could occur in such ASR systems towards different genders, age groups, or dialects. But, to analyze and reduce bias, the models require significant amounts of data to be trained on, and some corpora lack that. This is where augmentation techniques can be used to generate more unique data without any further collection of it. This paper explores the use of SpecAugment's frequency masking on such a corpus, JASMIN-CGN, for the Transitional regional accent of Dutch, with a hybrid GMM-HMM architecture, in order to reduce the bias for gender or age, for this specific dialect. The experiments show that SpecAugment does not manage to lower the WER (20.8% overall compared to the baseline model, which achieves 19.5% performance), on the contrary, it even increases the bias for age. The results are mainly attributed to the combination of low amounts of data + the hybrid architecture used, which proves SpecAugment to be a useful augmentation policy only for end-to-end models.

Automatic speech recognition (ASR) does not perform equally well on every speaker. There is bias against many attributes, including accent. To train Dutch ASR, there exists CGN(Corpus Gesproken Nederlands) and as an extension, the JASMIN corpus with annotated accented data. This paper focuses on improving ASR performance for NRAD (Northern regional accented Dutch) speech, training on speakers from the region of Overijssel. To achieve this improvement, the corpus data is augmented using Vocal Tract Length Perturbation (VTLP), which entails randomly warping the frequency of each recording using a factor in the range [0.9, 1.1]. The baseline and augmented ASR systems are trained using trigram GMM-HMM (Gaussian mixture model hidden Markov models) through the Kaldi toolkit on the DelftBlue supercomputer. This leads to improvements on word error rates (WER) for all speaker groups and styles, with an overall relative improvement of 14,64% and the biggest improvement observed for male speakers - from 25.15% WER to 19,68% WER. The impact of this augmentation on other accents and non-accented speech is not explored. This experiment can serve as a stepping stone for developing overall more robust and less biased Dutch ASR.

Building Automatic Speech Recognizers (ASRs) has been a challenge in languages with insufficiently sized corpora or data sets. A further large issue in language corpora is biases against regionally accented speech and other speaker attributes. There are some techniques to improve ASR performance and reduce biases in these corpora, known as data augmentations. One audio data augmentation, pitch shifting, has had successes in other experiments for increasing ASR performance. Pitch shifting it is tested in this paper on the JASMIN-CGN speech data set from the Southern regions of the Netherlands. Using a hybrid GMM-HMM ASR, two baselines are developed, one using all speech data from the region, the other only using native speech. For the former ASR, pitch shifting is found to not improve Word Error Rate (WER) performance or reduce bias, but the latter succeeds in improving WER performance and reduced bias for certain speaker groups when augmented.

A problem prevalent in many modern-day Automatic Speech Recognition (ASR) systems is the presence of bias and its reduction. Bias can be observed when an ASR system performs worse on a subset of its speakers compared to the rest rather than having the same overall generalization for everyone. This can be seen by using Word Error Rates (WER) as a metric. Depending on the ASR system in question the type of bias differs. However, techniques have been proposed and shown to succeed in reducing WER, and subsequently bias, by the use of data augmentation techniques for the recorded speech. These techniques perturb the audio in a certain way. Afterward, it is added to a model's training set and the model is retrained with the added data. One such technique is SpecSwap. This paper explores how using SpecSwap affects the WER performance of a hybrid-model ASR system using the JASMIN-CGN dataset's West-Dutch region. For comparison, a state-of-the-art data augmentation technique, VTLP, was also used, which has been shown to be effective in other cases. The experiments both led to a consistent WER increase. Therefore it was concluded that the data provided for the region was too little for the augmentation policy to be effective in any of the subcategories or in the overall performance of the system. However, SpecSwap shows potential in mitigating the widely discussed gender bias in ASR systems by reducing the difference between male and female speakers' WER.