Automatic Detection of Mitosis and Nuclei From Cytogenetic Images by CellProfiler Software for Mitotic Index Estimation

The objective of this paper was to apply the CellProfiler open-source software for automatic detection of mitotic and nuclei figures from digitized images of cultured human lymphocytes for MI assessment, and to compare its performance to that performed through semi-automatic and visual detection. Lymphocytes were irradiated and cultured for mitosis detection. Sets of images from cultures were analyzed visually and findings were compared with those using CellProfiler software. The CellProfiler pipeline includes the detection of nuclei and mitosis with 80% sensitivity and more than 99% specificity. We conclude that CellProfiler is a reliable tool for counting mitosis and nuclei from cytogenetic images, saves considerable time compared to manual operation and reduces the variability derived from the scoring criteria of different scorers. The CellProfiler automated pipeline achieves good agreement with visual counting workflow, i.e. it allows fully automated mitotic and nuclei scoring in cytogenetic images yielding reliable information with minimal user intervention.

Source: Radiation Protection Dosimetry - December 22, 2016 Category: Radiology Authors: Gonzalez, J. E., Radl, A., Romero, I., Barquinero, J. F., Garcia, O., Di Giorgio, M. Tags: Paper Source Type: research