Dr Ben Cardoen MSc PhD

• PhD in Computing Science, Simon Fraser University, 2024
• MSc in Computer Science, University of Antwerp, 2017
• BSc in Computer Science, University of Antwerp, 2015

Ben Cardoen's research develops mathematical and computational methods to extract reliable structure from noisy, high-dimensional data, with a focus on graph-based representations and operator-theoretic approaches. He is particularly interested in settings where direct structural inference is either computationally intractable or statistically unstable, and where meaningful signals must be recovered under perturbation, sparsity, or heterogeneity.

A central theme of Ben's work is the development of spectral and operator-level tools that characterise stability, separability, and identifiability in complex systems. This includes recent work on perturbation-aware graph representations and filtration methods that amplify informative structure while suppressing noise, as well as theoretical analyses of how geometric and stochastic perturbations affect graph spectra. Alongside this, he develops data-driven frameworks for multichannel and temporal data, with an emphasis on learning under sparse supervision and preserving interaction structure across channels and time.

These methods are motivated and validated through applications in biological imaging, where data are often noisy, multiscale, and only partially observed. Dr Cardoen's work aims to bridge mathematical theory with practical pipelines for analysing complex biological systems, enabling reproducible and interpretable discovery.

Ben completed a PhD in Computer Science, focusing on functional discovery in multichannel superresolution microscopy. Prior to this, he obtained an MSc in Computing Science on distributed optimisation methods and a BSc in Computing Science on simulation and computational modelling. Following his PhD, Ben held a short postdoctoral position at University of British Columbia,and subsequently joined the University of Birmingham as a Research Fellow in Mathematics, continuing work at the interface of mathematics, computer science, and biomedical imaging.

Before joining the University of Birmingham as a Research Fellow in Mathematics, Ben worked across interdisciplinary settings spanning mathematics, computer science, and biomedical imaging.

• 4TAM – Spring 2026 Signal Processing on Biological Graphs

Dr Cardoen co-supervises two PhD students, spanning fractal geometry in biomedical imaging and mathematical modelling of memory-encoding hormesis in biological systems.

Current Research

Dr Cardoen's current work focuses on developing operator and spectral methods for analysing noisy, high-dimensional systems, with particular emphasis on graph-based representations. A central aim is to understand how meaningful structure can be identified, preserved, or recovered under perturbation, sparsity, and measurement noise.

One strand develops perturbation-aware frameworks for graphs, analysing how geometric or stochastic changes affect spectral properties and the separability of structure. This includes work on spectral fragility and stability, characterising when graph-based representations remain informative under realistic noise models.

A second strand introduces filtration-based approaches that amplify informative signal while suppressing noise, providing a principled way to improve robustness across scales. These methods connect to existing graph descriptors while extending them to settings with higher variability and uncertainty.

A third, emerging direction explores agentic and learning-based approaches to navigating large structural search spaces. Here, efficiently computable operator-level signals are used to guide the exploration of candidate structures, enabling data-driven identification of salient motifs and higher-order organisation without exhaustive enumeration. This connects to ongoing work on evidential reasoning in complex networks and aims to bridge mathematical structure with adaptive, model-guided discovery.

In parallel, Ben develops data-driven frameworks for multichannel and temporal systems, particularly in biological imaging. This includes active learning approaches for detection under sparse supervision, and methods that preserve interaction structure across channels and time. The goal is to enable reproducible and interpretable discovery in complex biological data.

Applications in Biological Systems

These methods are applied to fluorescence microscopy and related imaging modalities, where data are often noisy, multiscale, and only partially observed. Current work includes analysing spatial and temporal organisation in intracellular systems, using graph-based representations to quantify proximity, interaction, and structural change over time.

Previous Research

Dr Cardoen's doctoral work focused on functional discovery in multichannel image analysis, developing computational methods to identify meaningful structure across heterogeneous data sources. This work combined image analysis, optimisation, and graph-based representations, and laid the foundation for subsequent research on robustness and structure in complex systems.