Deteksi Out of Distribution (OOD) Menggunakan Extreme Value Theory (EVT) Pada Klasifikasi Penyakit Kulit

Yasin, Yordan (2020) Deteksi Out of Distribution (OOD) Menggunakan Extreme Value Theory (EVT) Pada Klasifikasi Penyakit Kulit. Undergraduate thesis, InstitutTeknologi Sepuluh Nopember.

[img] Text
07211640000013-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only

Download (9MB) | Request a copy

Abstract

Sampai saat ini tingginya tingkat penderita penyakit kulit dan tidak meratanya tenaga medis masih menjadi masalah utama dalam bidang kesehatan di Indonesia. Dengan kemajuan Teknologi Informasi dan Komunikasi, diagnosis penyakit kulit dapat dilakukan sedini mungkin agar pasien segera mendapatkan perawatan. Salah satunya adalah dikembangkan sebuah sistem klasifikasi penyakit kulit yang akan diterapkan pada teknologi Teledermatology. Sistem ini akan mengklasifikasi penyakit kulit pada citra dermoscopic menggunakan algoritma Deep Learning yaitu Convolutional Neural Network. Teknologi ini sudah mampu mengklasifikasikan 7 penyakit kulit yang termasuk dalam kategori kanker kulit dengan akurasi sebesar 72%. Namun, 7 penyakit yang dapat diklasifikasikan tersebut bukanlah penyakit yang umum terjadi di Indonesia. Selain itu Deep Learning yang digunakan pada sistem klasifikasi tersebut saat ini masih bersifat closed set dan belum dapat menanangani data gambar yang bersifat open set. Oleh karena itu, perlu sistem klasifikasi yang mampu mengklasifikasikan penyakit yang umum terjadi di negara tropis dan mendeteksi Out of Distribution (OOD) agar sistem mampu menangani data gambar yang bersifat open set dengan menggunakan Extreme Value Theory (EVT). =============================================================================================== Until now the high rate of sufferers of skin diseases and unequal medical personnel is still a major problem in health sector in Indonesia. With the advancement of Information and Communication Technology, the diagnosis of skin diseases can be made as early as possible so that patients immediately get treatment. False one of them was developed a skin disease classification system that will be applied to teledermatology technology. This system will classify skin diseases in dermoscopic images using the Deep Learning algorithm, Convolutional Neural Network. This technology is able to classify 7 diseases skin that is included in the category of skin cancer with accuracy by 72%. However, 7 diseases that can be classified are not a common disease in Indonesia. other than that Deep Learning used in the classification system is currently still closed set and cannot handle data images that are open set. Therefore, it needs a classification system that is able to classify common diseases that occur in tropical countries and detect Out of Distribution (OOD) so that the system is able to handle image data that is open set with using Extreme Value Theory (EVT).

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Penyakit Kulit, Deep Learning, Out of Distribution, Extreme Value Theory, Skin Disease, Deep Learning, Out of Distribution, Extreme Value Theory.
Subjects: Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
R Medicine > RL Dermatology
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User: Yordan Yasin
Date Deposited: 25 Aug 2020 03:55
Last Modified: 25 Aug 2020 03:55
URI: https://repository.its.ac.id/id/eprint/79703

Actions (login required)

View Item View Item