Fig. 1From: Clinical application of the HM-1000 image processing for HER2 fluorescence in situ hybridization signal quantification in breast cancerNoise optimization by machine learning model. Noise optimization and background subtraction for CEP17 signals, showing (a) the original image colored in 3-3-2 RGB; (b) the image after application of the rolling ball algorithm, where the radius pixels used for background subtraction is 15; and (c) after application of the threshold (1490), which was automatically obtained from the linear regression modelBack to article page