Skripsi

MULTI OBJECT TRACKING PADA CITRA JANTUNG ANAK MENGGUNAKAN DEEP LEARNING

---

Congenital heart disease (CHD) in children, such as atrial septal defects (ASD), ventricular septal defects (VSD), and atrioventricular septal defects (AVSD), requires accurate diagnosis through dynamic analysis. However, existing methods for analyzing echocardiographic video are often limited to frame-by-frame analysis and are not yet capable of consistently tracking temporal changes. This study proposes a multi-object tracking approach based on instance segmentation using deep learning to address these limitations. The proposed method integrates the You Only Look Once Version 8 (YOLOv8) and You Only Look Once Version 26 (YOLOv26) architectures to perform instance segmentation, which is then integrated with the ByteTrack algorithm for consistent object tracking in echocardiography videos. This approach was applied to video data of ASD, AVSD, and normal heart conditions. The results show that the instance segmentation model performance indicates YOLOv26 outperforms YOLOv8, with an average Mean Average Precision (mAP) of 88.7% for bounding boxes (mAP BBox) and 89.1% for masks (mAP Mask) in the Normal case. In cases of heart abnormalities, such as ASD and VSD, YOLOv26 also showed improved performance compared to YOLOv8, particularly in mask segmentation results. Meanwhile, in the multi-object tracking evaluation, the best performance was also achieved under Normal conditions, with a Multiple Object Tracking Accuracy (MOTA) of 59.1%, an ID F1 Score (IDF1) of 56.7%, and a Localization Accuracy (LocA) of 80.4%. These results indicate that the model is capable of tracking objects quite stably, especially under normal conditions, although performance in certain specific cases still shows variation.

---

Availability

Detail Information

Series Title

-

Call Number

T2002432026

Publisher

Indralaya : Prodi Sistem Komputer, Fakultas Ilmu Komputer Universitas Sriwijaya., 2026

Collation

xv, 104 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

006.307

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Sistem Komputer
Model Deep Learning

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment