Klasifikasi Citra X-Ray Covid-19 Menggunakan Ekstraksi Fitur HOG dan Transfer Learning CNN

Kembara, Bayu (2024) Klasifikasi Citra X-Ray Covid-19 Menggunakan Ekstraksi Fitur HOG dan Transfer Learning CNN. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025212003-Master_Thesis.pdf Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Penyakit Corona Virus Disease (COVID-19) adalah infeksi virus yang mempengaruhi sistem pernapasan dan telah masuk ke Indonesia dengan total kasus baru mencapai 1.831. Dalam rangka mengidentifikasi pasien COVID-19 atau pasien normal, penelitian ini memanfaatkan citra X-ray dada (chest X-ray atau CXR) yang diperoleh dari situs Kaggle.
Pendekatan arsitektur Convolutional Neural Networks (CNN) yang meliputi beberapa tahapan: penggunaan image enhancer CLAHE untuk meningkatkan kualitas citra, konversi citra ke grayscale, ekstraksi fitur menggunakan metode Histogram of Oriented Gradients (HOG), serta klasifikasi dengan menerapkan metode transfer learning. Arsitektur CNN ini memadukan teknik CLAHE dan grayscale untuk memperbaiki kualitas citra sebelum melaksanakan ekstraksi fitur, yang berperan penting dalam meningkatkan akurasi pengenalan pola pada citra.
Penelitian ini mengintegrasikan empat model populer: AlexNet, GoogleNet,ResNet-50, dan VGG-16, serta menggunakan perubahan warna grayscale dan penambahan image enhancer CLAHE dengan penambahan Optimizer Nadam dan Adam guna memaksimalkan performa akurasi model yang diusulkan. Hasil penelitian menunjukkan efektivitas model ini dalam klasifikasi citra X-ray COVID-19, dengan kemampuan membedakan antara pasien positif COVID-19 dan pasien normal.
Performa terbaik dicapai oleh model VGG-16 yang dioptimalkan dengan Nadam dan proses grayscale, menghasilkan nilai akurasi, presisi, recall, dan f1- score sebesar 90% (0,90). Selanjutnya, model GoogleNet dengan pengoptimalan serupa mencatat nilai akurasi, presisi, recall, dan f1-score sebesar 91% (0,91), menegaskan potensi besar pendekatan ini dalam mendukung diagnosa COVID-19 melalui analisis citra medis.
===================================================================================================================================
Coronavirus Disease (COVID-19) is a viral infection that affects the respiratory system. With a total of 1,831 new cases, COVID-19 has entered Indonesia. To identify COVID-19 patients or normal patients, this study utilizes chest X-ray images (CXR) obtained from the Kaggle website.
The approach involves using a Convolutional Neural Networks (CNN) architecture, which includes several stages: using the CLAHE image enhancer to improve image quality, converting the images to grayscale, extracting features using the Histogram of Oriented Gradients (HOG) method, and classification by applying transfer learning methods. This CNN architecture combines CLAHE and grayscale techniques to enhance image quality before feature extraction, which plays a crucial role in improving pattern recognition accuracy in the images.
This study integrates four popular models: AlexNet, GoogleNet, ResNet-50,and VGG-16. It utilizes grayscale conversion and the addition of the CLAHE image enhancer, along with the Nadam and Adam optimizers, to maximize the performance accuracy of the proposed models. The results demonstrate the effectiveness of these models in classifying COVID-19 X-ray images, with the ability to distinguish between COVID-19-positive patients and normal patients.
The GoogleNet model optimized with Nadam and grayscale processing achieved the best performance, resulting in accuracy, precision, recall, and f1- score of 90% (0.90). Furthermore, the Vgg-16 model with similar optimization recorded an accuracy, precision, recall, and f1-score of 91% (0.91), highlighting the significant potential of this approach in supporting COVID-19 diagnosis through medical image analysis

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Convolutional Neural Networks (CNN), Covid-19, Histogram of Oriented Gradients (HOG), Optimizer Nadam, Optimizer Adam, Transfer learning, X-Ray
Subjects:	Q Science > QM Human anatomy R Medicine > RC Internal medicine > RC346 Nervous system--Diseases--Prognosis. R Medicine > RC Internal medicine > RC771 Pneumonia. R Medicine > RC Internal medicine > RC78 Diagnosis, Radioscopic--Examinations, questions, etc.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Kembara Bayu
Date Deposited:	05 Aug 2024 06:20
Last Modified:	05 Aug 2024 06:20
URI:	http://repository.its.ac.id/id/eprint/113048

Actions (login required)

View Item