A new optical density granulometry-based descriptor for the classification of prostate histological images using shallow and deep Gaussian processes.

•A public database of annotated prostate cancer images from the Clinical Hospital of Valencia.•We demonstrate the importance of optical density images in color deconvolution to encode the relevant features of prostate cancer.•We formulate a novel morphological descriptor based on granulometries for prostate cancer classification.•We introduce probabilistic models based on shallow and deep Gaussian Processes to address the discrimination between healthy and tumoral prostate tissue.•A fast and automatic method that detects almost perfectly prostate cancer on Whole Slide Images providing a useful tool to pathologists.