Text

SEGMENTASI PEMBULUH DARAH PADA CITRA RETINA DENGAN MENGGUNAKAN MASK-RCNN (REGION CONVOLUSIONAL NEURAL NETWORK)

---

Penilaian

---

Blood vessels in the retina have a lot of information related to human health. Image analysis on the retina is one of the first steps in the main diagnosis of the disease. This research will focus on the process of segmentation of blood vessels in retinal images. The first step is to prepare the data using the DRIVE and STARE datasets, then filter the data. The second step is pre-processing the data by changing the DRIVE and STARE image formats, namely TIF and PPM, into JPG file formats to improve image quality. The third step is to perform data augmentation to multiply the data using the Horizontal Flip, Rotation technique. Translation, Zoom Range and Brightness. The fourth step is to process the image label to form a blood vessel object in the retina, then give a class, namely Blood Vessel. The next step is the segmentation process using the ResNet-101 architecture with the DRIVE and STARE datasets. The results obtained using the Epoch 500 parameter with a Learning Rate of 10−4 from the proposed method on the DRIVE dataset are 86.06% accuracy, 76.69% precision, 73.69% sensitivity, 61.71% IOU and 0.69% MAP, while the results obtained in the STARE dataset are 87.01 % accuracy, precision 78.09%, sensitivity 75.95% IOU 73.58% and MAP 0.72%.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2207003873	T79507	T795072022	Central Library (Referens)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T795072022

Publisher

Inderalaya : Jurusan Sistem Komputer, Fakultas Ilmu Komputer Universitas Sriwijaya., 2022

Collation

xv, 41 hlm.; ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

006.754 07

Content Type

-

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Jurusan Sistem Komputer
Situs Jejaring Sosial

Specific Detail Info

-

Statement of Responsibility

SEPTA

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment