Klasifikasi Tumor Ginjal Benign Dan Malignant Pada Citra CT Scan Menggunakan Metode Deep Learning Berbasis CNN

Adini, Mahira Ilmi (2026) Klasifikasi Tumor Ginjal Benign Dan Malignant Pada Citra CT Scan Menggunakan Metode Deep Learning Berbasis CNN. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07311940000014_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Tumor ginjal merupakan salah satu masalah kesehatan dengan angka insidensi yang terus meningkat. Berdasarkan data Global Cancer Observatory (GLOBOCAN) tahun 2022, di Indonesia tercatat 2.676 kasus baru dengan angka kematian mencapai 1.563 kasus. Deteksi dini berperan penting dalam meningkatkan keberhasilan pengobatan dan menurunkan angka kematian. Penelitian ini bertujuan mengembangkan sistem klasifikasi bertingkat (multi-stage classification) pada citra Computed Tomography (CT) Scan ginjal serta mengevaluasi pengaruh metode pembagian data, yaitu image-based split dan patient-based split, terhadap performa model deep learning. Evaluasi dilakukan menggunakan Stratified 5-fold Cross Validation serta final testing pada data uji yang tidak digunakan selama pelatihan. Dataset dipraproses menggunakan ekstraksi Region of Interest (ROI) berbasis crop contour, diikuti proses resize dan normalisasi Model yang digunakan untuk pelatihan meliputi Custom CNN, VGG16, dan ResNet50. Sistem klasifikasi dilakukan secara bertingkat, yaitu tahap pertama mendeteksi citra normal dan tumor, kemudian hasil taha kedua mengklasifikasikan tumor menjadi benign dan malignant. Hasil penelitian menunjukkan bahwa Resnet50 memberikan performa terbaik pada image-based split dengan akurasi 92,55%, presisi 94,15%, recall 94,44%, dan F1-score 94,30%. Pada patient-based split, VGG16 menjadi model terbaik dengan akurasi 71,43%, presisi 75,81%, recall 81,59%, dan F1-score 78,59% pada tahap deteksi normal dan tumor serta VGG16 menunjukkan akurasi 70,95% pada tahap klasifikasi tumor ginjal benign dan malignant.
=====================================================================================================================================
Kidney tumors are one of the health problems with a continuously increasing incidence rate. According to data from the Global Cancer Observatory (GLOBOCAN) 2022, Indonesia recorded 2.676 new kidney cancer cases and 1.563 deaths. Early detection plays an important role in improving treatment outcomes and reducing mortality rates. This study aims to develop a multi-stage classification system for kidney Computed Tomography (CT) scan images and to evaluate the effect of data splitting strategies, namely image-based split and patient-based split, on the performance of deep learning models. The evaluation was conducted using Stratified 5-fold Cross Validation and final testing on an independent test set that was not used during the training process. The dataset was preprocessed through Region of Interest (ROI) extraction using a crop contour approach, followed by image resizing and normalization. Three deep learning models, namely Custom CNN, VGG16, and ResNet50, were employed for model training. The proposed multi-stage classification system consisted of 2 stages: the first stage distinguished normal and tumor images, while the second stage classified tumors into benign and malignant categories. The results showed that ResNet50 achieved the best performance on the image-based split, with an accuracy of 92,55%, precision of 94,15%, recall of 94,44%, and an F1-score of 94,30%. In the patient-based split, VGG16 achieved the best performance, with an accuracy of 71,3%, precision of 75,81%, recall of 81,59%, and an F1-score of 78,59% for normal and tumor detection, while achieving an accuracy of 70,95% for benign and malignant kidney tumor classification.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Tumor Ginjal, CT Scan, Deep Learning, VGG16, Kebocoran Data, Patient-Based Split Kidney Tumor, CT Scan, Deep Learning, VGG16, Data Leakage, Patient-Based Split
Subjects:	Q Science > QR Microbiology > QR 201.T84 Tumors. Cancer R Medicine > R Medicine (General) > R858 Deep Learning
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Biomedical Engineering > 11410-(S1) Undergraduate Thesis
Depositing User:	Mahira Ilmi Adini
Date Deposited:	03 Aug 2026 01:34
Last Modified:	03 Aug 2026 01:34
URI:	http://repository.its.ac.id/id/eprint/141287

Actions (login required)

View Item