Засоби оптимізації кількості кластерів при кластеризації МРТ-знімків ракових пухлин мозку за допомогою алгоритму K-MEANS

Abstract

In this paper, the working principle of the medical image-clustering algorithm for image medical images has been reviewed, and ways to overcome its shortcomings have been proposed. As these ways we analyzed the application of elbow method and silhouette method to find the optimal number of clusters. The accuracy of these methods has been evaluated, and the peculiarities of their application to improve the quality of image segmentation by k-means method have been assessed.