Text

ARSITEKTUR VOLUMETRIC U-NET DAN TRANSFORMER UNTUK SEGMENTASI TUMOR OTAK PADA CITRA HASIL MAGNETIC RESONANCE IMAGING

---

Penilaian

---

Brain tumors are abnormal tissue growths within the brain that can lead to death. The components of a brain tumor can be classified into background, enhancing tumor, peritumoral edema, and non-enhancing tumor, which are identified in three-dimensional Magnetic Resonance Imaging (MRI) scans. The separation of these components can be achieved through automatic segmentation. This study proposes a combination of Vision Transformer (ViT) and Volumetric U-Net architectures for the segmentation of brain tumor components in MRI images. ViT is utilized in the encoder to capture global spatial relationships, while the decoder maintains the Volumetric U-Net structure to preserve local spatial details. The performance of the proposed architecture achieved accuracy, sensitivity, specificity, IoU, and f1-score values of 98.78%, 80.33%, 97.01%, 74.6%, and 83.3%, respectively. These results indicate a good performance in brain tumor segmentation from MRI images. At the label level, the model achieved accuracy, sensitivity, specificity, IoU, and f1-score values ranging from 78% to 98% for background, enhancing tumor, and peritumoral edema. However, the performance for the non-enhancing tumor label was relatively low, with an accuracy of 98.6%, sensitivity of 43.6%, specificity of 99.9%, IoU of 43.2%, and f1-score of 60.2%. This lower performance is attributed to the relatively small feature size and unclear boundaries of the non-enhancing tumor region. Based on these findings, future studies are encouraged to explore new approaches capable of better detecting small regions in 3D brain tumor segmentation.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2507003135	T174040	T1740402025	Central Library (Reference)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T1740402025

Publisher

: Prodi Ilmu Matematika, Fakultas Matematika Dan Ilmu Pengetahuan Alam Universitas Sriwijaya., 2025

Collation

xii, 77 hlm.; ill.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

510.07

Content Type

-

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Matematika

Specific Detail Info

-

Statement of Responsibility

EM

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment