Automatic Segmentation and Counting Method of Fluorescent Microscopic Images of Retinal Cells

Abstract

The invention discloses an automatic cutting and counting method for fluorescent microscopic images of retinal cells. The method comprises the following steps: (a) preprocessing, wherein the images are preprocessed and noise points in the images are filtered away; (b) boundary encoding, wherein profiles of the cells are extracted and the extracted profiles are encoded; (c) concave point detection, wherein concave points in the profiles are found and marked; (d) cutting, wherein adherent cells are cut. The method combines a conventional algorithm and various algorithms proposed in modern time, and retains the advantages of simplicity for calculation, low calculation cost and high running efficiency in threshold filtration and edge detection algorithms. At the same time, the method combines Freeman chain code and polygonal concavity and convexity methods, better cuts the adherent cells and enables a cutting result to have relatively good accuracy and relatively high efficiency.

Description

Technical field [0001] The design involves the automatic segmentation and counting method of retinal cell fluorescence microscopic images, and belongs to the technical field of cell image segmentation and image processing. Background technique [0002] In the process of disease diagnosis, treatment and repair, the study of biological cell structure and morphological changes can provide help for disease diagnosis and treatment. The most important and difficult thing is the recognition and segmentation of cell morphology in biological cell images. Due to the complexity and diversity of cell images, there is no completely universal segmentation method that can achieve correct segmentation of all cell images. The segmentation of overlapping and adhering cells is a major problem in cell image segmentation. In order to improve the accuracy of the analysis results, a separation algorithm must be used to automatically separate them into individual cells. [0003] There are many methods f...

AI Technical Summary

Problems solved by technology

However, the determination of the threshold mainly depends on the gray histogram, and seldom considers the spatial position relationship of the pixels in the image. Therefore, when the background is complex, especially when several research targets overlap on the same background, part of the boundary information is easily lost, resulting in segmentation failure. incomplete

The edge-based segmentation algorithm is complex and computationally expensive, it is difficult to completely detect the edge, and once there is noise interference, the effect of the edge detection operator is not ideal.

The region-based segmentation algorithm has a large overhead. When the predetermined error value introduced in the calculation process is not selected properly, it is easy to introduce misjudgment, and it is easily affected by the overlapping interference between the internal organizations of the analysis target.

In recent years, new algorithms have greatly promoted cell segmentation technology. For example, wavelet transform segmentation has a certain guiding effect on fine segmentation, but the size and shape of mathematical morphological structural elements have a great impact on the segmentation results.

The segmentation algorithm based on digital morphology can accurately locate the edge, and the effect of suppressing noise is better, but sometimes isolated points or unclosed points may appear on the obtained edge

Images