From Van Essen Lab

Revision as of 19:21, 27 March 2010 by David (Talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search


Foci Search in Caret6

Goal of Foci Search

What is the goal of foci search? Is the user searching for foci or is the user searching for studies? DVE: Both, but foci are paramount in my view

 Should the search return a list of foci or a list of studies?

DVE: Both.

It appears that the ultimate goal is to find studies and that foci are an intermediate step in the search process. DVE: Only in some cases.

Is there a log available showing the search that users are currently performing? DVE: YES: Quick-search (summary) is at Quick-search (details) is at (admin privileges needed) There are also reports on WebCaret and SumsDB download usage (summary and details). I have used these many times in recent years to get a sense of overall usage, and it has been on my mind for a while that we should get a more extensive reporting system in place. Now that Erin is starting to work with Google Analytics, perhaps we can see whether that (or something else) would be useful for extracting summary statistics. Ping, can you look into this?

Search Methods

There are two methods for searching studies, spatial location and functionality. A search query may be either or both of these methods. For instance, a user may search a spatial location (angular gyrus) and a functionality (mathematics).

Spatial Location Search

There are two ways to specify a spatial location, stereotaxic and partitioning scheme.

A stereotaxic spatial search consists of a coordinate and a radius. The stereotaxic coordinate is either entered by the user or from clicking on a surface.

A partitioning scheme spatial search consists of one or more areas selected from a partitioning scheme such as Brodmann or Geography (gyri and sulci). Areas in a partitioning scheme are chosen from list of checkboxes with the name of each area, by clicking on a surface, or by clicking annotations. For the list of areas, it may be useful to list the known functionality of each area and/or provide links to Wikipedia, NeuroLex, NIF, Neuronames, etc.

Sub-cortical structures should also be added to the PALS atlas so that interested users could search for data in those structures. DVE: This is a good summary. As a semantic point, I now favor the term 'parcellation scheme' instead of 'partitioning', because it fits naturally with the tem 'parcel' that can be a functional or geographic subdivision of cortex or subcortical regions.

Functionality Search

Keywords appear to be the attribute for a functionality search. However, the major problem with the study data is that it is "raw" data. For example, viewing the foci keywords page presents keywords that are synonyms (Alzheimer, Alzheimer disease, Alzheimer's disease) or irrelevant (axial, brain, fMRI, human, etc). In addition, with 600+ studies loaded more than 1200 keywords are presented.

What is needed is a "curation" of studies. That is, one examines each study (abstract, title, keywords) and assigns each study a single term that describes the functionality examined in the study. This would result in a small list of functionality terms that could be listed and selected by the user. This list might be similar to the list at the bottom of

DVE: I think the issue is more complex in several ways. 1) A given study often contains foci related to two or more distinct tasks (e.g., semantic memory vs working memory; face activations vs 'place' activations, etc). Keywords, while useful, cannot discriminate between these. This is why Erin has put a huge effort into assigning 'focus-specific' metadata. The most important is what ends up as the focus 'class' - a brief encoding of the most important task attribute. 2) Foci searches are currently set up to enable both 'narrow' and 'broad' searches. Narrow searches target the foci class and perhaps a few other categories and aim to minimize false negatives (but must tolerate missing many potentially valid foci). Broad searches tolerate lots of false positives (requiring more post-hoc filtering and analysis) but have a higher overall yield. The extreme form of broad search includes title, keywords, all text in the abstract, and other metadata fields as well.

I think we need to give this more thought and discussion before figuring out what will be best for Caret6 functional searches.

Listed below are examples of keywords from studies. A guess at the functionality studied is within parenthesis.

  • PET; autobiographical memory; episodic memory; retrieval; cingulate; limbic system (Memory)
  • extrastriate cortex; face perception; functional MRI; fusiform gyrus; ventral visual pathway; word perception (Perception)
  • sequential finger movements; PET; supplementary motor area; premotor cortex; bimanual coordination (Motor)
  • remapping; spatial perception; arm; eye; spatial memory; spatial updating; parietal cortex; LIP; PRR (Spatial)
  • working memory; prefrontal cortex; cortical deactivation; sleep deprivation; functional imaging; BOLD fMRI (Memory)
  • prefrontal; frontal; inferior; temporal; cortex; lobes; frontopolar; anterior; dorsolateral; ventrolateral; episodic; associative; working; visual; memory; long term; short term; executive; control; fMRI; neuroimaging; event related (Memory)

A search of keywords could be left to an advanced search mode.

Advanced Search

Which attributes (area, authors, citation, class, comment (focus), comment (study), data format, data type, geography, keywords, MESH terms, name, ROI, spatial, species, stereotaxic space, structure, table header, table subheader, title) are truly useful for searching?

The advanced search interface could be similar to the current online foci search interface.

Personal tools
Sums Database