Fractal geometry predicts dynamic differences in structural and functional connectomes
Anca Rădulescu, Eva Kaslik, Alexandru Fikl, Johan Nakuci, Sarah Muldoon, Michael Anderson · 2025
Read Fractal geometry predicts dynamic differences in structural and functional connectomes by Anca Rădulescu, Eva Kaslik, Alexandru Fikl, Johan Nakuci, Sarah Muldoon, Michael Anderson (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-11477.txt.
Introduction
Anca Rădulescu and five colleagues wrote a 2025 paper on a new way to read brain networks. A connectome maps how brain areas link to each other, in both structure and function. Standard graph tools often miss how these networks behave as a whole over time.
The team instead uses fractal geometry, the math of rough, self-similar shapes. They apply ideas from Mandelbrot-like sets and repeated quadratic maps to the network data. Clear gaps appear between maps of brain structure and maps of brain activity. One gap is a flip in the cusp of the shapes they plot. The method may give fresh clues about thought and brain disease.
Catalog record
| CREATOR | Anca Rădulescu, Eva Kaslik, Alexandru Fikl, Johan Nakuci, Sarah Muldoon, Michael Anderson |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 77 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-11477
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Fractal geometry predicts dynamic differences in structural and functional connectomes Anca Rădulescu, Eva Kaslik, Alexandru Fikl, Johan Nakuci, Sarah Muldoon, Michael Anderson Abstract Understanding the intricate architecture of brain networks and its connection to brain function is essential for deciphering the underlying principles of cognition and disease. While, traditional graph-theoretical measures have been widely used to characterize these networks, they often fail to fully capture the emergent properties of large-scale neural dynamics. Here, we introduce an alternative approach to quantify brain networks that is rooted in complex dynamics, fractal geometry, and asymptotic analysis. We apply these concepts to brain connectomes and demonstrate how quadratic iterations and geometric properties of Mandelbrot-like sets can provide novel insights into structural and functional network dynamics. Our findings reveal fundamental distinctions between structural (positive) and functional (signed) connectomes, such as the shift of cusp orientation and the variability in equi-M set geometry. Notably, structural connectomes exhibit more robust, predictable features, while functional connectomes show increased variability for non-trivial tasks. We further demonstrate that traditional graph-theoretical measures, when applied separately to the positive and negative sub-networks of functional connectomes, fail to fully capture their dynamic complexity. Instead, size and shape-based invariants of the equi-M set effectively differentiate between rest and emotional task states, which highlights their potential as superior markers of emergent network dynamics. These results suggest that incorporating fractal-based methods into network neuroscience provides a powerful tool for understanding how information flows in natural systems beyond static connectivity measures, while maintaining their simplicity. keywords Keyword1, Keyword2, Keyword3, Keyword4 1 Introduction Fractal research transformed how we understand complexity, revealing how simple rules can generate intricate, emergent behavior. The Mandelbrot set – created by iterating a basic quadratic function – became an iconic example, illustrating self-similarity and nonlinear feedback in complex systems. Fractal geometry has long been recognized as a powerful framework for describing self-similar structures in nature, from vascular branching gabrys2005fractal ; jayalalitha2008fractal to neuronal morphology karperien2016morphology ; karperien2024morphology . Building on this legacy, we previously developed a generalizable framework for analyzing dynamics in complex networks. At its core is the equi-M set, a new object that extends classical complex dynamics into high-dimensional, evolving systems. We explore the role of equi-M sets as a tool for characterizing the global organization of brain networks. By mapping the asymptotic behavior of iterated quadratic dynamics onto empirical connectomes, we generate novel descriptors that integrate connectivity structure with properties of the emerging dynamics. Unlike traditional network analysis, which emphasizes static features such as node degree or clustering, our approach captures how behavior emerges over time through interactions across the network. Structural and Functional brain connectomes exhibit complex topologies that reflect the brain’s capacity for information processing rezaeinia2020identifying , adaptability stampanoni2019synaptic , and resilience puxeddu2024relation . Applying this methodology to Structural and Functional connectomes derived from the Human Connectome Project, we uncover key differences in the topological and geometric features of equi-M sets. Structural connectomes, which are inherently symmetric and positive, exhibit equi-M sets with robust and consistent cusp-like features that reflect unifying features in network organization. In contrast, Functional connectomes, shaped by the interplay of excitatory and inhibitory interactions, display more diverse and task-dependent fractal structures. Notably, equi-M sets derived from resting-state and emotion-processing Functional networks differ noticeably in their geometry. This suggests that task engagement drives greater dynamical diversity across individuals. Our findings indicate that fractal dynamics offer a novel lens for studying brain organization that bridges the gap between network topology and emergent neural computation. By leveraging equi-M sets as network dynamics assessment tools, we open new avenues for the dynamical classification of brain states. 1.1 Complex quadratic dynamics Historically, the study of single map complex iterations emerged at the beginning of the twentieth century, with the work of Pierre Fatou fatou1920equations and Gaston Julia julia1918memoire . It describes (by means of very simple iterative rules) rich, complex systems that produce objects with intricate fractal geometry. Despite the seemingly elementary and natural structure of such systems, investigating them required the development of new areas of mathematics, encompassing real and complex analysis, topology, and computational methods. However, even as important progress has been made, many questions remain open devaney2007open . Although emerging areas of dynamics have been widely applied to understanding nonlinear and chaotic phenomena in various natural systems, the application of complex dynamics has yet to see widespread exploration in many domains, including neuroscience. Discrete dynamics of single iterated transformations in the complex plane have developed over a few decades into a rich research field carleson1996complex . In particular, quadratic iterations represent a simple, yet incredibly rich, family of examples. They have first been studied on the real line, where they are associated with cascades of period-doubling bifurcations and universality. In turn, complex quadratic iterations in the family f_{c}(z)=z^{2}+c , with c\in\mathbb{C} , provide textbook recipes for creating asymptotic fractal sets in the complex plane. For iterations of a single quadratic map, one of the most striking results is that the trajectory (or orbit) of the critical point z=0 under iterations of the map captures global information on the behavior of all other orbits. The property of the critical orbit being bounded is particularly important in the context of long-term iterations. The Mandelbot set appeared in the 1960s from the need to obtain an overarching description of this property over the whole family of quadratic maps. It is defined as the set of all c\in\mathbb{C} for which the critical orbit of the complex quadratic function f_{c}(z)=z^{2}+c is bounded. This is a fractal with fascinating characteristics that have been extensively analyzed since its introduction. The topological organization of the Mandelbrot set into hyperbolic bulbs (leading to its shape and fractal geometry) classifies all possible combinatoric behaviors that orbits can have in the family f_{c} . This makes the Mandelbrot set both a staple in discrete dynamical systems and a canonical representation for more general phenomena. 1.2 Complex quadratic maps on networks Our work extends this framework from its original scope of single map iterations to the broader context of complex dynamical networks (CGNs). More specifically, we consider networks in which the coupled nodes are identical complex quadratic maps, with coupling specified by the binary adjacency matrix A\in\mathbb{R}^{n\times n} and weights g\in\mathbb{R}^{n\times n} . The overall evolution of this discrete system takes the form of an iteration in \mathbb{C}^{n} , which acts in each component as z_{k}\mapsto\left(\sum_{j=1}^{n}A_{kj}g_{kj}z_{j}\right)^{2}+c. (1) Since nodes have the same intrinsic dynamics dictated by the parameter c , the emerging network dynamics are a direct reflection of the value of c and of the specific (A,g) network architecture at play. By analogy to the single map case, we define the equi-M set of the network as the set of all c\in\mathbb{C} for which the critical point 0\in\mathbb{C}^{n} remains bounded under iteration. In this generalized context, many of the results on the properties and structure of the traditional Mandelbrot set are lost. Our theoretical work shows that the equi-M set no longer provides a full combinatorial description of the asymptotic behavior of the map radulescu2023complex . However, our results strongly suggest that equi-M sets have quantifiable topological and fractal properties that reflect the network architecture, making them good candidates for testing and classifying how dynamic patterns evolve for