Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Clusters in Focus: A Simple and Robust Detail-On-Demand Dashboard for Patient Data

Lukas Schilcher, Peter Waldert, Benedikt Kantz, Tobias Schreck

TL;DR

Clusters in Focus addresses the challenge of comparing data cohort structures across many feature-pair projections by introducing a three-panel interactive dashboard. It projects data onto two features $F_A$ and $F_B$, computes clusters via algorithms like $K$-Means or DBSCAN, and uses a Jaccard similarity $J$ to compare the resulting source cluster $C_{AB}$ with clusters from all other pairs, presented as a ranked List View and a reorderable Matrix View. The key contributions are the cross-pair cluster similarity analysis, an interpretable visualization workflow, and a demonstrated use case on Parkinson's disease speech biomarkers, showing robustness of identified cohorts across related feature families. This approach enables rapid validation and discovery of biomarkers by revealing how cohorts persist across multiple, interpretable two-dimensional subspaces, with practical impact for biomedical data exploration and hypothesis generation.

Abstract

Exploring tabular datasets to understand how different feature pairs partition data into meaningful cohorts is crucial in domains such as biomarker discovery, yet comparing clusters across multiple feature pair projections is challenging. We introduce Clusters in Focus, an interactive visual analytics dashboard designed to address this gap. Clusters in Focus employs a three-panel coordinated view: a Data Panel offers multiple perspectives (tabular, heatmap, condensed with histograms / SHAP values) for initial data exploration; a Selection Panel displays the 2D clustering (K-Means/DBSCAN) for a user-selected feature pair; and a novel Cluster Similarity Panel featuring two switchable views for comparing clusters. A ranked list enables the identification of top-matching feature pairs, while an interactive similarity matrix with reordering capabilities allows for the discovery of global structural patterns and groups of related features. This dual-view design supports both focused querying and broad visual exploration. A use case on a Parkinson's disease speech dataset demonstrates the tool's effectiveness in revealing relationships between different feature pairs characterizing the same patient subgroup.

Clusters in Focus: A Simple and Robust Detail-On-Demand Dashboard for Patient Data

TL;DR

Clusters in Focus addresses the challenge of comparing data cohort structures across many feature-pair projections by introducing a three-panel interactive dashboard. It projects data onto two features and , computes clusters via algorithms like -Means or DBSCAN, and uses a Jaccard similarity to compare the resulting source cluster with clusters from all other pairs, presented as a ranked List View and a reorderable Matrix View. The key contributions are the cross-pair cluster similarity analysis, an interpretable visualization workflow, and a demonstrated use case on Parkinson's disease speech biomarkers, showing robustness of identified cohorts across related feature families. This approach enables rapid validation and discovery of biomarkers by revealing how cohorts persist across multiple, interpretable two-dimensional subspaces, with practical impact for biomedical data exploration and hypothesis generation.

Abstract

Exploring tabular datasets to understand how different feature pairs partition data into meaningful cohorts is crucial in domains such as biomarker discovery, yet comparing clusters across multiple feature pair projections is challenging. We introduce Clusters in Focus, an interactive visual analytics dashboard designed to address this gap. Clusters in Focus employs a three-panel coordinated view: a Data Panel offers multiple perspectives (tabular, heatmap, condensed with histograms / SHAP values) for initial data exploration; a Selection Panel displays the 2D clustering (K-Means/DBSCAN) for a user-selected feature pair; and a novel Cluster Similarity Panel featuring two switchable views for comparing clusters. A ranked list enables the identification of top-matching feature pairs, while an interactive similarity matrix with reordering capabilities allows for the discovery of global structural patterns and groups of related features. This dual-view design supports both focused querying and broad visual exploration. A use case on a Parkinson's disease speech dataset demonstrates the tool's effectiveness in revealing relationships between different feature pairs characterizing the same patient subgroup.
Paper Structure (9 sections, 2 figures)

This paper contains 9 sections, 2 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. The Clusters in Focus -Tool
  4. Panel 1: Data & Feature Exploration
  5. Panel 2: Focused 2D Clustering
  6. Panel 3: Cross-Pair Cluster Similarity
  7. Use Case: Parkinson's Disease Speech Biomarkers
  8. Discussion
  9. Conclusion

Figures (2)

  • Figure 1: Panel 1 views in Clusters in Focus: (a) Heatmap and (b) Condensed feature summary.
  • Figure 2: The Cluster Similarity Panel in Matrix View configuration. The panel provides two switchable views (top right): a List View for targeted ranking and this Matrix View for global pattern discovery. Each cell's color represents the aggregated Jaccard similarity, with the aggregation method (Maximum) and matrix ordering (Optimal Leaf Ordering) selected by the user. The reordering algorithm groups features with similar profiles, revealing a distinct high-similarity block around the source features and exposing, through the bright horizontal and vertical lines, their strong relationship with other feature groups related to vocal stability (e.g., HNR, DFA) and jitter (Jitter:DDP).