Study Deteksi Dini Kanker Kulit Menggunakan Metode Region Growing Dan Deep Learning

yunandar, rian and Yusuf, Suhendro (2023) Study Deteksi Dini Kanker Kulit Menggunakan Metode Region Growing Dan Deep Learning. Masters thesis, IIB Darmajaya.

	Text Cover.pdf Download (314kB)
	Text Halaman Pernyataan.pdf Download (240kB)
	Text Halaman Persetujuan.pdf Download (269kB)
	Text Halaman Pengesahan.pdf Download (455kB)
	Text Abstrak.pdf Download (659kB)
	Text Daftar Isi.pdf Download (666kB)
	Text Bab I.pdf Download (847kB)
	Text Bab II.pdf Download (838kB)
	Text Bab III.pdf Download (1MB)
	Text Bab IV.pdf Download (7MB)
	Text Bab V.pdf Download (320kB)
	Text Daftar Pusaka.pdf Download (1MB)
	Text Lampiran.pdf Download (2MB)

Abstract

Skin cancer is a disease caused by mutations in skin cells. It is generally classified into two main categories: melanoma and nonmelanoma skin cancer. It is reported that skin cancer accounts for 5.9% to 7.8% of all cancer cases per year. The most common types of skin cancer in Indonesia are basal cell carcinoma (65.5%), followed by squamous cell carcinoma (23%), malignant melanoma (7.9%), and other types of skin cancer. Melanoma, in particular, can lead to high mortality rates, especially if not detected early. The problem in skin cancer detection lies in the manual detection method, which is not efficient and effective. Currently, detection is performed by experts through the examination of dermoscopy images of the skin as a basis for diagnosis. To address this issue, this research proposes a skin cancer image detection using the Region Growing segmentation method and Deep Learning Long Short-Term Memory (LSTM). In the first stage, oversampling is conducted to balance the amount of data in all classes, followed by data preprocessing before feeding them into the LSTM algorithm. For the Region Growing+LSTM algorithm, after the segmentation process, a highlight area segmentation is applied, followed by classification with LSTM. Both algorithms are evaluated using Accuracy, Precision, and Recall metrics, and the results are compared in terms of their evaluation performance and computational speed. The evaluation results show that the Region Growing+LSTM algorithm achieved an accuracy of 75%, while the LSTM algorithm achieved an accuracy of 54%. The training and prediction times for the LSTM algorithm to predict skin cancer images were 39.3 seconds and 3.2 seconds, respectively. On the other hand, the Region Growing+LSTM algorithm required 17 minutes and 2 seconds for training and 3 minutes and 49 seconds for prediction. In conclusion, the Region rowing+LSTM algorithm exhibits better accuracy but requires more time compared to LSTM on the skin cancer image dataset used in this research

Item Type:	Thesis (Masters)
Subjects:	Ilmu Komputer eTheses
Divisions:	Bahan Ajar > Buku
Depositing User:	rian yunandar
Date Deposited:	05 Nov 2024 07:54
Last Modified:	05 Nov 2024 07:54
URI:	http://repo.darmajaya.ac.id/id/eprint/18805

Actions (login required)

View Item