An automatic segmentation algorithm for high-throughput digital PCR fluorescence images

Digital polymerase chain reaction (dPCR) fluorescence images are interfered by uneven illumination, which is difficult to be accurately segmented by traditional threshold methods. This paper developed a gray fluctuation threshold segmentation algorithm to solve this problem. Firstly, the image is preprocessed based on the Gaussian scale space (GSS) theory, which deeply improves the contrast of the image. Then the grayscale fluctuation curves of the image are extracted from the horizontal and vertical directions respectively, and all the fluctuation curves are iteratively searched for the large-scale wave peaks satisfying the specific fluctuation amplitude. Finally, each wave curve is segmented by the floating threshold, and the final segmented image is formed by the logical operation of the segmentation results in the two directions. Experiment results demonstrate that the proposed method is highly accurate compared to the traditional methods. The segmentation accuracy of the proposed method is above 97%, which is much higher than traditional segmentation methods.