Topological Analysis of Mouse Brain Vasculature via 3D Light-sheet Microscopy Images

Jiachen Yao, Nina Hagemann, Qiaojie Xiong, Jianxu Chen, Dirk M. Hermann, Chao Chen

Abstract

Vascular networks play a crucial role in understanding brain functionalities. Brain integrity and function, neuronal activity and plasticity, which are crucial for learning, are actively modulated by their local environments, specifically vascular networks. With recent developments in high-resolution 3D light-sheet microscopy imaging together with tissue processing techniques, it becomes feasible to obtain and examine large-scale brain vasculature in mice. To establish a structural foundation for functional study, however, we need advanced image analysis and structural modeling methods. Existing works use geometric features such as thickness, tortuosity, etc. However, geometric features cannot fully capture structural characteristics such as the richness of branches, connectivity, etc. In this paper, we study the morphology of brain vasculature through a topological lens. We extract topological features based on the theory of topological data analysis. Comparing of these robust and multi-scale topological structural features across different brain anatomical structures and between normal and obese populations sheds light on their promising future in studying neurological diseases.

Introduction

The brain is an exceptionally energy-intensive organ, accounting for approximately 20% of the body's energy consumption despite representing only 2% of its total weight. This energy is supplied through a vast and intricate vascular network. Disruptions to this network are critically linked to various neurological conditions, including stroke, Alzheimer’s disease, and cognitive deficits. Understanding the structural changes associated with these conditions is paramount for developing diagnostic and therapeutic strategies.

Historically, vascular analysis has relied on local geometric features such as vessel diameter or tortuosity. While useful, these metrics fail to capture the global connectivity and branching complexity of the network. This study proposes a topological approach, utilizing Topological Data Analysis (TDA) to provide a more robust, multi-scale representation of the brain's "wiring diagram."

Methods

The researchers employed Persistent Homology (PH), a mathematical framework that characterizes the "shape" of data by tracking topological features across multiple scales.

  1. Persistent Homology (PH): This technique identifies connected components (0D), loops (1D), and voids (2D). It observes how these features appear and disappear as a spatial parameter is varied, effectively measuring the persistence of structural holes.
  2. Persistent Homology Transform (PHT): To ensure a comprehensive analysis, PH was applied from multiple viewpoints and angles, creating a multi-directional "topological signature" of the vasculature.
  3. Betti Curves: The complex topological data was summarized into Betti Curves. These curves represent the number of topological features at each scale, transforming abstract shapes into 1D vectors suitable for statistical comparison.
  4. Statistical Analysis: The study utilized t-tests and Maximum Mean Discrepancy (MMD) to evaluate the statistical significance of structural differences between healthy and obese populations.

Datasets and Regional Analysis

The study utilized a public dataset of 3D light-sheet fluorescence microscopy (LSFM) images from 9-12 week old male mice. The analysis compared healthy (wild-type) mice against obese (leptin-deficient ob/ob) mice across four distinct brain regions: the Cortex, Hippocampus, Striatum, and Midbrain.

Vascular networks across four brain regions: Cortex, Hippocampus, Striatum, and Midbrain.

The dataset included approximately 12 healthy and 8–10 obese samples per brain region, providing a diverse set of vascular architectures for topological characterization.

Key Results

The most significant finding of this research is that topological analysis reveals structural changes that are invisible to traditional geometric methods. While previous geometric studies identified differences in the cortex and striatum of obese mice, they failed to detect changes in other critical areas such as the hippocampus and midbrain.

Vascular structures of the Hippocampus (top) and Midbrain (bottom). Topological analysis successfully identified significant structural differences in these regions that geometric measures missed.
  • Hippocampus: Topological analysis identified a significant difference (p = 0.04) in mean feature values between the healthy and obese groups.
  • Midbrain: Using the MMD statistical test, researchers found a significant difference (p = 0.049) in vascular structure.
  • Striatum Complexity: The study confirmed that the striatum possesses the most complex vascular network, reflected in high 1D Betti numbers (representing loops).

Discussion and Future Work

Topological features offer a robust alternative to geometric measures, as they are less sensitive to local imaging noise and artifacts. By focusing on connectivity rather than just vessel size, TDA provides a complementary perspective on brain health.

However, several limitations remain. The sample size, while typical for high-resolution 3D imaging, is relatively small. Furthermore, the current model does not distinguish between arteries and veins, which serve distinct physiological roles. Future research should aim to incorporate vessel classification and investigate the biological mechanisms driving these topological shifts, particularly in the context of neuroregeneration and recovery from ischemia.

Summary updated

Abstract

Vascular networks play a crucial role in understanding brain functionalities. Brain integrity and function, neuronal activity and plasticity, which are crucial for learning, are actively modulated by their local environments, specifically vascular networks. With recent developments in high-resolution 3D light-sheet microscopy imaging together with tissue processing techniques, it becomes feasible to obtain and examine large-scale brain vasculature in mice. To establish a structural foundation for functional study, however, we need advanced image analysis and structural modeling methods. Existing works use geometric features such as thickness, tortuosity, etc. However, geometric features cannot fully capture structural characteristics such as the richness of branches, connectivity, etc. In this paper, we study the morphology of brain vasculature through a topological lens. We extract topological features based on the theory of topological data analysis. Comparing of these robust and multi-scale topological structural features across different brain anatomical structures and between normal and obese populations sheds light on their promising future in studying neurological diseases.
BibTeX
@inproceedings{yao2024topological,
  author = {Jiachen Yao and Nina Hagemann and Qiaojie Xiong and Jianxu Chen and Dirk M. Hermann and Chao Chen},
  title = {{Topological Analysis of Mouse Brain Vasculature via 3D Light-sheet Microscopy Images}},
  booktitle = {ISBI},
  year = {2024},
  url = {https://openreview.net/forum?id=mNdP0Yq5KJ},
}
Figure