SULAKSONO, Nanang and Adi, Kusworo and Isnanto, R. Rizal (2025) PENGEMBANGAN MODEL CONVOLUTIONAL NEURAL NETWORK UNTUK MENDETEKSI PENYAKIT GINJAL BERDASARKAN CITRA AXIAL COMPUTED TOMOGRAPHY (CT) SCAN. Doctoral thesis, UNIVERSITAS DIPONEGORO.

	Text Cover.pdf Download (186kB)
	Text COVER DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Restricted to Repository staff only Download (987kB)
	Text BAB I DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Download (96kB)
	Text BAB II DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Download (604kB)
	Text BAB III DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Restricted to Repository staff only Download (403kB)
	Text BAB IV DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Restricted to Repository staff only Download (2MB)
	Text BAB V DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Restricted to Repository staff only Download (73kB)
	Text DAFTAR PUSTAKA DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Download (227kB)
	Text LAMPIRAN DISERTASI DSI NAMA NANANG SULAKSONO (NIM 30000321520031).pdf Restricted to Repository staff only Download (3MB)

Abstract

Penyakit ginjal kronis (PGK) merupakan masalah kesehatan masyarakat global dengan prevalensi angka kematian yang tinggi. Deteksi penyakit ginjal pada medis masih terdapat permasalahan, di antaranya; Keterbatasan jumlah dokter spesialis radiologi di rumah sakit, membutuhkan waktu dalam menentukan penyakit ginjal dalam pembacaan citra Computed Tomography (CT) Scan. Oleh karena itu perlunya sistem kecerdasan buatan dengan menggunakan Convolution Neural Network (CNN) dalam membantu deteksi penyakit ginjal kista, tumor, batu dan normal secara otomatis. Tujuan penelitian untuk mengembangkan dan mengimplementasikan model sistem deteksi penyakit ginjal tumor, kista dan batu pada ginjal dalam peningkatan akurasi diagnosis kesehatan. Penelitian menggunakan model algoritma CNN Keras Tensor Flow python programming, dengan menggunakan data citra CT Scan irisan axial. Pra pengolahan dilakukan dengan menggunakan ROI, SMOTE dan tomek link. Dataset citra CT Scan yang digunakan dari data publik Kaggle sebesar kista 2247, tumor 1339, batu 848 dan ginjal normal 3284. Sedangkan data klinis atau primer diambil dari rumah sakit sebesar kista 602, tumor 494, batu 446 dan ginjal normal 540. Penelitian ini meliputi masukan, proses, dan keluaran. Penelitian melibatkan para pakar ahli Dokter Spesialis Radiologi dalam menilai jenis penyakit ginjal di antaranya kista, tumor, batu dan ginjal normal. Hasil penelitian pengujian dengan menggunakan arsitektur densenet201, inceptionV3, inceptionresnetV2, resnet152V2, didapatkan hasil pengujian model terhadap penyakit ginjal kista, tumor, batu dan ginjal normal, pada individu dataset publik dan klinis didapatkan nilai akurasi 92%-100%. Pengujian model pemisahan dataset gabungan dan ROI dengan akurasi100%. Pengujian model ensemble (weighted voting) dataset gabungan, didapatkan hasil performa akurasi yang sangat baik pada data gabungan publik dan klinis 95%-100%. Penelitian berbeda dengan sebelumnya hanya deteksi batu ginjal, sedangkan penelitian ini dilakukan dengan mengembangkan sistem multi-klasifikasi untuk mendeteksi tumor, kista, batu dan ginjal normal berbasis citra axial CT Scan dengan optimasi CNN menggunakan ensemble learning. Sehingga penggunaan teknik ensemble (weighted voting) dataset gabungan dengan model densenet201, InceptionV3, inceptionresnetV2, dan resnet152V2, dapat direkomendasikan untuk mendeteksi penyakit ginjal diantaranya kista, tumor, batu, normal.
Kata Kunci: Pembelajaran mendalam; Jaringan saraf tiruan; Penyakit ginjal; Axial CT Scan, CNN

Chronic kidney disease (CKD) is a global public health problem with high prevalence and incidence of kidney failure. The field of radiology in detecting kidney disease still has problems, including; Limited number of radiology specialists in hospitals, requiring time in determining kidney disease in reading Computed Tomography (CT) Scan images. Therefore, an artificial intelligence system using Convolution Neural Network (CNN) is needed to help detect kidney disease cysts, tumors, stones and normal automatically. The purpose of the study is to develop and implement a kidney disease detection system model for tumors, cysts and stones in the kidneys in improving the accuracy of health diagnosis. The study used the CNN Keras Tensor Flow python programming algorithm model, using axial slice CT Scan image data. Pre-processing was carried out using ROI, SMOTE and tomek link. The CT Scan image dataset used from Kaggle public data was 2247 cysts, 1339 tumors, 848 stones and 3284 normal kidneys. While clinical or primary data was taken from hospitals amounting to 602 cysts, 494 tumors, 446 stones and 540 normal kidneys. This study includes input, process, and output. The study involved expert Radiology Specialists in assessing the types of kidney diseases including cysts, tumors, stones and normal kidneys. The results of the test study using the densenet201, inceptionV3, inceptionresnetV2, resnet152V2 architectures, obtained the results of model testing on kidney disease cysts, tumors, stones and normal kidneys, in individual public and clinical datasets obtained an accuracy value of 92% -100%. Testing the combined dataset separation model and ROI with an accuracy of 100%. Testing the combined dataset ensemble (weighted voting) model, obtained very good accuracy performance results on combined public and clinical data of 95% -100%. The study is different from the previous one only detecting kidney stones, while this study was conducted by developing a multi-classification system to detect tumors, cysts, stones and normal kidneys based on axial CT Scan images with CNN optimization using ensemble learning. So that the use of the combined dataset ensemble (weighted voting) technique with the densenet201, InceptionV3, inceptionresnetV2, and resnet152V2 models can be recommended for detecting kidney diseases including cysts, tumors, stones, normal. Keywords: Deep learning; Artificial neural network; Kidney disease; Axial CT Scan, CNN

Actions (login required)

View Item