2014 Project Week:HD 4DShapes
- Utah: James Fishbaugh, Guido Gerig
- Iowa: Hans Johnson
In the past we have estimated personalized growth models for a small number of HD subjects. For this project we want to estimated models for 84 subjects: 30 CTRL, 16 LOW, 24 MED, 14 HIGH. We want to explore individual models of sub-cortical change, as well as the evolution of white matter. We also want to test the robustness of the estimation by comparing the results estimated forward and backward in time.
Sub-cortical change in HD:
- Estimate individual models for all subjects and compare volume extracted from regression.
- Compare models estimated forward and backward in time to explore sensitivity to initial conditions.
White matter evolution:
- Prepare white matter surfaces for regression.
- Experiment on one subject as proof of concept.
- Prep all subject data for estimation -- start estimating models, though it likely won't finish during the project week.
Models of subcortical change have been estimated forward and backward in time. Results for most subjects look qualitatively similar. However, for some subjects, the estimated evolution differs depending on the direction of estimation. Perhaps the best solution is to start the middle of the time interval, and estimate a geodesic in both directions.
White matter surfaces have been segmented and prepared for regression. Regression is currently running for all 84 subjects and I will update when the experiments complete. Preliminary results (after only several iterations) show little noticable change. I believe the segementations must be further prepared/cleaned to focus on only the important regions.
Below is an animation of subcortical evolution for a single subject over a 2 year period, estimated forward and backward in time.
We extract the volume of just the caudate (left+right) after shape regression and plot that along with the observed volume measurements taken directly from the data, shown as circles.