PERFORM Colloquium: Can neuroinformatics advances solve the biomedical reproducibility crisis?
In this talk, I will first review some evidence for reproducibility issues by taking examples in the fields of biomedicine, neuroimaging and imaging genetics. I will then discuss the most likely causes for irreproducibility, focusing on statistical and informatics aspects of the problem in the context of recent controversies about false positive rates. As new neuroinformatics tools and standards emerge, I will consider to what extent these new tools may be able to help us curb the reproducibility issues and improve the question of research efficiency. I will last consider the sociological factors and our publishing research culture as potential factors and reflect on what could help the biomedical community make their research culture evolve to a greater efficiency.
Since 1990, I have worked on the development of methods for the analysis of functional imaging data (mostly fMRI), and more specifically in the statistical modeling and inference aspects. While a post-doctoral fellow in London (1994-1996), I co-authored SPM, the most popular fMRI statistical analysis software, and developed novel data analysis techniques. I maintain close interactions with neuroscientists to ensure that analysis methods are answering actual needs. I have taught neuroimaging data analysis courses and was elected twice educational chair of the organization for Human Brain Mapping. During the past five years, I have developed an interest in neuroimaging network analyses and imaging genetics and was responsible for a large multi-centric neuroimaging genetic database. I believe that neuroinformatics is a fundamental part of neuroimaging, and I chair the neuroimaging data sharing task force of the International Neuroinformatics Coordinating Facility. I am the co-editor in chief of Frontiers in Brain Imaging Methods, which I co-founded in 2012 partly to address the need for reproducibility and the development of methods in the brain imaging field. Currently, I am developing methods for neuroimaging data as well as teaching in Python and participate in the NIH funded “ReproNim” project.