Statistical Challenges for Clinical Trials with Neuroimaging
David C. Zhu, Department of Radiology, Michigan State University, USA
Clinical trials with neuroimaging often collect data from various techniques of MRI (magnetic resonance imaging), various techniques of PET (positron emission tomography), genetics, cerebrospinal fluid and blood biomarkers, cognitive tests and clinical measurements. Ideally, all data are used to define the condition of a brain. However, these data are complex with multiple dimensions and multiple image modalities, and thus pose a true big data challenge. Researchers often focus on one data type or incorporate two or three modalities together to investigate the brain and how the brain function and structure change over time. In this talk, I will provide an overview of the data collected from the large multi-site ADNI (Alzheimer's Disease Neuroimaging Initiative, http://adni.loni.usc.edu) project and the multi-site rrAD (Risk Reduction for Alzheimer’s Disease, http://www.rradtrial.org) project. The ADNI project collects brain data from older subjects (normal, mild cognitive impairment (MCI) and Alzheimer’s disease (AD)) over multiple time points with intervals of months and years to observe the brain modification. The rrAD project is an intervention study to understand whether aerobic exercise and intensive medical management of blood pressure and cholesterol can reduce the risk of AD in older adults who are at risk of AD. From an imaging researcher perspective, I will also introduce a method of incorporating image data collected from various modalities, along with genetics, cognitive tests and clinical data. Nevertheless, vigorous statistical integration of big data remains a challenge.