Desktop version

Home arrow Engineering arrow Cellular Image Classification

The ICIP2013 Training Dataset

The HEp-2 cell images of ICIP2013 dataset is obtained by using a monochrome high dynamic range cooled microscopy camera which is fitted on a microscope with a plan-Apochromat 20x/0.8 objective lens and an LED illumination source. At least two scientists are involved in the labeling process. In dubious cases, a third expert is asked to adjudicate the conflict between the two scientists. So far, only the training dataset is available. However, the training dataset is big enough to evaluate different methods. The ICIP2013 training dataset contains 13596 cells which are categorized into six classes: homogeneous (ho), speckled (sp), nucleolar (nu), centromere (ce), nuclear membrane (nm) and golgi (go). The dataset includes two patterns less frequent occurring in the practical clinic as follows [24]:

  • Nuclear membrane: characterized by a thin membranous fluorescence around the nucleus in the interphase cells;
  • Golgi: characterized by speckled staining of a polar organelle adjacent to one part of the nucleus and composed of irregular large granules.

Thus, it offers a more realistic evaluation on the automatic classification algorithms. We partition the ICIP2013 training dataset into a training set consisting of 6842 cells from 42 slide images and a test set consisting of 6754 cells from 41 slide images. See Table 1.2 for detailed information about the dataset.

< Prev   CONTENTS   Source   Next >

Related topics