PERBANDINGAN KINERJA ARSITEKTUR RESNET-50 DAN GOOGLENET PADA KLASIFIKASI PENYAKIT ALZHEIMER DAN PARKINSON BERBASIS DATA MRI

PIETERSZ, SHAWN HAFIZH ADEFRID (2024) PERBANDINGAN KINERJA ARSITEKTUR RESNET-50 DAN GOOGLENET PADA KLASIFIKASI PENYAKIT ALZHEIMER DAN PARKINSON BERBASIS DATA MRI. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 19081010055.-cover.pdf Download (1MB)
	Text (Bab 1) 19081010055.-bab1.pdf Download (198kB)
	Text (Bab 2) 19081010055.-bab2.pdf Restricted to Repository staff only until 21 June 2026. Download (801kB)
	Text (Bab 3) 19081010055.-bab3.pdf Restricted to Repository staff only until 21 June 2026. Download (734kB)
	Text (Bab 4) 19081010055.-bab4.pdf Restricted to Repository staff only until 21 June 2026. Download (3MB)
	Text (Bab 5) 19081010055.-bab5.pdf Download (189kB)
	Text (Daftar Pustaka) 19081010055.-daftarpustaka.pdf Download (157kB)

Abstract

Alzheimer's and Parkinson's disease are neurodegenerative diseases related to the brain. Alzheimer's disease causes a decline in cognitive function and behavior. Meanwhile, Parkinson's disease causes motor and non-motor impairments. Both diseases have a significant impact on patients' health and quality of life with the impact increasing in recent years. The causes of both diseases are still not known clearly and in detail as there is no specific test used in detecting both diseases. However, one way to detect these diseases is by performing Magnetic Resonance Imaging (MRI) scans, an imaging technique that is widely used to examine human brain activity. In helping this technology, an approach can be taken using the Convolutional Neural Network (CNN) algorithm with the aim of helping classification using MRI scan images. Based on this explanation, the author classifies MRI scan images of patients with Alzheimer's and Parkinson's disease using two CNN models, namely ResNet50 and GoogLeNet to determine the performance of the two models. This research identifies the best model performance using data splitting parameters 60:20:20 (5400 training data, 1800 validation data, 1800 testing data), optimizer Adam, epoch 64, and batch size 20. Based on this configuration, the ResNet-50 model gets the highest accuracy rate reaching 83.55%. While the GoogLeNet model gets an accuracy rate of 79.27%. The model that has been created is able to distinguish MRI images of patients with Alzheimer's disease, healthy patients from Alzheimer's and Parkinson's disease, and patients with Parkinson's disease well.

Item Type:

Thesis (Undergraduate)

Contributors: