Time efficient cell detection in histopathology images using convolutional regression networks

Accurate cell detection is often an essential prerequisite for subsequent cellular analysis in computer aided diagnosis (CAD) for histopathlogy images. It is challenging due to high cell density, touching cells, low contrast, variant cell shapes and sizes, weak boundaries and the use of different image acquisition techniques. Existing methods are often struggling at tackling with the challenges at the same time. More importantly, the detection time efficiency, which is also crucial for cell detection in histopathology images, is limited by the complicated and redundant computing for many exsiting methods. In this paper, we propose a novel end-to-end cell detection pipeline based on convolutional regression neural networks to achieve competitive cell detection accuracy and better time efficiency at the same time. We evaluate our method on two challenging cell datasets and the comparative experiments demonstrate the superior performance of our method over existing state of the art.