A public-domain image processing tool for automated quantification of fluorescence in situ hybridisation signals

摘要

Aims: To develop and evaluate an automated method for quantification of HER2 fluorescence in situ hybridisation (FISH) signals. Methods: Using a popular, open source image manipulation tool, ImageJ, a macro for FISH signal assessment was created. A comparison against traditional manual counting was performed in breast cancer specimens from 42 patients. The tumour specimens were hybridised with probes for HER2 and chromosome 17 centromere (CEP17) and selected areas were digitised for image processing. Hybridisation signals were calculated both manually and automatically with the ImageJ custom macro. Results: The correlation coefficient between the automatic and manual HER2/CEP17 ratios was 0.98. The corresponding percentage agreement was 90% and the kappa value was 0.82. Conclusions: This study shows that it is possible to automate the determination of HER2 amplification by the use of open-source software, with results comparable to manual counting. The automated counting decreases the time needed for sample analysis and provides possibilities to enhance inter- and intralaboratory reproducibility of results. The FISH quantification tool (FishJ) is available for download as an ImageJ macro or alternatively it can be utilised through a web interface with an option of uploading FISH images for hybridisation signal counting. Combined with digitisation of FISH samples, the FishJ macro enables gene copy number to be assessed and re-evaluated on any area of a digitised specimen.