Jaringan Syaraf Tiruan Perambatan Balik untuk Klasifikasi Covid-19 Berbasis Tekstur Menggunakan Orde Pertama Berdasarkan Citra Chest X-Ray

Abstract

<p class="Abstrak"><span lang="IN">COVID-2019 pertama kali muncul di kota Wuhan, Cina pada Desember 2019, kemudian menyebar dengan cepat ke seluruh dunia dan menjadi pandemi. Pandemi COVID-19 telah menyebabkan dampak yang cukup </span><span lang="EN-US">fatal</span><span lang="EN-US">untuk</span><span lang="IN">kesehatan masyaraka. Meru</span><span lang="EN-US">pakan hal yang s</span><span lang="IN">angat penting untuk mendeteksi kasus positif sedini mungkin untuk</span><span lang="EN-US"> pencegahan </span><span lang="IN">penyebaran lebih lanjut dari </span><span lang="EN-US">virus</span><span lang="IN"> ini. Teknik tes paling umum yang saat ini digunakan untuk</span><span lang="EN-US"> mendiagnosa</span><span lang="IN"> COVID-19</span><span lang="IN">adalah <em>reverse-transcriptase polymerase chain reaction</em> (RT-PCR). Pencitraan radiologis dada seperti </span><em><span lang="EN-US">chest </span><span lang="IN">X-ray</span></em><span lang="IN"> memiliki peran penting dalam diagnosis dini</span><span lang="IN">penyakit ini. Karena sensitivitas RT-PCR rendah 60% -70%, bahkan jika </span><span lang="EN-US">h</span><span lang="IN">asil negatif diperoleh, gejala dapat dideteksi dengan pemeriksaan gambar radiologi pasien</span><span lang="EN-US">. T</span><span lang="IN">eknik kecerdasan buatan</span><span lang="IN">yang digabungkan dengan pencitraan radiologis dapat membantu untuk mendiagnosis C</span><span lang="EN-US">O</span><span lang="IN">V</span><span lang="EN-US">I</span><span lang="IN">D-19 dengan lebih cepat dan akurat</span><span lang="EN-US">.</span><span lang="IN">Proses klasifikasi pada penelitian ini terdapat beberapa tahapan yaitu pra-pengolahan, segmentasi, ekstraksi ciri, dan klasifikasi. Ekstraksi ciri yang digunakan adalah berdasarkan tekstur orde pertama dan klasifikasi yang digunakan adalah jaringan syaraf tiruan perambatan balik. Sistem</span><span lang="EN-US"> klasifikasi pada penelitian ini</span><span lang="IN"> menghasilkan</span><span lang="EN-US"> rata-rata</span><span lang="IN"> akurasi klasifikasi sebesar </span><span lang="EN-US">94,17</span><span lang="IN">% untuk kelas</span><span lang="EN-US"> normal</span><span lang="IN"> dan </span><span lang="EN-US">77,5</span><span lang="IN">% untuk </span><span lang="EN-US">COVID-19</span><span lang="IN">. </span><span lang="EN-US">Hasil akurasi tertinggi didapat pada skenario pertama dengan hasil akurasi sebesar 88,8%. Nilai rata-rata sensitivitas yang didapat pada penelitian ini sebesar 94,17% untuk kelas normal dan 76,67% untuk kelas COVID-19. Nilai rata-rata spesifisitas yang didapat pada penelitian ini sebesar 76,67% untuk kelas normal dan 94,17% untuk kelas COVID-19.</span></p><p class="Abstrak"><span lang="EN-US"><br /></span></p><p class="Abstrak"><em><strong><span lang="EN-US">Abstract</span></strong></em></p><p class="Abstract"><em>Covid-2019 first appeared in Wuhan, China, in December 2019, then quickly spread throughout the world and became a pandemic. The Covid-19 pandemic has had a fatal impact on public health. It is crucial to detect positive cases as early as possible to prevent the further spread of this virus. The most common test technique currently used to diagnose Covid -19 is the reverse-transcriptase polymerase chain reaction (RT-PCR). Chest radiological imaging such as chest X-ray has a vital role in the early diagnosis of this disease. Due to the low RT-PCR sensitivity of 60%-70%, symptoms can be detected by examining the patient's radiological images even if a negative result is obtained. Artificial intelligence techniques combined with radiological imaging can help diagnose Covid -19 more quickly and accurately. The classification process in this study consists of several stages, namely pre-processing, segmentation, feature extraction, and classification. The feature extraction used is based on the first-order texture, and the classification used is a backpropagation neural network. The classification system in this study resulted in an average classification accuracy of 94.17% for the normal class and 77.5% for Covid -19. The highest accuracy results were obtained in the first scenario, with an accuracy of 88.8%. The average sensitivity value obtained in this study was 94.17% for the normal class and 76.67% for the Covid -19 class. The average specificity value obtained in this study was 76.67% for the normal class and 94.17% for the Covid -19 class.</em></p><p class="Abstrak"><em><strong><span lang="EN-US"><br /></span></strong></em></p>

Optimization Detection of Skin Cancer using Deep Learning

“…In this study, the normal class' average sensitivity value was 94.17%, and the Covid-19 class' average sensitivity value was 76.67%. In this study, the normal class' average specificity value was 76.67%, and the Covid -19 class' average specificity value was 94.17% [9].…”

Backpropagation Design for Authenticating Blood Vessel Patterns of the Back of the Hand Using GLRLM

“…JST memiliki kemampuan untuk belajar dari data latih dan dapat menyesuaikan diri dengan variasi pola yang kompleks [17]. Dalam konteks pengenalan biometrik, penggunaan JST perambatan balik dapat meningkatkan kemampuan sistem untuk mengenali pola unik pada citra pembuluh darah punggung tangan [18].…”

Optimization Heart Disease Prediction using Machine Learning Models

“…Chest radiological imaging such as chest X-ray has an important role in the early diagnosis of this disease. Due to the low sensitivity of RT-PCR of 60%-70%, even if a negative result is obtained, symptoms can be detected by examining the patient's radiological images [8]. An important phase in facial motion recognition is feature extraction, which determines the level of recognition.…”