Transactions on Mass-Data Analysis of Images and Signals (ISSN:1868-6451)
Volume 1 - Number 2 - September 2009 - Pages 147-159
Automated classification of Immunofluorescence staining of HEp-2 cells in Clinical Routine Diagnostics
Ch. Plata1, H. Perner2, S. Speth1, K. J. Lackner1 and P. von Landenberg1
1Institute of Clinical Chemistry and Laboratory Medicine, Medical Center of the Johannes Gutenberg-University, D-55101 Mainz, Germany
1ImageInterpret GmbH, D-04107 Leipzig, Germany
Detection of antinuclear autoantibodies (ANA) based on Indirect Immunofluorescence (IIF) is of major importance for the diagnosis of several autoimmune diseases, such as Systemic Lupus Erythematosus. In clinical routine diagnostics, ANA are mostly visualized using HEp-2 cells, showing nuclear and cytoplasmatic immunofluorescence patterns. The manual evaluation of this screening test is labor intensive and still subjective. Standardization and further automation is needed to achieve comparable results based on objective parameters. Aim of this study was to evaluate results of an automated, software-based image analysis of HEp-2 cell fluorescence patterns in clinical routine diagnostic by comparing them to results of the manual method. Main focus was the solid differentiation between positive and negative HEp-2 cell fluorescence, the accuracy of computer analyzed HEp-2 cell fluorescence patterns identification as well as reliability of computer analyzed titer level determination. 299 serum samples were screened for presence of ANA and 2028 pictures of HEp-2 cell fluorescence patterns were generated. 381 pictures showed negative and 1647 positive staining. Nearly all negative samples (99,7%) were correctly classified by the software program. The automated pattern analysis classified the homogeneous pattern (81,2%), the centromere pattern (87,1%) and the nuclear dots pattern (76,7%) correctly resulting in an overall rate of 75,9% of correct classification. Titer analysis with a deviation of +/-1 titer levels was performed by the software program with 90,4% correct graduation. Using this software a reliable automated fluorescence pattern analysis, including positive/negative differentiation, fluorescence pattern classification and titer analysis is possible in routine diagnostics.