You are here

Medical Clustering

Clustering of body part x-ray

The primary objective of this task is to group digital x-ray images into four clusters: head-neck, upper-limb, body, and lower-limb. The secondary goal of this task is farther partitioning the initial clusters into sub-clusters, for example the upper-limb cluster can be farther divided into: Clavicle, Scapula, Humerus, Radius, Ulna, and Hand.


  • 01.11.2014: registration opens for all ImageCLEF tasks(Register here)
  • 15.11.2014: development data release starts (19.11.2014-Training Data is Available Now)
  • 18.04.2015: test data release starts
  • 18.05.2015: deadline for submission of runs by the participants
  • 19.05.2015: release of processed results by the task organizers (Reviced results)
  • 07.06.2015: deadline for submission of working notes papers by the participants (Submit Here)
  • 30.06.2015: notification of acceptance of the working notes papers
  • 15.07.2015: camera ready working notes papers
  • 08.-11.09.2014: CLEF 2015, Toulouse, France


In 2013 we received about 5,000 digital x-ray images from a hospital, which is collected over about a year of operation of the newly acquired digital x-ray talking facility at that hospital. We wanted to device a software that will be able to first partition the x-rays into four major clusters, and then farther sub-divide initial clusters into sub-clusters. With digital x-ray technology it is easy to take higher resolution images and tag them while data is taken, however, there are millions of old school plate based x-rays available for archiving and analyzing. This large number of data will remain unused if a faster organizer software is not built.

Tasks overview

  • Image Type:
    All images of thins project is are of .dcm type. Sample Image
  • Feature Detection:
    Though the task in hand suggests that first cluster the data into four major groups such as: head-neck, upper-limb, body, and lower-limb, and then construct subgroups with in those groups. We believe that identifying features that determines membership of an x-ray image to one of the four initial clusters dependent on correct identification of components that defines a sub-group. For example, to identify limbs it might be a good idea to identify the longer bone structures.
  • Clustering or Classification:
    We are looking for four major groups: head-neck, upper-limb, body, and lower-limb. Point to be noted is that there will be some overlapping as x-rays may contain bones from different part of body in the same image. In such cases providing a membership value to both or all the classes will be an excellent addition, however, classifying such image to both or all the classes is allowed.
  • Evaluation :
    Developers should determine the performance of their systems based on the correct identification of bone structures. Then defining the characteristics of those structures to numerically represent discriminating features. And finally using those characteristics or features to perform classification or clustering performance is measured.


  • Md Ashraful Amin, PhD, Department of Computer Science and Engineering, Independent University, Bangladesh (IUB), aminmdashraful(replace-by-an-at)
  • Dr. Mahmood Kazi Mohammed, Sir Salimullah Medical College, Dhaka, Bangladesh, mkmohammed86(replace-by-an-at)