Segmentasi Gambar Ultrasound Arteri Radialis Menggunakan Convolutional Neural Network Untuk Akses Insersi Arteri

Yoga, Cokorda Gede Sedana (2020) Segmentasi Gambar Ultrasound Arteri Radialis Menggunakan Convolutional Neural Network Untuk Akses Insersi Arteri. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

Text 07311640000026-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Kateterisasi dan kanulasi melalui arteri radialis sudah menjadi prosedur umum yang dilakukan dokter spesialis jantung untuk masa perioperatif. Meskipun tingkat kesuksesan yang tinggi untuk para ahli berpengalaman menggunakan teknik palpasi, ada beberapa kasus yang mana secara teknis sedikit sulit (hipotensi dan obesitas). Komplikasi yang paling sering terjadi pada saat kateterisasi melalui arteri radialis adalah oklusi arteri sementara (19,7 %) dan hematoma (14,4%), dengan infeksi pada tempat pemasukan (1,3%), haemorrhage (0,53%), dan bacteremia (0,13%). Suatu sistem ultrasound menjadi pilihan untuk membantu visualisasi para ahli terkait kelebihannya dalam aspek kenyamanan, ekonomis, dan non-ionisasi. Sistem yang dibutuhkan saat ini adalah sistem segmentasi otomatis yang mana pada kali ini mengusulkan penggunaan metode deep learning convolution neural network (CNN) untuk mendapatkan visualisasi citra pembuluh darah arteri radialis sebagai alat dukung para ahli pada saat melakukan insersi ke intra-arterial. Dari hasil pengujian, didapatkan bahwa proses segmentasi citra mendapatkan nilai rata-rata dice similarity coefficient sebesar 0,9389576565621461 dan rata-rata nilai error 12,4757% pada dataset testing. Serta nilai rata-rata dice similarity coefficient sebesar 0,9699205117090076 dan rata-rata nilai error 3,70003% pada dataset phantom.
====================================================================================================================
Catheterization and cannulation through radial artery have become common procedures that performed by cardiologists for the perioperative period. While the success rate using palpation techniques, there are some problems with technical difficulties (hypotension and obesity). The most common complication during catheterization through this access were temporary arterial occlusion (19.7%) and hematoma (14.4%), with infection at the access site (1.3%), bleeding (0.13%). An ultrasound system is an alternative to help experts visualize in term of comfort, economics, and non-ionization. Automatic segmentation system is needed which at this time proposes the use of deep learning method called Convolution Neural Network (CNN) to obtain visualization of radial artery as a tool to support experts while inserting intra-arterial. From the test results, it was found that the image segmentation process gets an average value of dice similarity coefficient of 0.9389576565621461 and average error value of 12.4757% on the testing dataset. As well as the average value of the dice similarity coefficient of 0.9699205117090076 and the average error value of 3.70003% on the phantom dataset.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Artery segmentation, convolutional neural network, ultrasound system, convolutional neural network, segmentasi pembuluh darah, sistem ultrasound
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms
Divisions:	Faculty of Electrical Technology > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Cokorda Gede Sedana Yoga
Date Deposited:	25 Aug 2020 07:20
Last Modified:	12 Sep 2025 02:59
URI:	http://repository.its.ac.id/id/eprint/79860

Actions (login required)

View Item