Challenges in population-based brain mapping
Imaging studies of clinical populations continue to uncover new patterns of altered structure and function, and novel algorithms are being applied to relate these patterns to cognitive and genetic parameters. Post-mortem brain maps are also beginning to clarify the molecular substrates of disease.
As imaging studies expand into ever-larger patient populations, population-based brain atlases (Mazziotta et al., 1995; Thompson et al., 2000a,b) offer a powerful framework to synthesize results from disparate imaging studies. These atlases use novel analytical tools to fuse data across subjects, modalities, and time. They detect group-specific features not apparent in individual patients' scans. Once built, these atlases can be stratified into subpopulations to reflect a particular clinical group, such as individuals at genetic risk for AD, patients with mild cognitive impairment (MCI) or different dementia subtypes (frontotemporal dementia/semantic dementia), or patients undergoing different drug treatments. The disease-specific features these atlases resolve can then be linked with demographic factors such as age, gender, handedness, as well as specific clinical or genetic parameters (Mazziotta et al., 1995; Toga & Mazziotta, 1996; Thompson et al., 2001 a–e).
New brain atlases are also being built to incorporate dynamic data (Thompson et al., 2002). Despite the significant challenges in expanding the atlas concept to the time dimension, dynamic brain atlases are beginning to include probabilistic information on growth rates that may assist research into pediatric disorders (Thompson et al., 2000a,b) as well as revealing patterns of degenerative rates in Alzheimer's disease (Fox et al., 1996; Thompson et al., 2001a–e, 2002; Chan et al., 2001). Imaging algorithms are also significantly improving the flexibility of digital brain templates.