Cross-disease Brain Image Modeling
Biomarkers Across Neurodegenerative Diseases, 2016
Individuals with Alzheimerís and Parkinsonís diseases share typical brain image patterns. The research field has long awaited an intelligent system that could interpret patterns associated with different types of neurodegenerative diseases.
We hypothesize that Gaussian Process Regression (a statistical model) can be used to compare brain images and identify common features characterizing diseases.
Recently, my lab developed a model that won top place in the recent DREAM Alzheimerís Disease Big Data Challenge, which aims to use brain images for disease diagnosis and establish a genetic model for predicting clinical outcomes. This competition, the most influential in the systems biology field, included groups that had over a hundred publications in brain image analysis. Our success in the challenge was due to development of our transformative feature identification and integrative method (customized Gaussian Process Regression) in analyzing brain images. In this proposed work, we will use this technique to investigate and compare brain images retrieved from the Alzheimerís Disease Neuroimaging Initiative and the Parkinsonís Progression Markers Initiative.
Impact on Diagnosis/Treatment of Parkinsonís Disease:
Our strategy aims to identify a series of brain regions that are predictive of clinical phenotypes and reveal shared and unique characteristics of the two diseases.
Next Steps for Development:
Our model may help quickly diagnosis Alzheimerís or Parkinsonís based on brain imaging.
Assistant Professor at University of Michigan
Location: Ann Arbor, Michigan, United States