titles and abstracts
Titles and abstracts of the talks will be progressively uploaded once submitted by the speakers
Kristina Ago
Palindromic (sub)words in the looking glass: From reconstruction challenges, down the rabbit hole, to a sophisticated palindromicity measure
Abstract: Palindromes play a central role in combinatorics on words and encode surprisingly rich structural information about finite words. In particular, every binary word is uniquely determined, up to reversal, by the set of its palindromic subwords, which highlights the role of palindromes in reconstruction problems. In this talk we present several results concerning palindromic subwords and the so-called MP-ratio, which is a kind of measure of palindromicity. We discuss some bounds on the MP-ratio and how they vary with respect to the arity of words, present some proof techniques, and map how much remains to be explored.
Benedetta Baldini
From CBCT to Virtual Head: AI-Based Automation and Multimodal Fusion in Maxillofacial Imaging
Abstract: Cone Beam CT (CBCT) is widely used in dental and maxillofacial radiology due to its ability to provide detailed 3D imaging of craniofacial structures. This work investigates the development of artificial intelligence (AI) approaches to automate key steps in CBCT-based maxillofacial image analysis, including cephalometric landmark detection and tooth segmentation. In addition, multimodal integration with radiation-free imaging modalities was explored to enhance the information content of CBCT data. The ultimate goal is the creation of patient-specific virtual head model that can support diagnosis, treatment planning, surgical simulation, and educational applications.
To this end, a multicentric annotated dataset of 1300 multicentric clinical examinations was collected. Several deep learning architectures, 3D V-Net, attention-based 3D U-Net, graph convolutional networks, and image registration algorithms, were trained and validated. The developed methods demonstrated excellent agreement with manual processes (gold standard), significantly reducing processing time and achieving clinically relevant accuracy: mean cephalometric landmark error < 1.8 mm, segmentation overlap (Dice score) of 0.93, and a fusion deviation of approximately 1 mm.
Claudia Balducci
Unmasking Alzheimer’s Disease: The Multifactorial Puzzle Beyond Protein Misfolding
Abstract: Alzheimer’s disease (AD), the most common form of dementia affecting nearly 33 million people worldwide, is a neurodegenerative disorder marked by synaptic and memory loss, chronic neuroinflammation, and neuronal death. While the accumulation of β-amyloid (Aβ) aggregates has long been considered the primary culprit, recent research has highlighted neuroinflammation as an equally central player in AD pathogenesis and progression. The hallmarks of AD—Aβ plaques and tau neurofibrillary tangles composed of hyperphosphorylated tau—initiate a cascade of immune responses in the brain, primarily mediated by microglia and astrocytes. Microglia, initially protective by clearing Aβ and damaged neurons, can become dysregulated over time, releasing pro-inflammatory cytokines and reactive oxygen species that further damage neurons. Similarly, activated astrocytes contribute to chronic inflammation, compromising synaptic function and the integrity of the blood-brain barrier. Emerging evidence also points to a bidirectional relationship between neuroinflammation and Aβ/tau pathology, where chronic inflammation amplifies Aβ and tau accumulation, creating a self-perpetuating vicious cycle. Genetic predispositions, including mutations in immune-related genes, increase AD susceptibility. Collectively, these insights have positioned immune cells at the forefront as promising therapeutic targets in the fight against AD.
Bojan Bašić
Constructing large patches layer by layer: how far can we go? Depicting a cross-section of Heesch’s tiling problem
Abstract: Given a geometric shape, a natural question arises: can ceramic tiles cut in that shape tile the entire (infinite) wall? Although this question may appear simple at first glance, some shapes turn out to be surprisingly deceptive. They appear fully “cooperative,” they seem to fit together naturally, suggesting that a tiling should exist. Piece after piece falls into place, everything looks promising—until, after extending the construction quite far, we suddenly find ourselves stuck.
To capture this phenomenon, the Heesch number of a shape is defined as the maximum number of layers of congruent copies that can be wrapped around it, without gaps or overlaps. Shapes with a large but finite Heesch number exemplify this deceptive cooperativeness: they allow impressively large patches to grow before the process inevitably breaks down. The central question, known as Heesch’s problem, asks whether shapes with arbitrarily large (finite) Heesch numbers can be constructed. Despite decades of investigation, the answer remains unknown.
In this talk we trace the historical development of constructions with increasingly large Heesch numbers, review the current state of the art, and briefly discuss several related questions.
Miranda Bellezza - Azzurra di Palma
Valentina Bordin - Paolo Finotelli
The DONAU project: Imaging and epidemiology as possible ways to approach the study of degenerative diseases
Alberto Bravin - Maurizio Santini
Micro-computed tomography for virtual histology: cutting edge tools to investigate biological samples
Abstract. Recent developments in micro-computed tomography have revolutionized the approach to X-ray microscopy. The development of highly spatially-coherent X-ray sources combined with efficient detection systems, the availability of performant control software able to guide users in achieving multiscale analysis of a single sample, and the possibility offered by commercial and custom-made artificial-intelligence-based software, has permitted the obtaining of performant sample analysis pipelines. This work will present the early results of the 3D-Virtual Histology laboratory recently installed at Milano-Bicocca’s premises, equipped with an X-radia Versa 615 micro-CT system and the realization of the X-ray laboratory presently under commissioning at the Bergamo University, equipped with a custom-made system optimized for phase-contrast tomography imaging.
Maria Michela Del Viva
Efficient encoding of dynamic visual scenes based on elementary 3D features
Abstract: Previous work has suggested that, given the need to rapidly process large amounts of visual data,the visual system extracts bottom-up saliency maps of static scenes using a limited set of features. These edge-and bar-like “optimal” 2D features can be derivedby applying constrained maximum-entropy principles to the frequency distribution of all possible features.It is interesting to ask the same question in the spatio-temporal domain, as motion information constitutes a crucial saliency cue in early visual analysis. However, extending the approach to real-life dynamic scenes leads to an exponential increase in the number of possible 3D features, making it too demanding to be performed with standard computing means.
We address this challenge by using specialized big-data reduction algorithms (Floating Top-k) adapted for execution on FPGA devices. This new approach makes it computationally feasible to identify optimal3Dfeatures by applying the constrained maximum-entropy method to a much larger space of potential features than previously possible.
We created movie sketches using optimal3D features (3×3pixels×3frames) extracted from a large video database and tested their effectiveness in a discrimination task.Observers’ accuracy remained very high across a broad range of model parameters, supporting the robustness of the model in predicting salient motion features.As a control, we generated alternative sketches by selecting sets of 3D features from frequency ranges different from the constrained maximum-entropy selection, matching either its information content or its constraint values. In both cases, these alternative“non-optimal”selections yielded poorer discrimination performance, indicating that neither total information northeconstraints alone are sufficient to produce meaningful representations ofmoving scenes.
Overall, these findings suggest that the visual system reduces complex dynamic inputs at an early processing stage by selecting a limited number of elementary featuresthatmaximizeinformation transmission under computational constraints.
Funding acknowledgments. This project was funded by the European Union –Next Generation EU, in the context of the grant PRIN 2022 (Project:"Real time reconstruction of data from LHC experiments with a distributed FPGA system", Grant no. 2022Z3K93E, CUP: I53D23001540006).
Gabriele Fici
String Reconstruction from (Multi)-Sets of Substrings
Abstract: I will discuss the problem reconstructing a string from a set (resp. a multiset) of substrings. This problem has application in several areas of computer science, e.g. bioinformatics and data mining.
Azzurra di Palma - Riccardo Giannuzzi
Eyes on the reels: tracking immersion in slot machines
Abstract: Gambling Disorder is driven by dynamic cognitive and motivational processes such as craving and cognitive distortions, which fluctuate during play and are difficult to capture through traditional self-report measures. In this study, we introduce SAFESLOT, an immersive virtual reality framework designed to objectively measure attentional and behavioral dynamics during slot-machine gambling. Eye-tracking data were acquired at 60 Hz and transformed through a preprocessing pipeline into structured attentional profiles based on six Areas of Interest (AOIs). Each participant was represented as a normalized six-dimensional feature vector describing gaze distribution across the slot interface. Unsupervised k-means clustering was applied to identify latent attentional phenotypes, with the optimal number of clusters determined using the Davies–Bouldin index. Four distinct profiles emerged, including a credits-focused pattern characterized by increased fixation on performance indicators. This cluster showed significantly higher post-session immersion and craving scores compared to other groups. Our findings suggest that immersion leaves a computationally identifiable signature in gaze behavior. These results highlight the potential of multimodal VR-based paradigms combined with machine learning to identify objective behavioral markers associated with gambling-related risk.
Peter Gritzmann
Diophantine Discrete Tomography: Structure, algorithms, and complexity
Abstract: Discrete Tomography deals with the inverse problem of deriving information about a function $f:D \rightarrow C$ with finite support which is only accessible through certain queries. Often, $D=\mathbb{R}^n$ or – in the lattice case - $D=\mathbb{Z}^n$ but also other domains have been studied. The most relevant queries include line and hyperplane queries specifying values of the X-ray and Radon transform, respectively. Typical codomains $C$ of interest are $\{0,1\}$, $\mathbb{N}_0$ and $\mathbb{Z}$, or appropriate variants in the polyatomic case. We will first briefly survey results which show how heavily the structure of the underlying tasks depends on the chosen codomain and then concentrate on the diophantine case $C=\mathbb{Z}$. In particular, we use the Hermite Normal Form to derive structural results and polynomial-time algorithms for general diophantine discrete tomography problems. Then we focus specifically on the polyatomic case.
Alexandrine Morand
Exploring the Neural Substrates of Prospective Memory in Aging: A Structural and Functional Connectivity Investigation
Abstract: Prospective memory, the ability to remember and execute intentions at a future time, is particularly vulnerable to aging, as it relies on controlled attention and executive functions. The first study presents data on time-based prospective memory in healthy younger and older adults, combining behavioral assessments with structural MRI and diffusion imaging. Reduced integrity of specific white matter tracts correlated with prospective memory deficits, emphasizing the role of white matter pathways in executive and monitoring control. Building on these findings, the second study examined resting-state functional connectivity. Age-related reductions in connectivity were observed in relation to prospective memory performance. These functional changes were linked to structural alterations in white matter, suggesting a combined structural-functional substrate for age-related decline. Overall, these findings discuss the neural profiles of aging, including dedifferentiation and compensation, in prospective memory decline.
Marco Pellegrini
Constructing signed magic arrays
Abstract:Let $m,n,s,k$ be four integers such that $1\leqslant s \leqslant n$, $1\leqslant k\leqslant m$ and $ms=nk$. A signed magic array $\mathrm{SMA}(m,n; s,k)$ is an $m\times n$ partially filled array whose entries belong to the subset $\Omega\subset \mathbb{Z}$, where $\Omega=\{0,\pm 1, \pm 2,\ldots, \pm (nk-1)/2\}$ if $nk$ is odd and
$\Omega=\{\pm 1, \pm 2, \ldots, \pm nk/2\}$ if $nk$ is even, satisfying the following requirements:
$(a)$ every $\omega \in \Omega$ appears once in the array;
$(b)$ each row contains exactly $s$ filled cells and each column contains exactly $k$ filled cells;
$(c)$ the sum of the elements in each row and in each column is $0$.
In this talk I will describe some constructions of these objects giving the necessary and sufficient conditions for the existence of an $\mathrm{SMA}(m,n; s,k)$, a problem posed in 2017 by Khodkar, Schulz and Wagner.
