Automated high throughput image analysis of the ND10 complex of KSHV cells

For a better understanding of inter- and intracellular events, in modern microbiology and virology analysis of huge amounts of data is required. In order to assist biologists with the evaluation of such data, automated image analysis soft- ware can be applied. In the following, methods based on Difference of Gaussian filters, k-means clustering, locally adap- tive thresholds and the watershed transform are described and applied for segmentation of intra-nuclear dots as well as cell nuclei. Using the described methods, activity of proteins from the ND10 domain can be automatically determined based on image analysis. Evaluating performance for splitting of touching cells shows that only 20 out of 3,267 cells require manual correction of the segmentation results. Analysing a subset of 1,000 intranuclear spots resulted in 971 true positive, 29 false negative and 14 false positive detections.