Text

KOMBINASI ARSITEKTUR VISUAL GEOMETRY GROUP, BOTTLENECK LAYER, GATE RECURRENT UNITS, DAN ATTENTION GATE PADA KLASIFIKASI PENYAKIT KANKER SERVIKS MENGGUNAKAN CITRA PAP SMEAR

---

Superficial-intermediate, parabasal, koilocytotic, dyskeratotic, and metaplastic are types of cells in the cervix that are at risk of causing cancer. Cervical cancer can be detected early using several methods, one of which is the Pap smear. Early detection of this disease can be achieved through the use of deep learning methods. This study proposes a model combining the Visual Geometry Group Bottleneck Layer architecture, Gate Recurrent Units, and Attention Gate for cervical cancer classification using Pap smear images. In the first block, the VGG architecture is used to capture detailed features in complex features. In the 14th block, before classification, a Bottleneck Layer is added to reduce the number of parameters and prevent overfitting before entering the classification layer. Gate Recurrent Units are added after the Bottleneck Layer to reuse features that were discarded due to feature dimension reduction from the Bottleneck Layer, and an Attention Gate is inserted to focus the model on relevant features while filtering out irrelevant ones. In the last block, the model uses a fully connected layer and a softmax activation function to provide prediction results. The proposed model achieved an average accuracy of 99.6%, sensitivity of 95.4%, specificity of 99.7%, F1-score of 95.4%, and Cohen's kappa of 95.2%, indicating that the model is capable of classifying Pap smear images effectively. In this study, the proposed model performed very well in recognizing the superficial-intermediate class compared to the parabasal, koilocytotic, dyskeratotic, and metaplastic classes, with a sensitivity of 99.8%, indicating that the model is very good at recognizing images from the superficial-intermediate class. Although the values obtained for the parabasal, koilocytotic, dyskeratotic, and metaplastic classes are still below those of the superficial-intermediate class, the results obtained are already very good because the sensitivity values for the parabasal, koilocytotic, dyskeratotic, and metaplastic classes are 90%.

---

Availability

Detail Information

Series Title

-

Call Number

T1841022025

Publisher

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

Collation

xv, 75 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

512.507

Content Type

Text

Media Type

unmediated

Carrier Type

other (computer)

Edition

-

Subject(s)

Aljabar linear

Specific Detail Info

-

Statement of Responsibility

SEW

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment