Task 1: Cell level staining pattern classification
Similar to the contest from the previous year, this task is focused on cell-level pattern classification. This task has been initially proposed at the “HEp-2 Cells Classification” contest hosted by the ICPR 2012 and then proposed again in all the successive initiatives hosted by the ICIP 2013 and ICPR 2014, using datasets of increased size and complexities of the patterns. In particular, the competitions hold on years 2013 and 2014 considered the following six HEp-2 patterns: homogeneous, centromere, speckled, nucleolar, mitotic spindle and Golgi. This task will be the continuation from the previous year to witness the advances made by the community.
The dataset has been collected between 2011 and 2013 at Sullivan Nicolaides Pathology laboratory, Australia. It utilizes 419 patient positive sera, which were prepared on the 18-well slide of HEP-2 IIF assay from Immuno Concepts N.A. Ltd. with screening dilution 1:80. The specimens were then automatically photographed using a monochrome high dynamic range cooled microscopy camera which was fitted on a microscope with a plan-Apochromat 20x/0.8 objective lens and an LED illumination source. Approximately 100-200 cell images were extracted from each patient serum. In total there were 68,429 cell images extracted: 13,596 images used for training, made available to the teams, and 54,833 for testing, privately maintained by the organizers. The images were automatically segmented by using the DAPI channel and manually annotated by specialists.
The labeling process involved at least two scientists who read each patient specimen under a microscope. A third expert’s opinion was sought to adjudicate any discrepancy between the two opinions. We used each specimen label for the ground-truth of cells extracted from it. Furthermore, all the labels were validated by using secondary tests such as ENA and anti-ds-DNA in order to confirm the presence and/absence of specific patterns.
Each cell image contained in the database is annotated with the following information:
- Cell pattern (one of the patterns defined above)
- Cell intensity
- Cell mask
- ID of the image which the cell belongs to
The application obtaining the highest value of the mean class accuracy in cell classification over the test set will be declared as the winner. The proclamation of the winner will be made during the contest session at ICPR 2016.