The principal mission of the Core is to help NIH researchers with analyses of their fMRI (brain activation mapping) data. Along the way, we also help non-NIH investigators, mostly in the USA but also some abroad. Several levels of help are provided, from short-term immediate aid to long-term development and planning. Consultations: The shortest term help comprises in-person consultations with investigators about issues that arise in their research. The issues involved are quite varied, since there are many steps in carrying out fMRI data analyses and there are many different types of experiments. Common problems include: - How to set up experimental design so that data can be analyzed effectively? - Interpretation and correction of MRI imaging artifacts (for example: subject head motion during scanning). - How to set up time series analysis to extract brain activation effects of interest, and to suppress non-activation artifacts? - Why don't AFNI results agree with SPM/FSL/something else? - How to analyze data to reveal connections between brain regions during specific mental tasks, or at rest? - How to recognize poor quality data? - How to carry out reliable inter-patient (group) statistical analysis, especially when non-MRI data (e.g., genetic information, age, disease rating) needs to be incorporated? - How to get good registration between the functional results and the anatomical reference images, and between the brain images from different subjects? - What sequence of programs is best for analyzing a particular kind of data? - Reports of real or imagined bugs in the AFNI software, as well as feature requests (small, large, and extravagant). There are familiar themes in many of these consultations, but each meeting and each experiment raises unique questions and usually requires digging into the goals and details of the research project in order to ensure that nothing central is being overlooked. The first question asked by a user is often not the right question to answer. Complex statistical issues are often raised. Often, software needs to be developed or modified to help researchers answer their specific questions. Helping with the Methods sections of papers, or with responses to reviewers, is often part of our duties. Educational Efforts: The Core developed (and updates) a 40-hour hands-on course on how to design and analyze fMRI data that was taught twice at the NIH during FY 2017 to over 250 students. All material for this continually evolving course (software, sample data, scripts, and PDF slides) are freely available on our Web site (https://afni.nimh.nih.gov). The course material includes several sample datasets that are used to illustrate the entire process, starting with images output by MRI scanners and continuing through to the collective statistical analysis of groups of subjects. By invitation, we also taught versions of this course at 5 non-NIH sites (expenses for these trips were sponsored by the hosts): Washington U (Missouri), Vanderbilt U (Tennessee), U Chieti-Pescara (Italy), U Nebraska, and UCSD. More than 1200 AFNI forum postings were made by Core members, mostly in answer to queries from users. Algorithm and Software Development: The longest-term support consists of developing new methods and software for fMRI data analysis, both to solve current problems and in anticipation of new needs. All of our software is incorporated into the AFNI package, which is Unix/Linux/Macintosh-based open-source and is available for download by anyone in source code (GitHub) or binary formats (Core server). New programs are created, and old programs modified, in response to specific user requests and in response to the Core's vision of what will be needed in the future. The Core also assists NIH labs in setting up computer systems for use with AFNI, and maintains an active Web site with a forum for questions (and answers) about fMRI data analysis. Notable developments during FY 2017 include: - A significant controversy erupted in mid-2016 over statistical accuracy in fMRI group studies, after a paper in PNAS was published calling into question over 40,000 papers in this field (that specific claim was later rescinded by the authors). The Core expended significant effort to assess the magnitude of the issues raised in this paper and then fix the problems that we identified. Two new methods were developed to deal with this issue, both of which control false detections at desired level. The first method was presented in two papers (1, 2), and another paper is currently being written about the second more complex method. This newer method is more intricate, but reduces the number of arbitrary choices required of the researcher, minimizing the possibility of p-hacking (searching for desirable results by playing with the analysis parameter choices). A two-hour presentation on the newer method was given to a packed room of NIH AFNI users, and the slides made available on the Web. - We provided assistance to a visiting developer from the ENIGMA group, to enable them to modify their resting state FMRI data analysis to use AFNI. Other open-source FMRI analysis software packages restrict their usage for any commercial applications, which made it impossible for the ENIGMA group to collaborate with industrial researchers. - Core programmers aided and participated in 2 workshops at the NIH: BrainHack and the Open and Reproducible Science using Python meeting. - Core scientists presented 4 posters at the Human Brain Mapping meeting in Vancouver BC in June. - Work continues in collaboration with other NIH group on brain template development, including one for macaques (now available), one for marmosets, and one for human toddlers. - Improvements were made to the skull stripping (removal of the non-brain parts of 3D images) methods in AFNI, in combination with improved methods for aligning brain image datasets -- including implementation of nonlinear warping for improving alignment of functional brain images to structural brain images. Public Health Impact: From Oct 2016 to Aug 2017, the principal AFNI publication has been cited in 385 papers (cf Scopus). Most of our work supports basic research into brain function, but some of our work is more closely tied to or applicable to specific diseases: - We collaborate with Dr. Alex Martin (NIMH) to apply our resting state analysis methods to autism spectrum disorder. - We consult frequently with NIMH researchers (e.g., Drs. Pine, Ernst, Grillon, Leibenluft) working in mood and anxiety disorders. - We consult with Dr. Elliot Stein (NIDA) in his research applying fMRI methods to drug abuse and addiction, and to Dr. Momenan (NIAAA) in his studies of alcoholism. - We collaborate with Dr Ernesta Meintjes (U Cape Town) on data analysis of the effects of prenatal alcohol exposure on the brains of infants and toddlers. - Our precise registration tools (for aligning fMRI scans to anatomical reference scans) are important for individual subject applications of brain mapping, such as pre-surgical fMRI planning. - Our real-time fMRI software is being used for studies on brain mapping feedback in neurological disorders, is used daily for quality control at the NIH fMRI scanners, and is also used at a few extramural sites. - Our statistical methods are being applied to epilepsy patients undergoing surgical planning with electro-corticography. Publications: The first 2 papers listed are our response to the PNAS paper that ignited a controversy about the accuracy of FMRI-generated brain maps. Other publications describe new data analysis methods, or are the results of collaborating with brain researchers.
Showing the most recent 10 out of 223 publications
