A Parallel Fish Image Processing Pipeline of High-Throughput Chromosomal Analysis

The detection of tumor cells is critical for early diagnosis of primary and metastatic tumor as well as for monitoring the anti-cancer treatment effect. Fluorescence in situ hybridization (FISH) probes have been utilized to detect the unique characteristics of either over-expressed oncogenes or abnormalities of chromosomes in tumor cell nuclei. Automated quantification of chromosomes and their interphase cell nuclei is required for high-throughput analysis of cellular dynamics, especially in clinical applications. We present a fully automated FISH image processing pipeline that is able to robustly detect and characterize both cells and chromosomes in a parallel way. Nuclei and chromosomes are treated with different strategies because of inconsistent internal features and external needs, and then re-associated for consolidated chromosomal analysis and tumor cell characterization. Experimental results of the combined workflow showed the high accuracy and efficiency of the both chromosome detection and cell segmentation.