Figure 2

Flowchart of the experiment workflow representing the data processing for the individual workflow steps. A set of images fully labeled for neoplastic nuclei were separated into a fixed set of training, test, and validation set. Measurements of small non-neoplastic lymphocytes were used as size reference and one set of images from dogs was used for the parameter optimization. The testing of the overall lymphoma classification performance was done on independent canine and feline lymphoma images.