Model for Detecting Globules in Images of Skin Neoplasms

Abstract: This article is devoted to the digital processing of images of skin neoplasms to detect significant structural elements in the diagnosis of melanoma–globules (clumps, lumps). A new processing model is proposed, which makes it possible to stably select globules in images of different contrasts without the need to manually adjust the parameters. The results of the experiment confirming the adequacy of the model are presented. The globule recognition accuracy ranged from 81 to 89%, depending on the contrast of the original images. The experimental sample of images contained 2868 globules. © 2022, Pleiades Publishing, Ltd.