KLASIFIKASI KANKER KULIT MELANOMA DENGAN PENGKLASIFIKASI BACKPROPAGATION NEURAL NETWORK
Total View This Week0
Total View This Week0
Institusion
Universitas Pembangunan Nasional Veteran Jakarta
Author
Pandu Ananto Hogantara, .
Subject
QA75 Electronic computers. Computer science
Datestamp
2021-12-21 07:51:33
Abstract :
Melanoma skin cancer is the deadliest type of skin cancer in the world. Melanoma skin cancer is almost impossible to cure when it has reached the final stages of cancer. Meanwhile, it is still curable in the early stages of the cancer. However, distinguishing benign lesions from melanoma cancer is hard because both have similar characteristics in the early stages of the development. The difficulty of distinguishing between benign lesions and melanoma cancer raises problems, namely underdiagnosis and overdiagnosis. Underdiagnosis is a condition in which melanoma is classified as a benign lesion, while overdiagnosis is a condition where a benign lesion is classified as a melanoma. Based on these problems, it is important to create an effective and accurate method that can classify melanoma cancer. This study uses the Gray Level Run Length Matrix (GLRLM) texture, HSV image color features, and shape features with backpropagation neural network to classify melanoma cancer. Based on this research, the color feature gets the best accuracy, sensitivity, and specificity among the three features, namely 77.33%, 80.33%, 74.33%. Meanwhile, the best overall accuracy was obtained when the three features were combined with the accuracy, sensitivity, and specificity score of 81.17%, 77.67%, 84.67%.
Melanoma skin cancer is the deadliest type of skin cancer in the world. Melanoma skin cancer is almost impossible to cure when it has reached the final stages of cancer. Meanwhile, it is still curable in the early stages of the cancer. However, distinguishing benign lesions from melanoma cancer is hard because both have similar characteristics in the early stages of the development. The difficulty of distinguishing between benign lesions and melanoma cancer raises problems, namely underdiagnosis and overdiagnosis. Underdiagnosis is a condition in which melanoma is classified as a benign lesion, while overdiagnosis is a condition where a benign lesion is classified as a melanoma. Based on these problems, it is important to create an effective and accurate method that can classify melanoma cancer. This study uses the Gray Level Run Length Matrix (GLRLM) texture, HSV image color features, and shape features with backpropagation neural network to classify melanoma cancer. Based on this research, the color feature gets the best accuracy, sensitivity, and specificity among the three features, namely 77.33%, 80.33%, 74.33%. Meanwhile, the best overall accuracy was obtained when the three features were combined with the accuracy, sensitivity, and specificity score of 81.17%, 77.67%, 84.67%.
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
