Text

DETEKSI AREA LESI TERINFEKSI PADA KASUS PRA-KANKER SERVIKS MENGGUNAKAN PENDEKATAN SEMANTIK SEGMENTASI DAN FASTER-RCNN

---

Penilaian

---

Cervical pre-cancer is the initial occurrence of cancer and referred to as cervical lesions and carcinoma (CA). Before the formation of pre-cancerous cells (lesions), abnormal cells appear on the cervix which is the prelude to cervical pre-cancer. The squamous columnar junction (SCJ) is a transitional area of the vaginal mucosa and cervical mucosa that looks like a distinct line between areas of columnar epithelium and squamous. A computer-assisted diagnostic system solves these problems, namely medical image segmentation and object detection. This study uses the Convolutional Neural Network (CNN) method, namely segmentation using U-Net architecture, V-Net, and detection using Faster R-CNN architecture. The initial data and augmented data resulted in 24 models from the U-Net architecture, 24 models from the V-Net architecture, and 12 models from the Faster R-CNN architecture using data from the best model results obtained from segmentation detection on the U-Net architecture. The best model from U-Net CA (II) which gets the highest evaluation results is 99.34% Pixel Accuracy, 94.16% Intersection Over Union (IoU), and 93.22% F1 Score. Lesion (II) 98.83% Pixel Accuracy, 94.48% Intersection Over Union (IoU), and 93.29% F1 Score. While the best model of the V-Net CA (II) architecture got the highest evaluation results of 99.32 Pixel Accuracy, 93.85 Intersection Over Union (IoU), and 93.03 F1 Score. Lesions (II) 99.00% Pixel Accuracy, 94.70% Intersection Over Union (IoU), and 94.09% F1 Score. The best model from Faster R-CNN gets the highest evaluation result of mAP with a value of 86.07%. Results Based on the experiment of segmentation using U-Net and V-Net, the U-Net segmentation result model is better than the model obtained from segmentation using V-Net.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2207002049	T73717	T737172022	Central Library (Referens)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T737172022

Publisher

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

Collation

xviii, 116 hlm.; ilus.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

003.07

Content Type

-

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Jurusan Sistem Komputer
Sistem analisis

Specific Detail Info

-

Statement of Responsibility

SEPTA

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment