Multi-scale visual analysis of time-varying electrocorticography data via clustering of brain regions

Sugeerth Murugesan, Kristofer E. Bouchard, Edward F. Chang, Max Dougherty, Bernd Hamann, Gunther H. Weber

TL;DR

This paper introduces ECoG ClusterFlow, a novel interactive visualization system designed to analyze complex, time-varying electrocorticography (ECoG) data. By integrating multi-scale views and advanced clustering techniques, ECoG ClusterFlow helps neuroscientists identify dynamic patterns in brain activity, crucial for understanding neurological disorders like epilepsy.

What Problem Does This Paper Address?

Analyzing brain activity data, particularly from electrocorticography (ECoG), presents a significant challenge due to its high spatio-temporal resolution and dynamic nature. ECoG data captures electrical signals directly from the brain's surface, revealing intricate patterns of functional connectivity that change over time. Understanding these changes is vital for diagnosing neurological diseases, such as epilepsy, and for developing effective treatments. Existing tools often struggle to handle the multi-scale nature of this data, making it difficult to simultaneously grasp both broad trends and fine-grained details. This gap hinders the ability of researchers to effectively identify seizure origins, track their spread, and understand the underlying neural mechanisms.

Methods

ECoG ClusterFlow employs a hierarchical, multi-scale approach to visualize dynamic cluster evolution in ECoG data. The system's core is a two-stage clustering process: first, it detects communities (clusters) at each time step, and then it tracks these communities over time. A key innovation is the "Cluster Evolution View," which uses an alluvial diagram to represent how clusters emerge, merge, split, and disappear over time. This view allows users to see the overall flow of community changes.

Complementing this is the "Electrode View," which places interactive glyphs representing electrodes onto a 3D brain model. These glyphs visualize cluster membership and electrical activity, providing spatial context. The system supports multiple levels of detail, allowing users to aggregate time steps to see broader trends or examine individual time steps for fine-grained analysis. This "overview first, then details on demand" approach, along with focus+context and highlighting/linking techniques, aims to reduce cognitive load and facilitate intuitive exploration.

Datasets

The paper utilizes two types of datasets for evaluation:

  • Synthetic Dataset: A custom-generated dataset with 54 electrodes and controlled activation values across 30 time steps. This dataset was designed to have known patterns, allowing for a thorough evaluation of the visualization method in a well-understood scenario.
  • Epileptic Seizure Dataset: Real-world ECoG data from patients with epilepsy. This dataset was pre-processed into two forms: a high-gamma dataset (70-170 Hz) and a full-range dataset (averaging all frequencies). These datasets were used to demonstrate the system's effectiveness in analyzing complex, real-world neurological data.

Key Insights and Ideas

  • Multi-scale visualization is crucial: The ability to switch between coarse-grained overviews and fine-grained details is essential for understanding complex spatio-temporal brain data.
  • Visualizing cluster evolution aids discovery: The alluvial diagram in the Cluster Evolution View provides an intuitive way to track dynamic changes in brain network communities, revealing patterns that might be missed by static or purely animated methods.
  • Spatial context is key: Integrating cluster information with a spatial representation of the brain (Electrode View) allows neuroscientists to pinpoint the location of dynamic changes and correlate them with anatomical regions.
  • Interactive exploration enhances usability: Features like brushing, linking, and adjustable granularity empower users to tailor the analysis to their specific research questions and cognitive preferences.
  • Bridging the gap between visualization and neuroscience: The system was developed in close collaboration with neuroscientists, ensuring that the design choices and features directly address their analytical needs and domain questions.

Key Results

The paper demonstrates the effectiveness of ECoG ClusterFlow through two case studies:

  • Synthetic Data: The system successfully identified known patterns, including stable cluster configurations, random distributions, and emergent focal activation points, validating the visualization's ability to represent underlying data structures.
  • Epileptic Seizure Data: Neuroscientists were able to identify distinct brain states (before-seizure, early-seizure, mid-seizure, late-seizure) and pinpoint seizure initiation zones. They observed that the high-gamma dataset often showed clearer community structures compared to the full-range dataset, and the system facilitated the comparison of these patterns. The tool was praised for its intuitiveness and ability to reveal relevant insights into seizure genesis and propagation.

Limitations and Future Work

  • Scalability: While the system can handle up to 64 time steps and 256 electrodes, scaling to much larger datasets (e.g., >500 time steps) in real-time remains a challenge. Future work will focus on developing more advanced data reduction and abstraction techniques.
  • Comparison of visualizations: The authors plan to conduct further evaluations on the usefulness of spatio-temporal clustering and how to consistently compare visualizations across different datasets.
  • Automated pattern identification: While the system facilitates visual exploration, future work could explore more automated methods for identifying significant spatio-temporal patterns.

Summary updated

Abstract

Background There exists a need for effective and easy-to-use software tools supporting the analysis of complex Electrocorticography (ECoG) data. Understanding how epileptic seizures develop or identifying diagnostic indicators for neurological diseases require the in-depth analysis of neural activity data from ECoG. Such data is multi-scale and is of high spatio-temporal resolution. Comprehensive analysis of this data should be supported by interactive visual analysis methods that allow a scientist to understand functional patterns at varying levels of granularity and comprehend its time-varying behavior. Results We introduce a novel multi-scale visual analysis system, ECoG ClusterFlow, for the detailed exploration of ECoG data. Our system detects and visualizes dynamic high-level structures, such as communities, derived from the time-varying connectivity network. The system supports two major views: 1) an overview summarizing the evolution of clusters over time and 2) an electrode view using hierarchical glyph-based design to visualize the propagation of clusters in their spatial, anatomical context. We present case studies that were performed in collaboration with neuroscientists and neurosurgeons using simulated and recorded epileptic seizure data to demonstrate our system’s effectiveness. Conclusion ECoG ClusterFlow supports the comparison of spatio-temporal patterns for specific time intervals and allows a user to utilize various clustering algorithms. Neuroscientists can identify the site of seizure genesis and its spatial progression during various the stages of a seizure. Our system serves as a fast and powerful means for the generation of preliminary hypotheses that can be used as a basis for subsequent application of rigorous statistical methods, with the ultimate goal being the clinical treatment of epileptogenic zones.
BibTeX
@article{murugesan2017multi,
  author = {Sugeerth Murugesan and Kristofer E. Bouchard and Edward F. Chang and Max Dougherty and Bernd Hamann and Gunther H. Weber},
  title = {{Multi-scale visual analysis of time-varying electrocorticography data via clustering of brain regions}},
  journal = {BMC Bioinformatics},
  year = {2017},
  doi = {10.1186/s12859-017-1633-9},
}
Figure