Implementasi Mask-RCNN dengan Image Enhancement untuk Deteksi dan Segmentasi Ventricular Septal Defect pada Anak berdasarkan Video Ekokardiografi

Putri, Dwi Amalia (2026) Implementasi Mask-RCNN dengan Image Enhancement untuk Deteksi dan Segmentasi Ventricular Septal Defect pada Anak berdasarkan Video Ekokardiografi. Undergraduate thesis, UPN Veteran Jawa Timur.

This is the latest version of this item.

	Text (Cover) 22083010042.-cover.pdf Download (1MB)
	Text (Bab 1) 22083010042.-bab1.pdf Download (271kB)
	Text (Bab 2) 22083010042.-bab2.pdf Restricted to Repository staff only until 16 July 2029. Download (1MB)
	Text (Bab 3) 22083010042.-bab3.pdf Restricted to Repository staff only until 16 July 2029. Download (990kB)
	Text (Bab 4) 22083010042.-bab4.pdf Restricted to Repository staff only until 16 July 2029. Download (3MB)
	Text (Bab 5) 22083010042.-bab5.pdf Download (248kB)
	Text (Daftar Pustaka) 22083010042.-daftarpustaka.pdf Download (217kB)
	Text (Lampiran) 22083010042.-lampiran.pdf Restricted to Repository staff only until 16 July 2029. Download (363kB)

Abstract

In Indonesia, cases of Congenital Heart Disease (CHD) are estimated to occur in approximately 9 babies per 1000 live births and more than 20% of total cases of CHD in children are Ventricular Septal Defect (VSD). Children with VSD abnormalities are at risk of experiencing worse conditions if identified late. The limited number of pediatric cardiologists and the varying quality of echocardiography videos due to speckle noise from heart movement are two obstacles in diagnosing VSD. The Mask-RCNN method integrated with FNLM-CLAHE preprocessing to reduce noise and clarify object structures in the heart, is proposed in the development of a prediction model for cardiac ventricular objects. The Mask-RCNN model is capable of predicting multi-tasks, namely classification, detection, and segmentation, thus enabling comprehensive identification of VSD, septum, and ventricles. Using echocardiography videos in the apical four-chamber view, this study examines the effectiveness of the Mask RCNN model in various parameter configurations and combinations of image enhancement scenarios. Model performance is measured using the Average Precision (AP), Intersection over Union (IoU), and Dice Similarity Coefficient (DSC) metrics. The results showed that the Mask-RCNN model with ResNet101 backbone and FNLM-CLAHE image enhancement provided the best performance, with the highest bbox/mAP50 value of 78.97% and segm/mAP50 of 76.16%. FNLM-CLAHE had a positive effect on spatial prediction of VSD objects, with an IoU of 64.061% and a DSC of 72.508%. The evaluation demonstrates that the model is capable of accurately predicting the location and shape of each object within the ventricle, including the VSD class. These findings indicate the importance of backbone components in image feature extraction and image quality improvement preprocessing methods, to improve prediction accuracy. The results of this study can be a reference for the development of VSD prediction models based on ultrasound images in children in identifying ventricular structures in heart movement.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Muhaimin, Amri	NIDN0023079502	amri.muhaimin.stat@upnjatim.ac.id
Thesis advisor	Nasrudin, Muhammad	NUPTK4241774675130323	nasrudin.fasilkom@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA76.6 Computer Programming
R Medicine > R Medicine (General)

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Dwi Amalia Putri

Date Deposited:

16 Jul 2026 03:47

Last Modified:

16 Jul 2026 03:55

URI:

https://repository.upnjatim.ac.id/id/eprint/55704