MeSH Term & Author - My Search for Optical Illusions in 2000

In HTML file title and heading you can see the bibliographic field types the user was interested in as well as his comment. He entered this information in Meva's consultation form. Here the user would like to analyze medical subject headings (MeSH terms; a controlled vocabulary of biomedical terms which is used to describe the subject of each journal article) and authors. Since he has choosen two field types instead of one, Meva displays relations between both fields.

This section shows system messages: warnings, recommendations to improve the search strategy, links to frequently asked questions or messages about data conversions performed by Meva. Please note in this example the links to the FAQ (frequently asked questions), in particular for problems whilst saving the file, for missing chart bars / contingency cell colors and for the possibility to change the font size of this result.

The parameter table lists all the parameters the user provided in Meva's consultation form.¹

• Top: The user wants to display only the 8 most frequent fields in the diagrams.
• Filter 1: An asterisk (*) as filter for MeSH terms causes Meva to analyze only major topics. Minor topics or generic subject headings like Human or Male will be left out.
• Filter 2: Empty. Since the user selected authors for field 2, an author-specific filter will be displayed in addition (here with the default value - all authors).
• Minimal Frequency 1: A MeSH term must appear 6 times at least to be considered for analysis.
• Minimal Frequency 2: 1 (default).
• Details: The datails will be sorted alphabetically in ascending order for MeSH terms (default).
• Link Restrictor: The user has entered the search expression from his PubMed session.
• MeSH Codes: No (default). Of interest if users want to have codes attached to MeSH terms.
• MeSH Tree: The user requires Meva to display a MeSH tree until a depth of 10. Subtree F (Psychiatry and Psychology) shall collect the frequencies of subject headings. Note that selecting a subtree does not affect the tree display. Meva will always show all possible subtrees.

¹ Besides, Mevas idea of date and time of the consultation as well as some Meva data limits are stated. These limits can be used to check if Meva's capacity has been exceeded. In: 96/5120 KB: Meva received 202 of 5120KB allowed from the user, Out: 383/15000 Tags: Meva printed (about) 383 of 15000 allowed detail section fields, TopVal: 48/1000: The maximal allowed number in a contingency table was 1000, the current 48. The data limits apply only to HTML, not to text results.

The totals table shows how Meva processes the bibliographic fields¹ contained in PubMed's result file in consecutive phases. Here, the PubMed search result contained 108 articles with 1010 subject headings and 269 authors among other bibliographic fields.

Let's continue now with the processing of subject headings:

From the 1,010 subject headings, 344 passed the user-supplied filter: to analyze only major topics², an asterisk (*) was provided as a filter by the user. Since most subject headings occur multiple times, the frequency of unequal (distinct) headings was stated as well (116).

Subject headings occurring less than 6 times were omitted. The minimal frequency 6 was supplied by the user in Meva's consultation form.

All remaining subject headings were taken over into Meva's database. This database represents consequently an user-defined subset of a PubMed result file containing only frequent fields (user specified a minimal frequency) or fields of special interest (user specified a filter). If the user would not supply any of these restrictions, the database would include the same data for field 1 and 2 as the original Pubmed result file. Meva's analysis will operate only on this database!

¹ Within a PubMed search result file in MEDLINE format, all bibliographic information of an article (e.g. author, title, subject headings, publication date etc.) is structured as a list of fields. Meva can read this data and display contents and relations.
² Major topics are MeSH terms that are one of the main topics discussed in the article denoted by an asterisk on the MeSH term or MeSH/Subheading combination, e.g. Sarin/*toxicity. An asterisk as a MeSH term filter causes Meva to process only major topics.

The following picture illustrates how Meva applies user-defined filters and minimal frequencies to bibliographic fields. For didactic purposes, lower numbers have been used:

The first frequency distribution graph shows you the most frequent field 1 values of the database. Since the user selected MeSH terms as field 1, a top value of 8, and an asterisk in the filter 1 inside Mevas consultation form, Meva displays the 8 most frequent MeSH major topics of the database. Optical Illusions e.g. occurred 48 times and thus was found to be the most frequent major topic inside of the PubMed file of the user. (This is not remarkable since the PubMed file was the result of a PubMed search for this term.) The following subject headings are more interesting since their high frequencies suggest that Optical Illusions emerge physiologically on the basis of Motion Perception, Pattern Recognition and Form Perception.

Since the user supplied a link restrictor (here: Optical Illusions[mh] 2000[dp]), a click on a field value in this table will trigger a new search in Pubmed. As an example, to go through the correlation between optical illusions and motion perception with a fine-tooth comb, click on the subject heading Motion Perception in this table to trigger a PubMed search for Motion Perception[mh] Optical Illusions[mh] 2000[dp]. The terms will be ANDed automatically by PubMed. The search field tags inside the square brackets denote PubMed's field types. The use of these qualified terms instead of a plain text search will result in more specific search results. For further particulars we refer to an article from Coletti MH and Bleich HL, 2001: Medical Subject Headings Used to Search the Biomedical Literature.

What is the link restrictor good for? To begin with by clicking a field value the user can make sure that the resulting solution set (here: Motion Perception[mh] Optical Illusions[mh] 2000[dp]) will not exceed PubMeds original solution set (here: Optical Illusions[mh] 2000[dp]). Secondly, the link restrictor prevents the user from searching for too general expressions. For example, a PubMed search for Motion Perception[mh] only would result in tens of thousands of articles!

You can find further information about the use of the link restrictor in Mevas Form Help.

The second frequency distribution graph displays the most frequent field 2 values of the database, provided the user supplied a second bibliographic field. Since the user selected Author as field 2 and a top value of 8 in Mevas consultation form, Meva displays the 8 most frequent authors. As you can see, Cavanagh published most of the articles related to optical illusions in 2000.

Tired of reading long lists of authors for each article? You can restrict Meva's analysis on first authors or last authors as well.

Since the user supplied a link restrictor (here: Optical Illusions[mh] 2000[dp]), a click on a field value in this table will trigger a new search in Pubmed. As an example, a click on Cavanagh P in this table would trigger a new PubMed search for Cavanagh P[au] Optical Illusions[mh] 2000[dp] and restrict the search radius thereby to publications from this author.

The first contingency table illustrates how often the most frequent field 1 values (here: MeSH terms) are combined within an article. This table will only be displayed if the field values can occur multiple times within an article, e.g. MeSH Term, Author, Publication Type oder EC/RN Number values. Note that row or column sums of a field can exceed its absolute frequency (which can be found in the frequency distribution or details table). More frequent values are displayed in darker colors.

As you can see, *Optical illusions have been found combined with *Pattern Recognition, Visual in 16 articles. From the frequency of shared ocurrences you can infer often the degree of correlation between two field values. In this example, the frequent socialisation of optical illusions with pattern recognition suggests that phenomena related to optical illusions can arise inter alia on the basis of pattern recognition.

You can use the contingency table for other purposes as well: If you are searching for diseases (instead of for Optical Illusions) in PubMed, you can retrieve via Meva's contingency tables important associated diagnostic or therapeutic procedures.

Since the user supplied a link restrictor (here: Optical Illusions[mh] 2000[dp]), a click on the shared field of Pattern Recognition, Visual and Orientation in this table would trigger a new PubMed search for Pattern Recognition, Visual[mh] Orientation[mh] Optical Illusions[mh] 2000[dp] and restrict the original search radius thereby to this context. If the user did not supply a link restrictor, Meva will produce a link to PubMed anyhow, since the the AND conjunction of both field values is specific enough.

The second contingency table illustrates how often the most frequent field 1 values are combined with the most frequent field 2 values within an article, provided the user supplied a second bibliographic field in Meva's consultation form. In this example, the degree of contingency between MeSH terms and Authors will be displayed. Note that row or column sums of a field can exceed its absolute frequency (which can be found in the frequency distribution or details table). More frequent values are displayed in darker colors.

A click on a field value will trigger a new PubMed search: e.g. clicking on the shared field of Cavanagh P and Motion Perception/*physiology in this table would trigger a new PubMed search for Cavanagh P[au] Motion Perception/*physiology[mh] Optical Illusions[mh] 2000[dp] and restrict the original search radius thereby to publications from this author. If the user did not supply a link restrictor, Meva will produce a link to PubMed anyhow, since the the AND conjunction of both field values is specific enough.

From the degree of contingency between 2 field values you can quite often draw important conclusions: In this example, the contingency between an author and subject headings allows to determine focal points of an author and consequently an author profile. Looking at this table you can see that the author Cavanagh P published most articles about Motion Perception/*physiology.

It might be astonishing that the most frequent subject heading for Cavanaghs articles is not Optical Illusions - after all the user has searched for it in PubMed and each article of the resulting PubMed file should contain this keyword, consequently also the 5 articles from Cavanagh (see frequency distribution of authors). The reason is simple: If you search for say Optical Illusions[mh] in PubMed, it will also find articles indexed by the NLM librarians with e.g. *Optical Illusions, Optical Illusions/*physiology, Optical Illusions/physiology or *Optical Illusions/physiology. Meva accounts each of these variations being a separate subject heading. Would you enter Optical Illusions instead of an asterisk as filter 1 in Meva's consultation form, you could see that from Cavanaghs articles, 1 has been indexed with *Optical Illusions, 3 with Optical Illusions/physiology and 1 with Optical Illusions. Consequently some of these subject headings could not pass the minimal frequency filter and thus could not pop up in Meva's result.

Conclusion: You must be aware of the fact that using filters and minimal frequencies will produce a database being a subset of the original PubMed search result. Interpretation of Meva's frequency distribution and contingency tables must take into account that these tables will be produced from this database.

The details table lists all bibliographic field values of Meva's database. In this example, the table has been shortened to keep this help file small. Table sort order and sort type has been set up by the user in the consultation form. Each table row contains an index i, the frequency n of a field 1 value, similarly the cumulated frequency nc, the field 1 value itself, the number of the article (PMID - PubMed ID) which the field 1 values have been found in and the corresponding field 2 values which have been found in the article as well.

As an example, the MeSH major topic *Attention (found 7 times) have been used to index the PubMed article no. 11273404, which has been published by the authors Suzuki S and Peterson MA. It has also been used to index the PubMed article no. 11273385, which has been published by the authors Crawford LE, Huttenlocher J and Engebretson PH etc.

A click on an PMID will lead you immediately to the according PubMed article. Since the user supplied a link restrictor, Meva provided links for the other fields as well: a click on the author Suzuki S in this table would trigger a new PubMed search for Suzuki S[au] Optical Illusions[mh] 2000[dp] and restrict the search radius thereby to publications from this author.

The MeSH table shows how Meva processes medical subject headings consecutively to get a hierarchical subject heading tree. It will only be listed if the user selected MeSH terms:

Unification: As you can see, from the 11 unequal (distinct) subject headings of the database (see totals table at the beginning of this side) only 7 subject headings have been found in PubMed's MeSH tree. This is normal since many subject headings match only with one subject heading inside the MeSH tree: the subject headings Orientation/physiology and Orientation e.g. match only with the term Orientation of the MeSH tree. Generic terms however like Human do not own a code and will be dropped. Thus this number indicates all MeSH Terms of the database, which are freed from leading asterisks and trailing subheadings and which could be found in the MeSH tree.

Association: Since a unified term can own several codes, their number is listed here as well. Orientation e.g. can be found both under F01.058.577 as well as under F02.830.606.

Interpolation: To complete the tree, Meva inserts terms (color marked inside the tree) whose codes are between the code of the associated terms and the tree root. At this stage, Meva has compiled all the information needed to draw the MeSH tree!

Totals: This is the sum of associated and interpolated terms - all distinct (unequal) terms of the MeSH tree. Accordingly you should count 26 entries in the MeSH tree displayed at the last page.

The following picture illustrates how Meva processes the subject headings in order to get a MeSH tree. For didactic purposes, lower numbers have been used:

This paragraph shows the success of internal charging strategies and serves only for background information purposes since most users will not deploy it. A Branch fit strategy of 100% indicates that Meva could find at least one code for each string in the user-defined branch (here: F) and charge it with its corresponding term frequency. Visual Perception e.g. could be found twice: under F02.463.593.932 as well as under G11.697.911.860. Meva however was told by the user to charge only branch F, so only the term Visual Perception of the subtree F will be charged with the frequency of 17 as you can see in the MeSH tree below.

The MeSH tree is organized hierarchically according to the codes of its subject headings. Codes and subject headings are provided by the National Library of Medicine, USA. Each entry of the following tree contains 4 values: code, its subject heading, its frequency and its cumulated frequency in square brackets, being the sum of its own frequency and of all of its subterms (even if those will not be displayed due to a low presentational depth supplied by the user). Thus frequencies are cumulated towards the top (root) of the tree!