ANALISIS DISTRIBUSI SPASIAL TUMOR HATI PADA CITRA CT SCAN BERDASARKAN HASIL SEGMENTASI MENGGUNAKAN NNU-NET

Setiawan Putri, Sasikirana Ramadhanty and Kuswardani, Dwina (2026) ANALISIS DISTRIBUSI SPASIAL TUMOR HATI PADA CITRA CT SCAN BERDASARKAN HASIL SEGMENTASI MENGGUNAKAN NNU-NET. Diploma thesis, Institut Teknologi PLN.

[thumbnail of COVER.pdf] Text
COVER.pdf

Download (139kB)
[thumbnail of LEMBAR PENGESAHAN.pdf] Text
LEMBAR PENGESAHAN.pdf

Download (308kB)
[thumbnail of ABSTRAK.pdf] Text
ABSTRAK.pdf

Download (202kB)
[thumbnail of BAB I.pdf] Text
BAB I.pdf

Download (245kB)
[thumbnail of BAB II.pdf] Text
BAB II.pdf

Download (865kB)
[thumbnail of BAB III.pdf] Text
BAB III.pdf
Restricted to Registered users only

Download (1MB)
[thumbnail of BAB IV.pdf] Text
BAB IV.pdf
Restricted to Registered users only

Download (2MB)
[thumbnail of BAB V.pdf] Text
BAB V.pdf
Restricted to Registered users only

Download (197kB)
[thumbnail of DAFTAR PUSTAKA.pdf] Text
DAFTAR PUSTAKA.pdf

Download (177kB)
[thumbnail of SKRIPSI_202231107_SASIKIRANA RAMADHANTY SETIAWAN PUTRI.pdf] Text
SKRIPSI_202231107_SASIKIRANA RAMADHANTY SETIAWAN PUTRI.pdf
Restricted to Registered users only

Download (6MB)

Abstract

Segmentasi tumor hati pada citra CT scan merupakan salah satu permasalahan dalam bidang analisis citra medis karena variasi ukuran, bentuk, lokasi, dan karakteristik intensitas tumor yang menyebabkan batas antara area tumor dan jaringan hati normal sulit dibedakan. Selain segmentasi, hasil segmentasi tumor dapat dimanfaatkan untuk memperoleh informasi tambahan mengenai karakteristik spasial, seperti jumlah lesi, pola persebaran, dan posisi relatif tumor. Penelitian ini bertujuan untuk menganalisis distribusi spasial tumor berdasarkan hasil segmentasi nnU-Net. Dataset yang digunakan adalah Liver Tumor Segmentation (LiTS) Dataset yang terdiri dari citra CT scan hati beserta anotasi ground truth berupa mask segmentasi tumor. Dataset berasal dari LiTS Part 1 dan Part 2 dengan total 131 data pasien, kemudian dilakukan seleksi berdasarkan anotasi tumor sehingga diperoleh 118 data pasien yang memiliki area tumor. Pada penelitian ini digunakan 50 data pasien untuk proses pelatihan dan analisis menggunakan skema 5-Fold cross-validation yang memang berdasarkan otomatis nnunet. Tahapan penelitian meliputi preprocessing, pelatihan model nnU-Net 2D, prediksi segmentasi tumor, serta analisis hasil segmentasi menggunakan Connected Component Labeling (CCL) dan pendekatan spasial berbasis koordinat voxel. Evaluasi segmentasi dilakukan menggunakan Dice Similarity Coefficient (DSC) dan Intersection over Union (IoU). Hasil evaluasi menunjukkan nilai rata-rata DSC tumor sebesar 50,52% dan IoU sebesar 39,33%. Hasil segmentasi selanjutnya digunakan untuk menganalisis distribusi spasial tumor melalui identifikasi jumlah lesi, persebaran tumor pada setiap slice CT scan, serta posisi spasial tumor berdasarkan koordinat centroid. Penelitian ini menunjukkan bahwa hasil segmentasi nnU-Net dapat dimanfaatkan sebagai dasar untuk memperoleh informasi karakteristik spasial tumor hati pada citra CT scan.

Liver tumor segmentation in CT scan images is a challenging problem in medical image analysis due to the high variability in tumor size, shape, and intensity, which obscures the boundaries between tumors and normal tissue. Considering that segmentation masks can be further explored to extract valuable spatial information, this study primarily aims to analyze the spatial distribution and characteristics of liver tumors by utilizing the output of automated segmentation using the nnU-Net architecture. This study uses the Liver Tumor Segmentation (LiTS) dataset, in which, out of a total of 131 patient data, 118 patients confirmed to have tumor areas were selected, and 50 of these data were further processed for experimentation. The segmentation process consists of preprocessing, training, and prediction using a 2D nnU-Net model, utilizing the built-in 5-Fold cross-validation scheme of the framework. Evaluation of the segmentation model performance resulted in an average Dice Similarity Coefficient (DSC) of 50.52% and an Intersection over Union (IoU) of 39.33%. The prediction results were then used as the primary computational basis for spatial analysis using the Connected Component Labeling (CCL) method and a voxel-coordinate-based approach. Through these stages, the spatial distribution of tumors was comprehensively mapped, including the identification of the number of lesions, the range of tumor distribution across each CT scan slice, and the determination of spatial position based on centroid coordinates. These results demonstrate that nnU-Net-based segmentation can be effectively utilized to further characterize the spatial characteristics of liver tumors.

Item Type: Thesis (Diploma)
Uncontrolled Keywords: Deep Learning, nnU-Net, Segmentasi Tumor Hati, CT Scan, Distribusi Spasial, Connected Component Labeling.
Subjects: Bidang Keilmuan > Deep learning
Skripsi
Bidang Keilmuan > Teknik Informatika
Bidang Keilmuan > Information Technology
Divisions: Fakultas Telematika Energi > S1 Teknik Informatika
Depositing User: FAK TELEMATIKA ENERGI
Date Deposited: 24 Aug 2026 06:55
Last Modified: 24 Aug 2026 06:55
URI: https://repository.itpln.ac.id/id/eprint/7106

Actions (login required)

View Item
View Item