Accurate Interactive Visualization of Large Deformations and Variability in Biomedical Image Ensembles

Area of Science:

• Medical image analysis
• Computer graphics
• Computational biology

Background:

• Large deformations in 3D image ensembles challenge current visualization and statistical analysis methods in biology and medicine.
• Linear interpolation introduces artifacts, hindering accurate shape analysis.

Purpose of the Study:

• To develop an efficient, interactive non-linear image interpolation and extrapolation method for high-quality rendering of large deformations.
• To enable novel interactive methods for shape ensemble analysis.

Main Methods:

• Utilized the theory of stationary velocity fields for non-linear image interpolation and extrapolation.
• Developed an efficient GPU-based image warping method.
• Showcased four novel visualizations: group mean shapes, interactive reformation, likelihood volumes, and streamline visualization.

Main Results:

• The GPU-accelerated non-linear warping method significantly improves visual quality and retains interactive frame rates compared to state-of-the-art methods.
• Demonstrated effectiveness on a real-world dataset and in a case study with rodent skull and mandible data.

Conclusions:

• The developed method enhances existing visualization techniques and enables new interactive shape ensemble analysis tools.
• The visualizations aid in distinguishing normal vs. pathological morphology, correlating shape with extrinsic factors, and assessing model quality.