Dense pooling layers in fully convolutional network for skin lesion segmentation

Ebrahim Nasr-Esfahani; Shima Rafiei; Mohammad H Jafari; Nader Karimi; James S Wrobel; Shadrokh Samavi; S M Reza Soroushmehr

doi:10.1016/j.compmedimag.2019.101658

Dense pooling layers in fully convolutional network for skin lesion segmentation

Comput Med Imaging Graph. 2019 Dec:78:101658. doi: 10.1016/j.compmedimag.2019.101658. Epub 2019 Oct 7.

Authors

Ebrahim Nasr-Esfahani¹, Shima Rafiei¹, Mohammad H Jafari², Nader Karimi¹, James S Wrobel³, Shadrokh Samavi⁴, S M Reza Soroushmehr⁵

Affiliations

¹ Department of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan, 84156-83111, Iran.
² Department of Electrical Engineering, University of British Columbia, Vancouver, BC V6T 1Z4, Canada.
³ Department of Internal Medicine, University of Michigan, Ann Arbor, 48109, USA.
⁴ Department of Electrical and Computer Engineering, Isfahan University of Technology, Isfahan, 84156-83111, Iran; Department of Electrical and Computer Engineering, McMaster University, Hamilton, ON L8S4L8, Canada; Michigan Center for Integrative Research in Critical Care, University of Michigan, Ann Arbor, 48109, USA.
⁵ Michigan Center for Integrative Research in Critical Care, University of Michigan, Ann Arbor, 48109, USA; Department of Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, 48109, USA. Electronic address: ssoroush@umich.edu.

PMID: 31634739
DOI: 10.1016/j.compmedimag.2019.101658

Abstract

One of the essential tasks in medical image analysis is segmentation and accurate detection of borders. Lesion segmentation in skin images is an essential step in the computerized detection of skin cancer. However, many of the state-of-the-art segmentation methods have deficiencies in their border detection phase. In this paper, a new class of fully convolutional network is proposed, with new dense pooling layers for segmentation of lesion regions in skin images. This network leads to highly accurate segmentation of lesions on skin lesion datasets, which outperforms state-of-the-art algorithms in the skin lesion segmentation.

Keywords: Deep neural networks; Dense pooling layer; Melanoma; Skin cancer.

MeSH terms

Dermoscopy*
Humans
Image Interpretation, Computer-Assisted / methods*
Melanoma / diagnosis*
Neural Networks, Computer*
Skin Neoplasms / diagnosis*