Circlet Transform in Cell and Tissue Microscopy Publisher

Summary: A study developed a method to detect circular cells in images, aiding faster disease diagnosis. #MedicalImaging #CellAnalysis

Sarrafzadeh O^{1, 3} ; Rabbani H^{2, 3} ; Mehri Dehnavi A^{2, 3} ; Talebi A⁴
Source: Optics and Laser Technology Published:2020

Abstract

Automatic detection of objects with circular pattern in digital images is an important topic in many fields of research especially in cell imaging. There are many cells in microscopic images that are circular such as Red Blood Cells (RBCs), White Blood Cells, hematopoietic cells and different types of parasite eggs. Automatic detecting, recognizing and quantifying of these cells provide rich information to pathologists to improve the study/diagnosis of different diseases. Many previously proposed methods utilize the edge information of a given image to detect circles that are not usually applicable for complex images. Fast Circlet Transform (FCT) is a new atomic representation based on using circular basis functions in different scales and frequencies which provides a novel and practical tool for circle detection and analysis of images with circular objects/patterns such as microscopic images. In this paper, three strategies based on FCT are proposed with the applications of FCT in cell and tissue microscopy as follows: In first application, a strategy is proposed for detecting and counting RBCs in microscopic images of blood smear in which an initial estimation of the number of RBCs is made and conflict circles are then removed to detect final and true RBCs. In second application, an algorithm is proposed to count and localize glomeruli in microscopic images of kidney sections by analyzing FCT coefficients in order to directly find circular objects. In third application, a method based on FCT is proposed to detect parasites in microscopic images with high unwanted impurities by modifying FCT coefficients and reconstructing the images. Our experimental results show the effectiveness and better performance of the proposed circlet-based methodologies in microscopic image analysis. © 2019 Elsevier Ltd