Klasifikasi Tingkat Keganasan Kanker Paru-paru pada Citra Computed Tomography (CT) Scan Menggunakan Metode Convolutional Neural Network = Classification Level Of Cancer Rate Rations In Computed Tomography (Ct) Scan Using Convolutional Neural Network Method

Anugerah, Andreas Galang (2018) Klasifikasi Tingkat Keganasan Kanker Paru-paru pada Citra Computed Tomography (CT) Scan Menggunakan Metode Convolutional Neural Network = Classification Level Of Cancer Rate Rations In Computed Tomography (Ct) Scan Using Convolutional Neural Network Method. Undergraduate thesis, InstitutTeknologiSepuluhNopember.

[img] Text
05111440000153-Undergraduate_Theses.pdf - Published Version
Restricted to Repository staff only

Download (2MB) | Request a copy

Abstract

Deteksi dini kanker akan memudahkan dalam proses pengobatan maupun tindak lanjut untuk proses penyembuhan sehingga dapat menyelamatkan jutaan nyawa di seluruh dunia setiap tahunnya. Diperlukan suatu metode yang dapat mengenali kanker paru-paru hanya dengan melalui citra Computed Tomography (CT) scan paru-paru seorang pasien. Salah satu metode pengenalan maupun klasifikasi citra yaitu metode Convolutional Neural Network (CNN). Tugas akhir ini mengimplementasikan metode CNN pada klasifikasi tingkat keganasan kanker paru-paru melalui citra CT scan seorang pasien. Dalam penyusunan Tugas Akhir ini dilakukan percobaan terhadap beragam arsitektur CNN dan optimasi yang biasa digunakan. Tugas Akhir ini membandingkan 4 arsitektur CNN untuk klasifikasi tingkat keganasan paru-paru, yaitu CanNet, LeNet, VGG16, dan VGG19. Dataset yang digunakan yaitu dataset LIDC-IDRI. Model akan dilatih menggunakan berbagai macam metode optimasi CNN. Model menghasilkan keluaran berupa kelas malignant atau kelas benign. Dari hasil uji coba, didapatkan nilai akurasi klasifikasi rata-rata terbesar pada data citra CT scan paru-paru yaitu 81.4% dengan menggunakan arsitektur LeNet. Arsitektur CanNet memiliki run time yang paling lama yaitu sekitar 34 jam. Penggunaan ukuran kernel terbukti berpengaruh terhadap run time sebuah arsitektur CNN. Metode optimasi Stochastic Gradient Descent (SGD) menjadi metode optimasi yang paling optimal dengan hasil akurasi 91.4%. Ketiga metode optimasi lainnya, yaitu Root Mean Square Propagation (RMSProp), Adaptive Gradient Algorithm (Adagrad), dan Adaptive Moment Estimation (Adam) tidak mengalami perubahan akurasi sejak epoch pertama. =================== Early detection of cancer is preferred for the treatment process as well as follow-up to the healing process so that it can save millions of lives around the world every year. It requires a method that can recognize lung cancer only through the image of a patient's computed tomography (CT) scan. One of recognition and image classification methods is Convolutional Neural Network (CNN). This final project implements the CNN method on the classification of lung cancer malignancy level using CT scan image. This Final Project, conducted experiments on various CNN architecture and optimization. This Final Project compares the 4 CNN architectures for the classification of lung malignancy rates, i.e. CanNet, LeNet, VGG16, and VGG19. The CT scan images used is from the LIDC-IDRI dataset. Models was trained using various methods of optimization for CNN. The model produces output in the form of malignant class or benign class. The experimental results achiev the greatest average classification accuracy value on the image of CT scan lung image is 81.4% using the LeNet architecture. The CanNet architecture has the longest run time of about 34 hours. The use of kernel size has been shown to affect the run time of a CNN architecture. Stochastic Gradient Descent (SGD) optimization method became the optimum optimization method with 91.4% accuracy. The three other optimization methods, Root Mean Square Propagation (RMSProp), Adaptive Gradient Algorithm (Adagrad), and Adaptive Moment Estimation (Adam), have not changed accurately since the first epoch

Item Type: Thesis (Undergraduate)
Additional Information: RSIf 005.133 Anu k-1
Uncontrolled Keywords: Convolutional Neural Network, Deep Learning, CT Scan Paru-paru, LIDC-IDRI,
Subjects: Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
T Technology > T Technology (General) > T57.5 Data Processing
T Technology > T Technology (General) > T58.5 Information technology. IT--Auditing
T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
Divisions: Faculty of Information and Communication Technology > Informatics > (S1) Undergraduate Theses
Depositing User: Andreas Galang Anugerah
Date Deposited: 08 Jan 2019 08:03
Last Modified: 08 Jan 2019 08:03
URI: http://repository.its.ac.id/id/eprint/53470

Actions (login required)

View Item View Item