Study on blood vessel segmentation of laser scattering microscopic image

Laser scatter imaging (LSCI) is a powerful method for analyzing blood flow in the human vasculature. Due to its advantages of non-contact, non-invasive and fast imaging, laser scattering imaging is increasingly being used in fundus vascular imaging. The research in this paper will contribute to the vascular segmentation of laser scattering microscopic images and their quantitative analysis. In this paper, the vascular segmentation algorithms for fundus scattered vascular images are investigated separately, and the segmentation effects of traditional and U-Net-based deep learning algorithms are compared. Based on the U-Net network, we propose a Recurrent-U-Net structure based on circular convolution and compare it with other algorithms. The results show that the deep learning segmentation algorithm outperforms traditional algorithms. Moreover, the Recurrent-U-Net network can reduce the problem of losing tiny vessels in the U-Net network and further improve the segmentation effect.