Development, implementation, and validation of a new method for meta-analysis of voxel-based neuroimage studies

Abstract

[eng] The objective of this thesis was to develop a method for the meta-analysis of neuroimaging studies that overcomes the drawbacks of the neuroimaging meta-analytic methods described above. To this end, we modified AES-SDM (Radua et al. (2014)) to incorporate several major changes that address each of the above limitations. This has resulted in a new method, called SDM-PSI, with profound differences from any previous neuroimaging meta-analytic method. I divided this overarching objective into the following individual objectives:

1. Accurately identify and circumscribe the drawbacks of the neuroimaging metaanalytic methods present at the time of the conception of this thesis. 2. Develop the statistical techniques of the new method. Optimize the usage of maximum likelihood estimation (MLE) to estimate the distribution of the missing information of the neuroimaging studies and combine it with the usage of multiple imputation techniques to recreate this non-published information. This combination shall intend to obtain better estimates of effect sizes in areas where CBM studies did not report results. 3. Develop an algorithm for the new method viable for being implemented efficiently, providing an implementation that can be successfully used in real-world scenarios. 4. Modify the SDM meta-analytic method to incorporate these new statistical techniques, as well as other novel changes in the field such as a standard subject-based permutation test instead of the spatial permutation approach used by previous methods.

5. Design and develop an algorithm for the usage of the standard permutation test as part of the SDM method. A computationally efficient implementation of such an algorithm shall also be feasible. 6. Adopt the scope not only of theoretical-methodological work but also of practical work with usable results and implementations ready to be used by the neuroimage meta-analysis community. This shall include the development of powerful software and a graphical user interface. 7. Disseminate the new method to ease the adoption of its application by the scientific community. This included the release of the new software in a user-friendly form, writing and recording a visual publication [117], and proactively seeking collaboration studies with scientific teams from other institutions and labs that would provide the first publications making use of the new method.

The first objective is addressed in the first published article that comprises this thesis; the second and third objectives are addressed in the second published article; the fourth, fifth and sixth objectives are addressed in the third published article, and the seventh objective is addressed in the fourth published article.