Benchmark set of synthetic images for validating cell image analysis algorithms

This article presents a synthetic image set for validation of cell image analysis algorithms. To address the problem of validation, we have previously developed a simulation framework for cell population images. Here, we apply the simulation for generating a benchmark set of cell images with varying characteristics. The value of simulation is in the ground truth information known for the generated images. Traditionally, the ground-truth has been obtained through tedious and error-prone manual segmentation of the images. While such approach cannot be fully replaced, we propose to use the simulated images for benchmarking along with manually labeled images, and present case studies of tuning and testing a cell image analysis algorithm based on simulated images.