Our Roadmap

NiMARE’s primary goal is to consolidate coordinate- and image-based meta-analysis methods with a simple, shared and comprehensive interface. This should reduce brand loyalty to any given algorithm, as it should be easy to employ the most appropriate algorithm for a given project. It also provides an environment where comparisons between methods are easier to perform.

A secondary goal of NiMARE is to implement some of the more cutting-edge methods for analyses built on meta-analytic neuroimaging data. There are many tools or algorithms that use meta-analytic data, including automated annotation, meta-analytic functional characterization analysis, and meta-analytic parcellation. Many of these methods are either tied to a specific meta-analysis package or never make it from publication to useable (i.e., documented and tested) code.

Ultimately, we plan to support all (or most) of the methods listed below in NiMARE:

  • Coordinate-based methods (nimare.meta)
    • Kernel-based methods
      • Activation likelihood estimation (ALE)

      • Specific coactivation likelihood estimation (SCALE)

      • Multilevel kernel density analysis (MKDA)

      • Kernel density analysis (KDA)

    • Model-based methods (nimare.meta.model)
      • Bayesian hierarchical cluster process model (BHICP)

      • Hierarchical Poisson/Gamma random field model (HPGRF)

      • Spatial Bayesian latent factor regression (SBLFR)

      • Spatial binary regression (SBR)

  • Image-based methods (nimare.meta.ibma)

  • Automated annotation (nimare.annotate)
    • TF-IDF vectorization of text (nimare.annotate.tfidf)

    • Ontology-based annotation (nimare.annotate.ontology)
      • Cognitive Paradigm Ontology (nimare.annotate.ontology.cogpo)

      • Cognitive Atlas (nimare.annotate.ontology.cogat)

    • Topic model-based annotation (nimare.annotate.topic)
      • Latent Dirichlet allocation (nimare.annotate.topic.LDAModel)

      • Generalized correspondence latent Dirichlet allocation (nimare.annotate.topic.GCLDAModel)

      • Deep Boltzmann machines (nimare.annotate.topic.BoltzmannModel)

    • Vector model-based annotation (nimare.annotate.vector)
      • Global Vectors for Word Representation model (nimare.annotate.vector.Word2BrainModel)

      • Text2Brain model (nimare.annotate.vector.Text2BrainModel)

  • Database extraction (nimare.extract)
    • NeuroVault

    • Neurosynth

    • Brainspell

    • PubMed abstract extraction

  • Functional characterization analysis (nimare.decode)
    • BrainMap decoding

    • Neurosynth correlation-based decoding

    • Neurosynth MKDA-based decoding

    • BrainMap decoding

    • Text2brain encoding

    • Generalized correspondence latent Dirichlet allocation (GCLDA)

    • Prediction framework (e.g. NeuroQuery)

  • Meta-analytic parcellation (nimare.parcellate)
    • Meta-analytic parcellation based on text (MAPBOT)

    • Coactivation-base parcellation (CBP)

    • Meta-analytic activation modeling-based parcellation (MAMP)

  • Common workflows (nimare.workflows)
    • Meta-analytic coactivation modeling (MACM)

    • Meta-analytic clustering analysis

    • Meta-analytic independent components analysis (metaICA)