Independent Research Fund Denmark awards DKK 575 million to 192 new research projects across all scientific research areas. Three projects from DTU Compute - Department of Applied Mathematics and Computer Science are among them.
Professor Tim Bjørn Dyrby from the research section Visual Computing
3D label-free optical mapping of neurite orientations and cell body density in brain tissue at micrometer resolution
The brain consists of a network of cable connections that ensure different functional brain areas can communicate and enable us to control our arms and legs. These cable connections are made up of bundles of microscopic tube-like axons and round cells that transmit signals. Brain diseases can attack the axons, causing parts of a cable connection to malfunction. Magnetic Resonance Imaging (MRI) can provide insight into the organization of these cable connections and any disease attacks, but the low image resolution blurs important details.
Professor Tim Bjørn Dyrby and his team have discovered a new method that uses optical imaging microscopy to visualize individual cable connections without the use of a contrast agent. Since it is an optical imaging method, chemicals are used to make the tissue sample transparent, like a glass brain. In this project, we will develop artificial intelligence algorithms to determine the direction of axons along a cable connection, even if two connections cross each other. Based on these directions, Dyrby and his team will recreate a detailed virtual image of the entire brain’s network at a micrometer scale, and we will test whether we can detect local disease attacks on the cable connections.
Their microscopy method could become a unique tool that provides detailed insight into even the smallest disease attacks on cable connections, which cannot be seen today. This gives doctors new opportunities to predict the impact of diseases on patients and to develop and test new drugs.
Professor and Head of section Philip Bille from the research section Algorithms, Logic and Graphs
Adaptive Pattern Matching
An important problem in processing large amounts of data is to handle large strings (documents, genome databases, internet traffic) to find patterns (small pieces of text, sequence motifs, virus signatures). Typically, pattern matching is done using algorithms whose efficiency depends only on the pattern size and the string. However, this is too pessimistic or impossible to achieve in many scenarios, and a more nuanced approach is needed. In this project, Professor and Head of section Philip Bille and his team will develop algorithms that efficiently can take advantage of combinatorial properties of strings and patterns (distribution of patterns occurrences, repetitions in the strings, compressibility, and the interplay between these) to obtain new adaptive results for pattern matching.
Associate professor Kristian Uldall Kristiansen from the research section Mathematics
Mixed-mode oscillations and the zero-Hopf bifurcation
Mixed mode oscillations (MMOs) are trajectories of a dynamical system with clear alternations between distinctly large and small amplitude motions. Mathematically, they occur in systems with multiple time scales but the theory is still incomplete; e.g. it is still an open problem whether MMOs are periodic or chaotic. However, MMOs bear some similarities to the zero-Hopf bifurcation, which marks a special onset of oscillations. In particular, the analysis of both problems relate to tracking special solutions in the presence of exponentially small terms. Until recently (2019), it was an open problem whether perturbations of the zero-Hopf bifurcation produce chaotic dynamics. For its solution, researchers developed new techniques. With this proposal, Associate professor Kristian Uldall Kristiansen and his team will for the first time extend and apply these methods to the analysis of MMOs.
