Skripsi

SEGMENTASI TUMOR HATI PADA DATA CT-SCAN MENGGUNAKAN KOMBINASI DEEPLABV3+, EFFICIENTNET DAN U-NET DECODER

---

Penilaian

---

Liver tumor is a condition where there is abnormal cells that grow in the liver and causes the important role of the liver organ to be distrupted. The process of separating abnormal cells and healthy tissue in liver tumors can be done by image segmentation using Convolutional Neural Network (CNN). This study proposes liver tumor segmentation on CT-scan data using a combination of DeepLabV3+, EfficientNet and U-Net architectures. The DeepLabV3+ encoder uses the EfficientNet architecture as its backbone due to its ability to learn deep features, but with fewer parameters. In the decoder, modifications were made using the U-Net architecture. The U-Net architecture applies several skip connections taken at each layer to overcome the limitations of DeepLabV3+, which only uses one skip connection. The application of skip connections in several layers aims to restore fine details lost at each resolution level. The result of liver tumor segmentation using a combination of DeepLabV3+ architecture, EfficientNet and U-Net Decoder achieved an accuracy of 98,94%, indicating that the model is very good at predicting all labels correctly overall. Specificity of 95,36% and sensitivity of 89,98% indicate that the model is capable of predicting background and tumor areas very well. An f1-score of 87,16% indicates that the nodel has a good balance between sensitivity and specificity. An IoU of 77,25% indicates the model’s ability to display segmentation result that overlap with the ground truth. The results of this study indicate that the application of DeepLabV3+, EfficientNet and U-Net architecture provides good performance in liver tumor segmentation because it is able to distinguish between tumor areas and background areas.

---

Availability

Inventory Code	Barcode	Call Number	Location	Status
2507006482	T186184	T1861842025	Central Library (Reference)	Available but not for loan - Not for Loan

Detail Information

Series Title

-

Call Number

T1861842025

Publisher

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

Collation

xiii, 90 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

616.990 7

Content Type

Text

Media Type

unmediated

Carrier Type

other (computer)

Edition

-

Subject(s)

Penyakit Tumor
Prodi Ilmu Matematika

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment