Saputra, Muhammad Dzikri Dwi (2020) Deteksi Kanker Kulit Melanoma Maligna Menggunakan Pengolahan Citra Digital Berbasis Metode Deep Learning. Tugas Akhir thesis, University of Technology Yogyakarta.
|
Text
Abstrak_5140711052_Muhammad Dzikri Dwi Saputra.pdf Download (280kB) | Preview |
Abstract
GLOBOCAN statistics data in 2012, a project of the International Agency for Research on Cancer (IACR), shows the average death rate from melanoma in Asia is 0.4-0.5 / 100,000, lower than Europe and North America. However, the mortality rate due to melanoma in men in Australia has increased over the past 30 years, plus clinical evidence of the poor prognosis of patients who have advanced melanoma, namely stage III or IV. Currently, cancer detection still uses biopsy techniques, which is less effective due to the length of time in determining the type of cancer. The solution offered in this study is the classification of cancer types using dermoscopy images to assist doctors in diagnosing cancer early. This study used a convolutional neural network method. The system classified two types of cancer, namely melanoma, and nevus. The system classified types of cancer through a dataset that conducted before training. The system predicted the accuracy of the type of cancer. The results of this study showed the classification of melanoma or nevus using the HAM1000 dataset. Furthermore, the accuracy of the training dataset was 83.07%, while validation was 79.25%. The test results of the dataset test obtained a 90% accuracy rate of the system, 100% precision, and 83% recall. Keywords: Convolutional Neural Network, Dermoscopy, Melanoma, Nevus
| Item Type: | Thesis (Skripsi, Tugas Akhir or Kerja Praktek) (Tugas Akhir) |
|---|---|
| Subjects: | T Technology > TK Electrical engineering. Electronics Nuclear engineering |
| Divisions: | Fakultas Sains Dan Teknologi > S1 Teknik Elektro |
| Depositing User: | Kaprodi Teknik Elektro |
| Date Deposited: | 19 Mar 2020 07:16 |
| Last Modified: | 19 Mar 2020 07:16 |
| URI: | http://eprints.uty.ac.id/id/eprint/4691 |
Actions (login required)
 |
View Item |
