DETEKSI ABNORMALITAS STRUKTUR JANTUNG ANAK DENGAN METODE MASK REGION CONVOLUTIONAL NEURAL NETWORK

MARSELA, ICHA DWI and Nurmaini, Siti (2023) DETEKSI ABNORMALITAS STRUKTUR JANTUNG ANAK DENGAN METODE MASK REGION CONVOLUTIONAL NEURAL NETWORK. Undergraduate thesis, Sriwijaya Univeristy.

	Text RAMA_56201_09011281924031.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (3MB) | Request a copy
	Text RAMA_56201_09011281924031_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (19MB) | Request a copy
	Text RAMA_56201_09011281924031_0002085908_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (527kB)
	Text RAMA_56201_09011281924031_0002085908_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (530kB) | Request a copy
	Text RAMA_56201_09011281924031_0002085908_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (705kB) | Request a copy
	Text RAMA_56201_09011281924031_0002085908_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_56201_09011281924031_0002085908_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (126kB) | Request a copy
	Text RAMA_56201_09011281924031_0002085908_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (134kB) | Request a copy
	Text RAMA_56201_09011281924031_0002085908_07_lamp.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (233kB) | Request a copy

Abstract

Detecting abnormal heart structures in infant is an important issue in the field of medicine. One method that can be used to detect abnormal heart structures in children is by using the Mask Region Convolutional Neural Network (Mask-RCNN) method. The process of training MRCNN in detecting abnormal heart structures in children can be done by utilizing well-segmented and labeled heart image datasets. After training is done, Mask-RCNN can be used to detect abnormal heart structures in heart images that have never been seen before.. The best Performance achieved in the segmentation of 7 heart structures was a DSC of 82.40%, an IoU of 70.43%, and an mAP of 90.79%. In the segmentation of 7 heart chambers and the septum, the best Performance achieved was 79.08% for DSC, 67.69% for IoU, and 89.60% for mAP. For the segmentation of Holes, the best model Performance on the 4CH data was 81.35%, 57.41%, and 89.9% for DSC, IoU, and mAP respectively. On the 5CH data, it was 79.23%, 65.72%, and 91.08%. For the Long Axis data, it was 80.13%, 48.98%, and 89.27%. For the Short Axis data, it was 72.83%, 56.32%, and 86.15%. Finally, for the Subcostal data, it was 76.42%, 60.72%, and 88.64% for each parameter of DSC, IoU, and mAP.By using the Mask-RCNN method, detection of abnormal heart structure in infants can be done accurately and efficiently. This method can help doctors in diagnosing and treating cases of abnormal heart structure in infants more quickly and accurately.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Abnormalitas, Deteksi, Infant, Instance Segmentation, Mask Region Convolutional Neural Network
Subjects:	Q Science > Q Science (General) > Q1-390 Science (General) > Q223.M517 Science -- Information services. Information storage and retrieval systems --Science.
Divisions:	09-Faculty of Computer Science > 56201-Computer Systems (S1)
Depositing User:	Icha Dwi Marsela
Date Deposited:	18 Apr 2023 04:08
Last Modified:	18 Apr 2023 04:08
URI:	http://repository.unsri.ac.id/id/eprint/96865

Actions (login required)

View Item