Pramulen, Aji Sapta (2021) Rekonstruksi 3D Carotid Artery berbasis Segmentasi Citra 2D Ultrasound menggunakan U-NET. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
07111950050003-Master_Thesis.pdf Download (16MB) | Preview |
Abstract
Penyakit kardiovaskular merupakan penyakit yang menjadi penyebab kematian utama di seluruh dunia, dimana 17,3 juta kematian per tahun disebabkan Cardiovascular Disease(CVD) dan diperkirakan meningkat 23,6 juta pada tahun 2030. Carotid Intima-Media Thickness (CIMT) sebagai pertanda awal dari aterosklerosis, terbukti berhubungan dengan factor risiko terjadinya kardiovaskular. Penggunaan Ultrasound sangat dibutuhkan untuk melihat kelainan tersebut. Penelitian sebelumnya banyak melakukan segmentasi dari citra Ultrasound untuk mempermudah menentukan carotid. Penelitian ini melakukan segmentasi Carotid Artery pada citra Ultrasound dengan menggunakan arsitektur U-Net. Tahapan yang dilakukan dalam penelitian ini adalah melakukan preprocessing pada citra ultrasound yang kemudian melakukan training yang bertujuan untuk menghasilkan model segmentasi dengan arsitektur U-NET dan hasil segmentasi akan di visualisasi. Hasil dari training dihasilkan rata-rata akurasi 98,41% dengan dice coeficient 85,67% sedangkan untuk data testing dihasilkan akurasi 98,46% dengan dice coeficient 82,82% untuk citra tanpa ada peningkatan kualitas citra Non-Local Means-Based Speckle Filtering. Hasil training untuk citra yang menggunakan peningkatan citra adalah 98,42% akurasi dan 85,29% dice coeficient sedangkan data testing dihasilkan akurasi 98,42% dan dice coeficient 82,38%.
======================================================================================================
Cardiovascular disease is the leading cause of death worldwide, with 17.3
million deaths per year caused by Cardiovascular Disease (CVD) and an
estimated 23.6 million by 2030. Carotid Intima-Media Thickness (CIMT) as
an early sign of atherosclerosis is associated with cardiovascular risk factors.
The use of ultrasound is urgently needed to notice the abnormality. Previous
research has segmented ultrasound imagery to make it easier to determine
carotids. This study segmented Carotid Artery on Ultrasound imagery using
U-Net architecture. The stage of this research is to do preprocessing on
ultrasound imagery which then conducts training that aims to produce a
segmentation model with U-net architecture and segmentation result will be
in visualization. The result of training generated an average accuracy of
98,41% with dice coefficient 85,67% while for the testing data generated 98,42%
accuracy with dice coefficient 82,82% for imagery without any improvement
in image quality Non-Local Means-Based Speckle Filtering. Training results
for images using image enhancement were 98,42% accuracy and 85,29% dice
coefficient while testing data generated accuracy of 98,42% and dice coefficient
82,38%.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Kardiovaskular, U-Net, Segmentasi, Carotid Artery, Cardiovascular, Segmentation, U-Net, Carotid Artery |
| Subjects: | Q Science > QA Mathematics > QA336 Artificial Intelligence R Medicine > RZ Other systems of medicine T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis |
| Depositing User: | Aji Sapta Pramulen |
| Date Deposited: | 24 Feb 2021 07:59 |
| Last Modified: | 09 Jul 2024 16:12 |
| URI: | http://repository.its.ac.id/id/eprint/82769 |
Actions (login required)
 |
View Item |