Skin Cancer Classification Using Dermoscopic Images based on Ranklet Transform, Co-occurrence Features and Random Forest Classifier

Authors

  • Chithra Devi M

DOI:

https://doi.org/10.37506/mlu.v20i3.1420

Keywords:

Dermoscopy images, Ranklet transform, Co-occurrence matrix, Random forest classifier

Abstract

The skin cell which grows abnormally is known as skin cancer and it is caused due to sun rays in the

uncovered skin. It spreads within a week, so the early diagnosis of skin cancer is required. Skin Cancer

Classification (SCC) based on ranklet transform, co-occurrence features and random forest classifier is

presented in this paper. Dermoscopic image in the PH2 database is used in this study for performance

evaluation. Initially, the dermoscopic images of three categories normal benign and malignant images are

preprocessed to smooth the images. Then the images are given to ranklet transform for decomposition. It

produces subband coefficients. The Ranklet Features based Co-occurrence Matrix (RFCM) is used to extract

the features and stored in database. The classification is made by Random Forest Classifier (RFC). The result

shows better classification accuracy of 93.5% sensitivity is 92% and specificity is 95% is obtained by RFCM

and RFC.

Author Biography

Chithra Devi M

Assistant Professor, Annai Vailankanni Arts and Science College, Thanjavur, (Affiliated to Bharathidasan

University)

Published

2020-07-24

How to Cite

Chithra Devi M. (2020). Skin Cancer Classification Using Dermoscopic Images based on Ranklet Transform, Co-occurrence Features and Random Forest Classifier. Medico Legal Update, 20(3), 344-350. https://doi.org/10.37506/mlu.v20i3.1420