Nuclei image segmentation using image processing techniques and heuristic methods

Digital pathology is becoming increasingly important in the modern scientific laboratory environment and is becoming an increasing technological requirement. In particular, the detection and segmentation of cell nuclei are extremely important for the development of interpretable models. It supports strong research programs in the fields of digital image processing, image enhancement and image-based pattern recognition. Various image processing techniques image segmentation plays a vital role in the step of analyzing the given image. Image segmentation is the basic step for analyzing images and extracting data from them. In this thesis, edge finding, thresholding, region magnification and clustering operations were performed to divide the segmentation of an image. If we compare the proposed algorithm with the previously proposed algorithms, this algorithm does not require a lot of parameters, and it is also faster, simpler and more flexible. In this thesis, PSB was performed on histological images created by H&E samples dyed from the 2015 Crowdsourcing Nucleiannotation and 2018 Data Science Bowl datasets. The noise in the images was first eliminated using morphological techniques, and then an Artificial Bee Colony based adaptive histogram algorithm was used for the segmentation of the nuclei images. The accuracy and effectiveness of the results have been tested by comparing them with other optimization algorithms. An average accuracy rate of 93.64% was achieved for cancer nuclei.