Determining the importance of figuresin journal articles to find representativeimages                            Henning M...
Introduction  Images are essential for diagnosis and  treatment planning  Diversity of modalities and protocols is growi...
Motivation  Image retrieval and classification make   medical visual content accessible for search  Images from PACS or ...
Examples for using figure importance  Interfaces and visualization require choices   Mobile information search has small...
Database used  ImageCLEF 2012 dataset   ~75,000 articles of PubMed Central, ~300,000 images  Subset was used   ~800 ar...
Types of figure importance  Five categories – sorted by importance  Most important figure – 1 figure that best   describ...
User test  User test participant profile:  45-year-old physician, surgeon  Working with medical images on a daily basis...
Results  All 50 articles were analyzed and all 641  images manually ranked by importance  Usually over 1 hour per articl...
Rules for determining importance  Diagnostic images are more important than  graphs/system overviews  Diversity in the r...
More rules for the importance  Magnification  Highest magnification most important  MRI for soft tissue diseases, CT fo...
Results  Distribution of figure importance                                       11
Results  Importance vs. position of the figure in the  article based on relative position in the text                    ...
Example for figure importance                                13
Conclusions  Relative importance of visual content in  scientific articles is important for many  applications  First st...
Future work  Repeat the same test with a second  person to measure inter-rater  disagreement  Quantify the subjectivity ...
Thank you for your attention!Further information:henning.mueller@hevs.chhttp://medgift.hevs.ch/http://www.khresmoi.eu/http...
Upcoming SlideShare
Loading in …5
×

Determining the importance of figures in journal articles to find representative images

577 views
409 views

Published on

When physicians are searching for articles in the medical literature, images of the articles can help determining relevance of the article content for a specific information need. The visual image representation can be an advantage in effectiveness (quality of found articles) and also in efficiency (speed of determining relevance or irrelevance) as many articles can likely be excluded much quicker by looking at a few representative images. In domains such as medical information retrieval, allowing to determine relevance quickly and accurately is an important criterion. This becomes even more important when small interfaces are used as it is frequently the case on mobile phones and tablets to access scientific data whenever information needs arise. In scientific articles many figures are used and particularly in the biomedical literature only a subset may be relevant for determining the relevance of a specific article to an information need. In many cases clinical images can be seen as more important for visual appearance than graphs or histograms that require looking at the context for interpretation.

To get a clearer idea of image relevance in articles, a user test with a physician was performed who classified images of biomedical research articles into categories of importance that can subsequently be used to evaluate algorithms that automatically select images as representative examples. The manual sorting of images of 50 journal articles of BioMedCentral with each containing more than 8 figures by importance also allows to derive several rules that determine how to choose images and how to develop algorithms for choosing the most representative images of specific texts.

This article describes the user tests and can be a first important step to evaluate automatic tools to select representative images for representing articles and potentially also images in other contexts, for example when representing patient records or other medical concepts when selecting images to represent RadLex terms in tutorials or interactive interfaces for example. This can help to make the image retrieval process more efficient and effective for physicians.

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
577
On SlideShare
0
From Embeds
0
Number of Embeds
11
Actions
Shares
0
Downloads
1
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Determining the importance of figures in journal articles to find representative images

  1. 1. Determining the importance of figuresin journal articles to find representativeimages Henning Müller, Antonio Foncubierta-Rodriguez, Chang Lin, Ivan Eggel
  2. 2. Introduction  Images are essential for diagnosis and treatment planning Diversity of modalities and protocols is growing  Medical images represent largest amount of medical information produced  Medical literature carries much of the medical knowledge Images are essential for medical articles  Images can help understand content of an article quickly  Much quicker than reading a text to determine relevance Search in the literature is frequent 2
  3. 3. Motivation  Image retrieval and classification make medical visual content accessible for search Images from PACS or from the medical literature (as in the ImageCLEF image retrieval campaign)  When searching the image needs to be put into context Text and visual data to understand context Often more than a single image  Identification of important figures Use of visual content to determine relevance quickly Limit search to important figures 3
  4. 4. Examples for using figure importance  Interfaces and visualization require choices  Mobile information search has small interface  Creation of ground truth and rules for importance ranking to select in the best way  Often the first N are taken 4
  5. 5. Database used  ImageCLEF 2012 dataset  ~75,000 articles of PubMed Central, ~300,000 images  Subset was used  ~800 articles contained more than 8 images  Final choice of 50 articles containing 641 diagnostic images 5
  6. 6. Types of figure importance  Five categories – sorted by importance Most important figure – 1 figure that best describes the article Essential figures – 0-3 figures Important figures – 0-3 figures Little importance figures – 0+ figures Totally unimportant figures – 0+ figures 6
  7. 7. User test  User test participant profile: 45-year-old physician, surgeon Working with medical images on a daily basis  User test scenario: Searching for relevant articles on the topic of the text  Which figures contain the most important information about the article?  Requires to read and understand article  All systematic findings or rules for the ranking were written down 7
  8. 8. Results  All 50 articles were analyzed and all 641 images manually ranked by importance Usually over 1 hour per article  Rules and findings were noted and discussed  Statistics on figure importance and places in the text were made  Compound figures were difficult to judge Information content is high as they contain different subfigures and links between them Compound figure separation new sub task in ImageCLEF 8
  9. 9. Rules for determining importance  Diagnostic images are more important than graphs/system overviews  Diversity in the results set is important Not several times the same modality if others exist  Size/resolution matters Image quality is importance  Position in article Most importance in the conclusions, then results  Common vs. rare modality Common modalities are often easier to understand 9
  10. 10. More rules for the importance  Magnification Highest magnification most important  MRI for soft tissue diseases, CT for bone related ones Modality adapted to the disease, can be derived from RadLex terms  Compound vs. other Compound figure contains much information  In non-diagnostic images, statistical graphs most important 10
  11. 11. Results  Distribution of figure importance 11
  12. 12. Results  Importance vs. position of the figure in the article based on relative position in the text 12
  13. 13. Example for figure importance 13
  14. 14. Conclusions  Relative importance of visual content in scientific articles is important for many applications First study to investigate this in a user test  Can improve IR system performance Rank images based on position or remove some images to search a smaller sub space  Can improve visualization For small interfaces such as mobile phones and tablets To combine text and images in an optimal way 14
  15. 15. Future work  Repeat the same test with a second person to measure inter-rater disagreement Quantify the subjectivity of the task  Model the rules and optimize an automatic selection strategy Modality classification to filter out unwanted images Clustering and use of cluster centers to increase diversity Find the place of figures in the text and rank based on this 15
  16. 16. Thank you for your attention!Further information:henning.mueller@hevs.chhttp://medgift.hevs.ch/http://www.khresmoi.eu/http://www.mitime.org/width/ 16

×