IBaI Institute | Home | Past events | ICDM´2002 | Abstracts

Mining Diagnosis Knowledge for HEp-2 Cell Classification

Dr. P. Perner and G. Schmidt, Institute of Computer Vision and Applied Computer Sciences, Leipzig/Germany

Pictures, stored in picture-archiving and communication systems or other image-data bases, provide new possibilities for a deep study of specific and temporal features of the image object, such as for e.g. lesion in medical images and for a dynamical study of the feature evolution. The application of data mining will help to get some additional knowledge about specific features of different classes and the way in which they are expressed in the image.

Data mining can elicit non-formalized expert knowledge; it can automatically create effective models for decision-making, and can help to find some inherent non-evident links between classes and their imaging in the picture. The new knowledge obtained as a result of data analysis in the database can enhance the professional knowledge of the expert or the user of the image data base. This knowledge can also be used for teaching novices or can support image analysis and iagnosis by the expert. An additional advantage of data mining application for the decision of medical or other tasks is on the long-run the opportunity for creation of fully automatic image-diagnosis systems that could be very important and useful in the case of lack of knowledge for decision-making.

In project LernBildZell (sponsored by the Germany Ministry of Economy BMWI) we are devoloping new methods for mining image data bases and the automatic image classification of medical cell images. We apply our methods to HEp-2 cell images. Hep-2 cells are used for the identification of antinuclear autoantibodies (ANA). They allow for recognition of over 30 different nuclear and cytoplasmic patterns, which are given by upwards of 100 different autoantibodies. The identification of the patterns has recently been done manually by a human inspecting the slides with a microscope.

In the talk we will describe how image mining can be performed, the methods that are used and how the results have been used for the developement of an automatic image analysis and classification system for HEp-2 cells.

mission | ICDM´2009 | past events | publications | tutorial days | contact