Replicability of large dataset data-driven results in smaller local datasets
An example of ICA network of Parkinson disease
The accumulation of large open datasets of different clinical cohorts (like PPMI) will enable the researchers to test their hypothesis with more powerful statistical inferences. The replicability of these large datasets results in smaller local datasets is a fundamental question. In this research, we focus on a very popular method in neuroimaging community called ICA, and test its replicability of the results from this method with both PPMI dataset (about 250 PD subjects and 120 NC subjects) and our local Abbas dataset (40 PD patients, 29 ET patients and 33 NC subjects). The scientific question of this research will be:
-
What the ICs derived from the small local dataset looks like and what are the differences and correlations between the the PD-ICA from our own dataset, we will try to interpret these results with Abbas and Meshal.
-
We will try to generate ICAs with the same number of local Abbas dataset PD/NC subjects by resampling the PPMI dataset and analyze the differences.
-
We will try to generate the ICAs by deforming to disease specific template and normal healthy template, this will help us to understand the significance of the disease cohort specific template.
People
