Determination of Appropriate Thresholding Method in Segmentation Stage in Detecting Breast Cancer Cells

Abstract

As in all cancer types, the early detection of breast cancer is vital in terms of patients holding on to life. Today, computer-aided image processing systems play an important role in the detection of diseases. Analyzing the images with accurate image processing methods is very important for professionals to interpret the images and to develop the treatment methods for diseases appropriately. The images contain-ing cancer cells (tumoroid) used in this study were obtained from the mini-Opto tomography device that creates 3D images by reconstruction of 2D images taken from different angles. It is an electronic, mechan-ical, and software-based device capable of 3D imaging of tumoroids up to 1 cm in diameter in size. Ob-serving an entire tumor spheroid that has the size of several centimeters in size in a single square image with a microscope is not possible, but with mini-Opto tomography it is possible. In our study, a few layers of 3D images of the tumoroid produced by MCF-7 breast cancer cells obtained on the different days from the mini-Opto device were used. Image thresholding offers many advantages at the segmenta-tion stage in order to distinguish the target objects. In this study, the determination of the most appropriate thresholding method for detecting the main tumor masses in the layered images was investigated. Moreo-ver, the contours of the tumoroid were determined in the original images based on applying the outcomes of thresholding. While various thresholding methods have been applied on diverse images in the literature, we have applied a few thresholding methods to small tumors up to 2 mm in size. As a result of the quali-tative assessment based on the results of the contour drawings on the thresholded images, the global thresholding and adaptive thresholding methods gave the best results.