Automatic detection of DNA double strand breaks after irradiation using an γH2AX assay

Tim Hohmann¹*, Jacqueline Kessler²*, Urszula Grabiec¹, Matthias Bache², Dirk Vordermark² and Faramarz Dehghani<sup>1

1</sup>Department of Anatomy and Cell Biology and ²Department of Radiotherapy, Martin Luther University Halle-Wittenberg, Halle (Saale) Germany
*These authors contributed equally

Offprint requests to: Faramarz Dehghani, Institute of Anatomy and Cell Biology, Martin Luther University Halle-Wittenberg, Grosse Steinstrasse 52, 06108 Halle (Saale), Germany. e-mail: faramarz.dehghani@medizin.uni-halle.de

Summary. Radiation therapy belongs to the most common approaches for cancer therapy leading amongst others to DNA damage like double strand breaks (DSB). DSB can be used as a marker for the effect of radiation on cells. For visualization and assessing the extent of DNA damage the γH2AX foci assay is frequently used. The analysis of the γH2AX foci assay remains complicated as the number of γH2AX foci has to be counted. The quantification is mostly done manually, being time consuming and leading to person-dependent variations. Therefore, we present a method to automatically analyze the number of foci inside nuclei, facilitating and quickening the analysis of DSBs with high reliability in fluorescent images. First nuclei were detected in fluorescent images. Afterwards, the nuclei were analyzed independently from each other with a local thresholding algorithm. This approach allowed accounting for different levels of noise and detection of the foci inside the respective nucleus, using Hough transformation searching for circles. The presented algorithm was able to correctly classify most foci in cases of "high" and "average" image quality (sensitivity>0.8) with a low rate of false positive detections (positive predictive value (PPV) >0.98). In cases of "low" image quality the approach had a decreased sensitivity (0.7-0.9), depending on the manual control counter. The PPV remained high (PPV >0.91). Compared to other automatic approaches the presented algorithm had a higher sensitivity and PPV. The used automatic foci detection algorithm was capable of detecting foci with high sensitivity and PPV. Thus it can be used for automatic analysis of images of varying quality. Histol Histopathol 33, 475-485 (2018)

Key words: γH2AX, Hough transformation, Foci Segmentation, Foci Detection, Foci

DOI: 10.14670/HH-11-945