Open AccessDot count proportion estimation in FISH specimens
Author(s) -
Castleman Kenneth R,
White Benjamin S
Publication year - 1995
Publication title -
bioimaging
Language(s) - English
Resource type - Journals
eISSN - 1361-6374
pISSN - 0966-9051
DOI - 10.1002/1361-6374(199506)3:2<88::aid-bio5>3.0.co;2-1
Subject(s) - fluorophore , fish <actinopterygii> , sample (material) , confusion , matrix (chemical analysis) , fluorescence in situ hybridization , sample size determination , computer science , algorithm , artificial intelligence , statistics , mathematics , fluorescence , biological system , pattern recognition (psychology) , chromosome , optics , physics , biology , chemistry , chromatography , fishery , psychology , biochemistry , gene , psychoanalysis , thermodynamics
Fluorescent in situ hybridization (FISH) is commonly used to tag individual chromosomes, or portions thereof, with a fluorophore. In interphase cells, a dot appears for each labeled chromosome. Multi‐band fluorescence microscopy permits one to view multi‐labeled specimens in color. Dot counting, whether conducted by human observation or implemented by image analysis algorithms, is subject to error. Dots can be missed, and false signals are sometimes counted. Dot counting errors undermine the accuracy of the clinical or research data being sought. This paper analyses FISH dot counting as a proportion estimation problem. It presents a method for removing the bias introduced into the estimate by dot counting errors, so that the accuracy is then restricted only by sample size considerations. It presents a theory that quantifies the RMS estimation error, given the sample size and a confusion matrix of dot counting error rates. This suggests guidelines for algorithm optimization and predicts what performance is practically attainable. The paper also illustrates the application of the theory with a small study.