Skip to main content
Fig. 7 | BMC Bioinformatics

Fig. 7

From: Slideflow: deep learning for digital histopathology with real-time whole-slide visualization

Fig. 7

Comparison between background filtering methods. a Comparison between slide background filtering methods. An example whole-slide image with large pen mark artifact is shown, with associated background filtering masks applied. Black areas indicate masked background. With Otsu’s thresholding alone, the pen mark is identified as foreground, and several parts of the tissue area are erroneously removed as background. Gaussian blur filtering removes the pen mark, but some smaller areas of background near edges are not removed. Performing Gaussian blur first, followed by Otsu’s thresholding, results in the most accurate background identification and removes the pen mark. b Example images identified as background using Otsu’s thresholding, but not when using grayspace filtering. c Example images identified as background using grayspace filtering, but not when using whitespace filtering

Back to article page