Penerapan Segmentasi U-Net untuk Deteksi Abnormalitas pada Citra X-Ray Paru-Paru

Sinatria, Tatia Shafwa (2026) Penerapan Segmentasi U-Net untuk Deteksi Abnormalitas pada Citra X-Ray Paru-Paru. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22083010001 - Cover.pdf Download (1MB)
	Text (BAB I) 22083010001 - BAB I.pdf Download (242kB)
	Text (BAB II) 22083010001 - BAB II.pdf Restricted to Repository staff only until 24 September 2029. Download (540kB)
	Text (BAB III) 22083010001 - BAB III.pdf Restricted to Repository staff only until 24 September 2029. Download (487kB)
	Text (BAB IV) 22083010001 - BAB IV.pdf Restricted to Repository staff only until 24 September 2029. Download (3MB)
	Text (BAB V) 22083010001 - BAB V.pdf Download (234kB)
	Text (DAFTAR PUSTAKA) 22083010001 - DAFTAR PUSTAKA.pdf Download (215kB)
	Text (DAFTAR LAMPIRAN) 22083010001 - DAFTAR LAMPIRAN.pdf Restricted to Repository staff only until 24 September 2029. Download (303kB)

Abstract

Chest X-ray examination is one of the imaging methods commonly used to assist in assessing lung conditions. However, manually identifying lung regions and abnormalities in X-ray images can be time-consuming and may be influenced by subjective interpretation. This study aims to implement the U-Net architecture for lung segmentation in chest X-ray images and utilize the segmentation results as a Region of Interest (ROI) to support the analysis of abnormal areas. The research stages include image preprocessing, U-Net model development and training, generation of a probability map, threshold determination, binary mask generation, segmentation evaluation, and abnormality analysis using a statistical approach. The experimental results show that a threshold of 0.5 provides the best segmentation performance, achieving a Mean Dice score of 0.9106, a median of 0.9555, and a standard deviation of 0.0870. Further evaluation resulted in an IoU of 0.8464, Precision of 0.9084, Recall of 0.9211, Pixel Accuracy of 0.9516, and Specificity of 0.9637. These results indicate that the U-Net model is capable of segmenting lung regions with good agreement with the ground truth. The resulting segmentation mask is then used to form the lung ROI, allowing abnormality analysis to be focused on the lung region. Analysis using the Absolute Local Z-Score and the distribution of normal image data is used as the basis for determining deviation thresholds and generating abnormality maps. Overall, this study produces a U-Net-based lung segmentation pipeline capable of generating lung masks and ROIs while supporting the analysis of abnormal areas in chest X-ray images.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Saputra, Wahyu Syaifullah Jauharis	NIDN198608252021211003	wahyu.s.j.saputra.if@upnjatim.ac.id
Thesis advisor	Adziima, Andri Fauzan	NIDN199505122024061001	andri.fauzan.fasilkom@upnjatim.ac.id

Subjects:

Q Science > Q Science (General)
Q Science > QA Mathematics
Q Science > QA Mathematics > QA76.6 Computer Programming

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Tatia Shafwa Sinatria

Date Deposited:

24 Sep 2026 03:34

Last Modified:

24 Sep 2026 03:34

URI:

https://repository.upnjatim.ac.id/id/eprint/60340