Klasifikasi Kanker Payudara Berdasarkan Citra Ultrasound Menggunakan Vision Transformer dan Contrastive Learning

Zuhla, Fazhira Farizatuz (2023) Klasifikasi Kanker Payudara Berdasarkan Citra Ultrasound Menggunakan Vision Transformer dan Contrastive Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 06111940000036-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2025. Download (4MB) | Request a copy

Abstract

Berdasarkan World Health Organization (WHO), kanker payudara menempati urutan pertama kasus kanker yang paling banyak dialami di dunia dan tanpa terkecuali di Indonesia. Kanker payudara menjadi penyebab kematian urutan kelima pada kasus kanker di dunia. Pemeriksaan dini diperlukan guna mencegah kanker menjadi semakin ganas. Pemeriksaan dini yang dilakukan dapat secara efektif meningkatkan kemungkinan bertahan hidup pasien sebesar 90%. Salah satu prosedur dalam pemeriksaan adanya kanker payudara adalah melalui screening menggunakan citra ultrasound. Screening menggunakan citra ultrasound memiliki keunggulan yaitu dapat membedakan antara benjolan yang berisi cairan (benign)/tumor jinak) dan benjolan padat (malignant/tumor ganas). Perkembangan teknologi berbasis jaringan syaraf tiruan terutama dalam proses klasifkasi citra telah meningkat pesat. Salah satu arsitektur jaringan syaraf tiruan yang memiliki keunggulan dalam klasifkasi citra adalah Vision Transformer (ViT). ViT menjadikan potongan-potongan dari suatu citra sebagai representasi ftur. Pada penelitian ini dilakukan klasifkasi kanker payudara berdasarkan citra ultrasound menggunakan vision transformer dan contrastive learning (ViT-CL). Penelitian dilakukan pada dataset citra ultrasound payudara dengan 3 kelas yakni normal, tumor jinak (benign), dan tumor ganas (malignant). Penelitian ini menghasilkan performa Model ViT-CL pada model dengan arsitektur pre trained ViT-B/16 yakni akurasi sebesar 82%, presisi sebesar 81%, recall sebesar 86%, dan F1-score sebesar 83%.
===============================================================================================================================
According to the World Health Organization (WHO), breast cancer ranks as the most common cancer case in the world, and also in Indonesia. Breast cancer is the fifth leading cause of cancer death in the world. Early screening is necessary to prevent cancer from becoming more malignant. Early screening can effectively increase the patient’s chance of survival by up to 90%. One of the procedures in breast cancer screening is through screening using ultrasound images. Ultrasound images has the advantage of being able to distinguish between fluid-filled lumps (benign tumours) and solid lumps (malignant tumours). The development of artificial neural network-based technology, especially in the process of image classification, has increased rapidly. One artificial neural network architecture that has advantages in image classification is the Vision Transformer (ViT). ViT makes patch of images as feature representations. In this study, the classification of breast cancer based on ultrasound images using vision transformer and contrastive learning (ViT-CL) was conducted. The research was conducted on a dataset of ultrasound image dataset with 3 classes: normal, benign, and malignant. This research resulted in the best ViT-CL model performance in the model with ViT-B/16: accuracy 82%, precision 81%, recall 86%, and F1-score of 83%.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Contrastive Learning, Kanker Payudara, Klasifkasi, Vision Transformer, Breast Cancer, Classifcation, Contrastive Learning, Vision Transformer
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Fazhira Farizatuz Zuhla
Date Deposited:	30 Aug 2023 07:33
Last Modified:	30 Aug 2023 07:33
URI:	http://repository.its.ac.id/id/eprint/103759

Actions (login required)

View Item