Application of Improved Chameleon Swarm Algorithm and Improved Convolution Neural Network in Diagnosis of Skin Cancer

Application of Improved Chameleon Swarm Algorithm and Improved Convolution Neural Network in Diagnosis of Skin Cancer

Wu Beibei, Nikolaj Jade
Copyright: © 2023 |Pages: 16
DOI: 10.4018/IJDWM.325059
Article PDF Download
Open access articles are freely available for download

Abstract

Skin cancer is affected by the uncommon evolution of skin cells and is a deadly type of cancer. In addition, skin lesion is affected by numerous factors, such as exposure to the sun, infections, allergies, etc. These skin illnesses have become a challenge in therapeutic diagnosis because of virtual resemblances, where image classification is vital to sufficiently diagnose dissimilar lesions. Therefore, early diagnosis is significant and can avert skin cancers like focal cell carcinoma and melanoma. A deep learning-based computer analyzing model can be an automatic solution in medical evaluations to overcome this issue. Hence, this paper suggests an improved chameleon swarm algorithm and convolutional neural networks (ICSA-CNN) for effective skin cancer identification and classification. The data are collected from the Kaggle dataset for classifying skin cancer. Chameleon swarm algorithm is a clustering technique utilized in data mining to the cluster dataset utilizing dynamic systems, and it can resolve constrained and global numerical optimization issues in skin cancer detection.
Article Preview
Top

Introduction

A significant rise in deaths may be attributed to a lack of awareness about the warning signs of skin cancer and how it can be prevented. Skin cancer is regarded as one of the most toxic forms of cancer (Kadampur et al., 2020). Exposure to UV rays from the sun is a major risk factor for developing skin cancer (Vijayalakshmi, 2019). Possible causes of cancer include sun exposure, weakened immune systems, genetic predisposition, and other factors (Nahata & Singh, 2020). Benign and malignant conditions may exhibit this erratic cellular development pattern. The term “mole” is often used to refer to benign tumors, a subset of malignancy (Amin et al., 2020). Malignant tumors, on the other hand, are considered cancer and are aggressively treated. They pose a risk to the body’s other tissues (Kumar et al., 2020). The skin’s outer layer comprises three distinct cell types: basal cells, squamous cells, and melanocytes (Goyal et al., 2020). These cells are responsible for the tissues becoming cancerous. Different skin cancers are considered dangerous; they include basal cell carcinoma (BCC), squamous cell carcinoma (SCC), and melanoma (Ashraf et al., 2020). Initially, cancer cells appear as flat patches in the skin, often with scaly, rough, reddish, or brown surfaces (Adegun, 2020). A huge brown area with darker specks is a sign of skin cancer. Other signs are a mole that mutates in appearance, whether in size, texture, or blood supply; a small lesion with an uneven border and varying shades of color; or an itchy or burning sore that causes discomfort (Togacar et al., 2021). Therefore, early detection is crucial in treating skin cancer. To diagnose skin cancer, doctors often do a biopsy. New skin patches or existing patches that change in size, shape, or color warrant a medical evaluation. Skin cancer can develop from any abnormality in the skin, including a sore, tumor, blemish, marking, or change in the skin’s appearance or texture. This treatment aims to get a tissue sample from a suspicious skin lesion to evaluate its malignant potential. The procedure is tedious, lengthy, and time-consuming (Khan et al., 2021).

In the current diagnostic system, image processing techniques are commonly employed to bring illness identification methods into remote treatment (Khan et al., 2019). Image processing techniques use one or multiple computer fusion models to examine melanoma and suspected lesions (Jinnai et al., 2020). To make this handcrafted imaging method fully automatic and more intelligent, academics implement deep learning (DL) models, where an extensive data set of skin lesion images is used for testing and training the multilayer neural network (Adegun, 2021). As a result, computer-based technology delivers a less expensive, more comfortable, and early diagnosis of skin cancer signs (Rehman et al., 2020). One of the best methods to accurately and swiftly identify skin cancer is using DL (Saba et al., 2019). Human diseases, treatments, symptoms, signs, and aberrant investigation findings are all cataloged and linked together in a diseases database. Hence, the database decides which pictures to identify with which diseases based on findings.

Recently, convolutional neural network (CNN) technology has been extensively used in medicinal image processing, particularly for therapeutic image segmentation (Wei et al., 2020). These CNN-based approaches can be categorized by pixels to differentiate background substances from foreground substances to attain the last segmentation (Fraiwan et al., 2022). The chameleon swarm algorithm (CSA) is a recently devised metaheuristic algorithm inspired by chameleons’ intellectual behavior in nature (Malibari et al., 2022). CSA is a bottom-up clustering technique used in data mining to cluster databases using dynamic models; CSA can resolve both constrained and global numerical optimization issues (Thurnhofer-Hemsi & Dominguez, 2021). The CSA performance is evaluated using the database containing real-world applications (skin cancer diagnosis) for feature selections (Ali et al., 2022).

This paper includes three major contributions:

Complete Article List

Search this Journal:
Reset
Volume 19: 6 Issues (2023)
Volume 18: 4 Issues (2022): 2 Released, 2 Forthcoming
Volume 17: 4 Issues (2021)
Volume 16: 4 Issues (2020)
Volume 15: 4 Issues (2019)
Volume 14: 4 Issues (2018)
Volume 13: 4 Issues (2017)
Volume 12: 4 Issues (2016)
Volume 11: 4 Issues (2015)
Volume 10: 4 Issues (2014)
Volume 9: 4 Issues (2013)
Volume 8: 4 Issues (2012)
Volume 7: 4 Issues (2011)
Volume 6: 4 Issues (2010)
Volume 5: 4 Issues (2009)
Volume 4: 4 Issues (2008)
Volume 3: 4 Issues (2007)
Volume 2: 4 Issues (2006)
Volume 1: 4 Issues (2005)
View Complete Journal Contents Listing