Faster R-CNN based microscopic cell detection

Abstract

The automatic analysis of microscopic images is an important subject of medical image processing, of which the cell detection is an important part. However, owing to the different size and shape, as also as the adhesion among cells, detecting and locating cells accurately seems to be a very challenging task. In this work, we investigate applying the Faster R-CNN, which has recently shown incredible performance on many public datasets, to cell detection. The Faster R-CNN contains both segmentation and classification. By training a Faster R-CNN model, a series of experiments are achieved. Experimental results show that the Faster R-CNN can detect almost all cells in a microscopic image. The proposed cell detector has improved detection performance, and it is easy-implemented and time-saving.