BKG

Mahasin, Muhammad Masdar (2020) Rancang Bangun Arsitektur Convolutional Neural Network Untuk Diagnosis Pneumonia Berdasar Citra X-Ray Dada. Sarjana thesis, Universitas Brawijaya.

Indonesian Abstract

Radiografi memegang peranan penting dalam bidang medis. Ada banyak penyakit dalam tubuh manusia yang dapat didiagnosis dengan tepat menggunakan Radiodiagnostik, seperti pneumonia hingga isu yang sedang hangat yaitu Covid-19. Diagnosis pneumonia pada umumnya memanfaatkan analisis ahli radiologi terhadap citra sinar-X dada pasien secara manual berdasarkan tampilan visual. Berbagai metode komputasi telah dikembangkan dalam proses diagnosis pneumonia, termasuk penggunaan deep learning dengan arsitektur VGGNet dan XCeption. Seiring berkembangnya teknologi, berbagai model komputasi berkembang pesat untuk mendapatkan hasil diagnosis yang lebih cepat dan akurat. Pada penelitian ini, dikembangkan model Deep learning CNN dengan arsitektur UBNet. UBNet adalah arsitektur CNN yang dirancang khusus untuk mengolah citra sinar-X dada dalam bentuk 2 dimensi dengan mempertimbangkan beban komputasi dan juga performa model. Kemudian, arsitektur UBNet disusun secara bertingkat sebagai sebuah sistem diagnosis pneumonia. Sistem diagnosis pneumonia ditampilkan dalam sebuah GUI (Graphic User Interface). Dalam pengujian, didapatkan akurasi UBNet mencapai +90% dengan waktu diagnosis kurang dari 10 detik ketika dijalankan dalam hardware tanpa GPU. Pada analisis error pada sistem diagnosis, diketahui bahwa sistem menunjukkan pola-pola unik ketika melakukan suatu kesalahan dalam proses diagnosis.

English Abstract

Radiography has an important role in the medical field. Many diseases in the human body can be diagnosed correctly using Radiodiagnostics, such as pneumonia to the current issue of Covid-19. The diagnosis of pneumonia in general utilizes radiologist's analysis of the patient's chest X-ray images manually based on visual appearance. Various computational methods have been developed in the process of diagnosing pneumonia, including the use of deep learning with the VGGNet and XCeption architectures. As technology develops, various computing models develop rapidly to get faster and more accurate diagnosis. In this study, a CNN Deep learning model was developed with UBNet architecture. UBNet is a CNN architecture specifically designed to process chest X-ray images in 2-dimensional form by considering the computational load and also the performance of the model. Then, UBNet's architecture was arranged in stages as a pneumonia diagnosis system. The pneumonia diagnosis system is displayed in a GUI (Graphic User Interface). In testing, obtained UBNet accuracy reached + 90% with a diagnosis time of less than 10 seconds when run on hardware without GPU. In the analysis of errors in the diagnosis system, it is known that the system shows unique patterns when making a mistake in the diagnosis process.

Other Language Abstract

-

Item Type: Thesis (Sarjana)
Identification Number: 0520090243
Uncontrolled Keywords: citra sinar-X, Deep learning, UBNet.
Subjects: 500 Natural sciences and mathematics > 537 Electricity and electronics > 537.5 Electronics
Divisions: Fakultas Matematika dan Ilmu Pengetahuan Alam > Fisika
Depositing User: ismanto
URI: http://repository.ub.ac.id/id/eprint/183327
Text
0520090243 - Muhammad Masdar Mahasin.pdf

Download | Preview

Actions (login required)

View Item View Item