Figure 1 | Scientific Reports

Figure 1

From: Automated Training of Deep Convolutional Neural Networks for Cell Segmentation

Figure 1

Data, training and testing pipelines for cell segmentation. (a) A z-stack of bright-field time-lapse microscopy images captured without any staining (time = t 1 to t N ). (b) A z-stack of bright-field images along with fluorescent images of nuclear and cytoplasmic stains captured after staining (time = t N+1). The green box represents the CellProfiler pipeline automatically generating illumination corrected images (data) and the segmented images (labels) for DCNN training. The data and the label are subjected to data augmentation to create the final dataset for training the DCNN. See Supplementary Fig. 2 for the detailed DCNN architecture. (c) The complete cell segmentation pipeline in CellProfiler (green box) that receives z-stack bright-field input images and outputs segmented cell images.

Back to article page