Description of the dataset to be used

In the following we provide few notes about the dataset that will be used for the competition. Note that the dataset will be made available only to researchers registered to the contest. Registrations will start on November 15, 2011.

Acquisition and manual labeling procedure

HEp-2 images were acquired by means of a fluorescence microscope (40-fold magnification) coupled with a 50Wmercury vapor lamp and with a digital camera (SLIM system by Das srl). The camera has a CCD with square pixel of 6.45 m. The images have a resolution of 1388x1038 pixels, a color depth of 24 bits and they are stored in BMP format (see the following picture for an example).

example of an IIF image

Specialists manually segmented and annotated each cell at a workstation monitor, and reported data on fluorescence intensity and staining pattern. Firstly, a biomedical engineer manually segmented the cells by the use of a tablet PC. Subsequently, each image was verified and annotated by a medical doctor specialized in immunology. The image database is annotated with the following information:

  • Image main pattern and intensity;
  • Cell seed points;
  • Cell pattern, which can be one of: homogeneous, fine speckled, coarse speckled, nucleolar, cytoplasmatic or centromere.


The dataset is constituted by 28 images, which, as a whole, contain 1457 cells, so divided among the different patterns:

  • Number of cells: 1457
  • Centromere: 388
  • Coarse Speckled: 239
  • Cytoplasmatic: 128
  • Fine Speckled: 225
  • Homogeneous: 345
  • Nucleolar: 257
The participants will receive 14 out of the 28 images to be used as the training set. The subdivision in training and test sets will be done maintaining approximately the same pattern distribution over the two datasets.


The dataset used for the competition is one of the outcomes the research project "Classification of Immunofluorescence Images for the Diagnosis of Autoimmune Diseases" supported by "Regione Campania (Italy)".