Skripsi

SEGMENTASI PEMBULUH DARAH PADA CITRA RETINA MENGGUNAKAN KOMBINASI ARSITEKTUR U-NET, RESNET, DAN BIDIRECTIONAL CONVOLUTIONAL LONG SHORT TERM MEMORY

---

Penilaian

---

The addition of layers to the U-Net architecture can cause more parameters and redundant computation. The use of skip connection in the ResNet architecture can solve this problem, but it can cause the loss of important features from the previous layer. Bi-directional Convolutional Long Short Term Memory (Bi-ConvLSTM) can store previous and current information for a long time so that the missing features of the ResNet architecture can be overcome. The purpose of this study was to determine the performance evaluation results of combination architectures of U-Net, ResNet, and Bi-ConvLSTM on retinal blood vessels segmentation. The evaluation measures used were accuracy, sensitivity, specificity, F1-Score, and IoU. The data used in this study is the DRIVE dataset which is divided into 20 training data and 20 test data. The methods used in this study are data collection, data pre-processing, architecture implementation, training, testing, and evaluation. The values of accuracy, sensitivity, specificity, F1-Score, and IoU obtained were 95.56%, 79.24%, 97.22%, 76.69%, and 62.18%, respectively. Based on these results, we can conclude that the proposed architecture has been successfully performed segmentation of retinal blood vessels and predicts background pixels very well, indic ated by the accuracy and specificity values above 90%. In addition, it is well enough in predicting retinal blood vessels and the harmonization between the sensitivity and specificity values indicated by the sensitivity and F1-Score values above 70%, but the similarity between the image segmentation results and ground truth is still not good, which is below 70%.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2107003454	T56091	T560912021	Central Library (REFERENCES)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T560912021

Publisher

Inderalaya : Prodi Ilmu Matematika, Fakultas Matematika dan Ilmu Pengetahuan Alam., 2021

Collation

xiii, 96 hlm. : ilus. ; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

519.707

Content Type

Text

Media Type

unmediated

Carrier Type

-

Edition

-

Subject(s)

Prodi Ilmu Matematika
Pembuluh Darah

Specific Detail Info

-

Statement of Responsibility

DS

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment