Text

SEGMENTASI TUMOR OTAK MENGGUNAKAN ARSITEKTUR 3D DENSE-INCEPTION U-NET PADA CITRA HASIL MAGNETIC RESONANCE IMAGING OTAK

---

Penilaian

---

Brain tumor is a condition of abnormal growth of brain cells that can lead to death. Early detection is important for tumor patterns in Magnetic Resonance Imaging (MRI) images. Early detection is done by separating tumor cell details through 3D image segmentation with labels of background, non-enhancing tumor, peritumeral edema, and enhancing tumor. 3D image segmentation can be done automatically by utilizing the Convolutional Neural Network (CNN) method. The CNN architecture that is often used in 3D image segmentation is the 3D U-Net architecture. In the 3D U-Net encoder, the feature extraction process consists of a convolution block and a downsampling process. Downsampling reduces the spatial dimension which causes loss of detail information and vanishing gradient. This research proposes a modified 3D U-Net architecture with the addition of Dense Block and Block Inception in the encoder part to overcome vanishing gradient without increasing the number of parameters. The average performance results on accuracy, sensitivity, specificity, IoU, and f1-score with respective values of 99.05%, 93.37%, 98.69%, 87.7%, 92.4%. The results obtained accuracy performance on each label reached more than 95%, indicating the architecture was able to perform segmentation very well. Sensitivity and f1-score values on tumor enhancing labels are still below 90%, which means the model is still limited in detecting these parts optimally. The IoU value on the background label reaches more than 90%, while the non-enhancing tumor, peritumeral edema, and enhancing tumor labels are still below 90%, indicating that the tumor area prediction is smaller than the ground truth. The results of this study show that the 3D Dense-Inception U-Net architecture is able to perform segmentation well. However, IoU enhancement is needed to improve the quality of brain tumor segmentation.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2507001933	T169758	T1697582025	Central Library (Reference)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T1697582025

Publisher

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

Collation

viii, 63 hlm.; 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