Classification of melanoma from Dermoscopic data using machine learning techniques

Melanoma is a skin disorder, occurring in melanocytes. They are classified as Benign and Malignant. The cure of melanoma is effective, if it can be recognized early. The most crucial part in the cure of melanoma is the exact classification and determining the group of melanoma. A comparative study f...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications 2020-02, Vol.79 (5-6), p.3713-3728
Main Authors: Janney.J, Bethanney, Roslin, S.Emalda
Format: Article
Language:English
Subjects:
Adaptive algorithms
Artificial intelligence
Classification
Classifiers
Cluster analysis
Clustering
Comparative studies
Computer Communication Networks
Computer Science
Data Structures and Information Theory
Decision trees
Feature extraction
Image classification
Image enhancement
Machine learning
Masking
Melanoma
Multilayers
Multimedia Information Systems
Special Purpose and Application-Based Systems
Support vector machines
Vector quantization
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Melanoma is a skin disorder, occurring in melanocytes. They are classified as Benign and Malignant. The cure of melanoma is effective, if it can be recognized early. The most crucial part in the cure of melanoma is the exact classification and determining the group of melanoma. A comparative study for classifying the group of melanoma using the supervised machine learning algorithms is discussed in this proposed work. Classification of melanoma from dermoscopic data is proposed to help the clinical utilization of dermatoscopy imaging methods for skin sores classification. The images were enhanced using anisotropic diffusion filter and unsharp masking. The melanoma was segmented from the background using adaptive k-means clustering algorithm with two clusters followed by feature extraction methods are based on intensity and texture features from the segmented data, which is followed by training of classifier and finally testing on unknown dermoscopic data. Classifiers such as k-nearest neighbour, support vector machine, multi-layer perceptron, decision tree and random forest were used. To test the performance of the classifiers, the area under the receiver operating characteristics curve (ROC) is utilized. The Random forest method is found to achieve 93% accuracy and classifies melanoma significantly good as compared to other classifiers.
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-018-6927-z