Mbirn: Scientific questions for later consideration

2005.05.17 and 2005.06.21 (C. Fennema-Notestine/R. Gollub)

Future questions to address when analysis pipleine is completed and statistical methods to compare these effects are developed:

1. Atlas used in the Spring of 2005 analysis of UCSD, MGH/BWH, WashU data is based on 40 subjects from Wash U (10 Young Normal Controls, 10 Middle aged NC, 10 Elderly NC, and 10 Alzheimers Disease). What are the performance consequences of changing the composition of the atlas?

• What are the performance consequences of changing the composition of the diagnostic groups of the atlas? For example, given that the MAD study is examining only data from ENC and AD groups, essentially, what is the consequence of using an atlas that contains only ENC and AD data?
• What are the consequences of changing the composition of the acquisition scan sequence

2. For any given set of biological effects in a dataset, which effects might be easily detectable with relatively crude levels of processing (e.g., large enough effect) and which might require more finely detailed processing? For example, given an automated application of subcortical segmentation, are effects of interest robust enough? Or might reliable, rule-based editing of the automated outcome enhance the ability to detect effects of interest?

---