DPVis: Visual Analytics With Hidden Markov Models for Disease Progression Pathways.

Area of Science:

• Computational biology
• Medical informatics
• Data visualization

Background:

• Disease progression models are crucial for understanding patient status from longitudinal health records.
• Hidden Markov Models (HMMs) identify distinct health states but present interpretation challenges for clinicians.
• Interpreting complex model parameters and clinical patterns remains a significant hurdle in disease progression analysis.

Purpose of the Study:

• To address the interpretability challenges of HMMs in disease progression modeling.
• To develop an integrated visualization tool for analyzing HMM outputs in chronic diseases.
• To facilitate clinical understanding and exploration of disease progression pathways.

Main Methods:

• A design study involving clinical scientists, statisticians, and visualization experts.
• Development of DPVis, a system integrating HMM parameters and outcomes into interactive visualizations.
• Application of DPVis to investigate disease progression in Type 1 Diabetes, Huntington's disease, Parkinson's disease, and COPD.

Main Results:

• DPVis successfully integrates HMM parameters and outcomes into interpretable visualizations.
• The tool enables visual summarization of disease states and interactive exploration of progression patterns.
• DPVis facilitates the analysis and comparison of clinically relevant patient subgroups.

Conclusions:

• DPVis enhances the evaluation and clinical interpretation of disease progression models.
• Interactive visualizations improve understanding of complex HMM outputs for chronic diseases.
• The tool supports clinical decision-making by providing clearer insights into patient health trajectories.