Computer-Aided Diagnosis Algorithm for Classification of Malignant Melanoma Using Deep Neural Networks

Chan-Il Kim,Seok-Min Hwang,Chang-Hee Won,Eun-Bin Park,Jong-Ha Lee

doi:10.3390/s21165551

Chan-Il Kim, Seok-Min Hwang + Show 3 more

Open Access

https://doi.org/10.3390/s21165551

Copy DOI

Journal: Sensors	Publication Date: Aug 18, 2021
Citations: 13	License type: CC BY 4.0

Affiliation: Keimyung University, Temple University

Abstract

Malignant melanoma accounts for about 1–3% of all malignancies in the West, especially in the United States. More than 9000 people die each year. In general, it is difficult to characterize a skin lesion from a photograph. In this paper, we propose a deep learning-based computer-aided diagnostic algorithm for the classification of malignant melanoma and benign skin tumors from RGB channel skin images. The proposed deep learning model constitutes a tumor lesion segmentation model and a classification model of malignant melanoma. First, U-Net was used to classify skin lesions in dermoscopy images. We implement an algorithm to classify malignant melanoma and benign tumors using skin lesion images and expert labeling results from convolutional neural networks. The U-Net model achieved a dice similarity coefficient of 81.1% compared to the expert labeling results. The classification accuracy of malignant melanoma reached 80.06%. As a result, the proposed AI algorithm is expected to be utilized as a computer-aided diagnostic algorithm to help early detection of malignant melanoma.

Highlights

Malignant melanoma accounts for approximately 1–3% of all the cases of malignant tumors diagnosed in Western countries
This paper proposed a malignant melanoma classification algorithm based on deep learning
This study verified that a training method based on two deep learning models can contribute to the early diagnosis of malignant melanoma through the extraction and classification of lesion areas

Summary

Introduction

Malignant melanoma accounts for approximately 1–3% of all the cases of malignant tumors diagnosed in Western countries. Quently, the CNN is utilized to predict malignant and be nignThtuemcoomrspuintetrh-aeidReGd Bdiiamgnaogseisomf tehtheoedxbtraascedteodnledseieopnleaarrenai.ng, which was proposed in this study, can perform a training process by itself under the supervision of a human b2e.inLgesuisoinngAthreacoSnecgempteonftdaetieopnly staking artificial neural networks for fields, including fN2eae.1ttu.[D1re5a]etiaxstruatciltiizoend, lesion area to segment extraction, and a lesion area in lesion classification [14]. Among 2000 dermoscopy images used in this study, the images including an excessive amount of hair (Figure 3a), unnecessary marks for the training process of the model (Figure 3b,c), and a great amount of magnifier noise (Figure 3d) were removed from the datasets for deep learning.

Results

Conclusion