Fig. 2

Pipeline to convert TL to pseudofluorescence. (A). TL image obtained by widefield microscopy. The red square represents a sliding window of 56 × 56 pixels used for image processing. (B). Patch classifier, which employs a U-NET architecture with five fully convolutional layers. An upscaled window (224px x 224px) is used as input. The class-one probability is used to generate a pseudofluorescent image as output (magenta colored). (C). The output is then downscaled and combined with the original imaging data to create a virtual imaging channel with pseudo-fluorescence (magenta)