Klasifikasi Penyakit Paru Obstruktif Kronik Berbasis Exhaled Breath Menggunakan Model VGG-1D Dan ResNet-1D Pada Sistem Electronic Nose

Aulia, Sheva (2026) Klasifikasi Penyakit Paru Obstruktif Kronik Berbasis Exhaled Breath Menggunakan Model VGG-1D Dan ResNet-1D Pada Sistem Electronic Nose. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026221145-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Penyakit Paru Obstruktif Kronik (PPOK) merupakan penyakit pernapasan progresif yang ditandai oleh keterbatasan aliran udara persisten dan menjadi salah satu penyebab utama kematian di dunia. Metode diagnosis konvensional, seperti spirometri dan Chest X-ray, masih memiliki keterbatasan sehingga diperlukan metode skrining yang lebih cepat, nyaman, dan non-invasif. Salah satu alternatifnya adalah pemanfaatan sistem electronic nose untuk menganalisis Volatile Organic Compounds (VOCs) pada udara napas. Penelitian ini bertujuan mengklasifikasikan subjek PPOK dan sehat menggunakan model satu dimensi Visual Geometry Group (VGG-1D) dan Residual Network (ResNet-1D). Dataset sekunder yang digunakan terdiri atas 70 sampel, yaitu 40 subjek PPOK dan 30 subjek sehat, yang diperoleh menggunakan 20 sensor gas berbasis Metal Oxide Semiconductor (MOS). Tahapan penelitian meliputi baseline removal, rekonstruksi sensor TGS4161, ekstraksi fitur berbasis nilai rata-rata, dan klasifikasi menggunakan Stratified K-Fold Cross Validation. Optimasi model dilakukan melalui variasi metode scaling, jumlah filter, dan ukuran kernel. Hasil penelitian menunjukkan bahwa VGG-1D menghasilkan performa terbaik dengan konfigurasi filter 32, 32, 64, dan 64, dense layer berisi 128 neuron, serta konfigurasi kernel 3, 3, P2, 3, 3, P2. Model tersebut memperoleh nilai accuracy 95,77%, precision 95,95%, recall 95,48%, specificity 95,48%, F1-score 95,65%, dan ROC-AUC 0,97, lebih tinggi dibandingkan ResNet-1D pada seluruh metrik evaluasi. Hasil penelitian menunjukkan bahwa VGG-1D mampu mengenali pola respons sensor gas secara efektif dan berpotensi menjadi alat bantu skrining awal PPOK bagi dokter dan tenaga medis di rumah sakit secara cepat, non-invasif, dan akurat..
==============================================================================================================================
Chronic Obstructive Pulmonary Disease (COPD) is a progressive respiratory disease characterized by persistent airflow limitation and is one of the leading causes of death worldwide. Conventional diagnostic methods, such as spirometry and chest X-ray, still have several limitations, creating the need for a faster, more convenient, and non-invasive screening approach. One promising alternative is the use of an electronic nose system to analyze Volatile Organic Compounds (VOCs) present in exhaled breath. This study aims to classify COPD and healthy subjects using one-dimensional Visual Geometry Group (VGG-1D) and Residual Network (ResNet-1D) models. The secondary dataset consisted of 70 samples, including 40 COPD subjects and 30 healthy subjects, collected using 20 Metal Oxide Semiconductor (MOS) gas sensors. The proposed methodology included baseline removal, TGS4161 sensor reconstruction, mean-based feature extraction, and classification using Stratified K-Fold Cross Validation. Model optimization was performed by varying the scaling method, number of filters, and kernel size. The experimental results showed that the VGG-1D model achieved the best performance using a configuration of 32, 32, 64, and 64 filters, a dense layer with 128 neurons, and a kernel configuration of 3, 3, P2, 3, 3, P2. The model achieved an accuracy of 95.77%, precision of 95.95%, recall of 95.48%, specificity of 95.48%, F1-score of 95.65%, and an ROC-AUC of 0.97, outperforming the ResNet-1D model across all evaluation metrics. These findings demonstrate that VGG-1D can effectively capture gas sensor response patterns and has the potential to serve as a rapid, non-invasive, and accurate COPD screening tool for physicians and healthcare professionals in hospitals.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	PPOK, Sensor Gas, Electronic Nose, VGG-1D, ResNet-1D COPD, Gas Sensors, Electronic Nose, VGG-1D, ResNet-1D
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Sheva Aulia
Date Deposited:	27 Jul 2026 07:58
Last Modified:	27 Jul 2026 07:58
URI:	http://repository.its.ac.id/id/eprint/137856

Actions (login required)

View Item