|
Abstract:
|
The neuroimaging field including technical advances for data collection due to scanner upgrades has made tremendous progress over the last 2 decades. New statistical techniques have been implemented in recent years, such as RAVEL, COMBAT and some variations of these, to address the challenges introduced by scanner variability, multiple data collection sites and other sources. However, some challenges still remain, such as: longitudinal multisite harmonization of multiple imaging modalities and different scanners, differences between normalization and harmonization techniques and methodology development to address these challenges remains to be tackled by the statisticians and others in the years ahead. In this roundtable I propose to discuss the importance of these developments as well as the differences between normalization and harmonization methods (RAVEL versus COMBAT) and how the downstream analyses may be affected if such sources of variability are not properly addressed. In addition, I will discuss why clinical and methodological grant proposals that have different focus than addressing these challenges could be potentially impacted at their analysis and inference level.
|
