Medical image reasoning with the convolutional neural network - based fuzzy logic
[ X ]
Tarih
2022
Yazarlar
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Adana Alparslan Türkeş Bilim ve Teknoloji Üniversitesi
Erişim Hakkı
info:eu-repo/semantics/openAccess
Özet
İlk olarak Aralık 2019'da Çin'in Hubei eyaletinde yerel bir vahşi hayvan pazarında tespit edilen yeni tip koronavirüs (Covid-19), geçen bir yıllık sürede tüm dünyayı etkisi altına alan bir pandemiye dönüşmüştür. Hastalığın kesin bir tedavisi henüz bulunmamaktadır. Aşı geliştirme ve uygulamaları olumlu sonuçlar verse de, salgın henüz dünya genelinde kontrol altına alınamamıştır. Salgının durdurulmasında en etkili yöntem aşılamayla birlikte vakaların hızlı tespiti ve karantina tedbirlerinin uygulanmasıdır. Son yıllarda bilgisayarlı görü(CV) alanında başarılı sonuçlar veren Evrişimli Sinir Ağları(CNN), pandemik vakaların teşhisinde radyologlara yardımcı bir yöntem olarak kullanımı bu tezin konusudur. Çalışma kapsamında, Covid-19 teşhis yöntemleri incelenmiştir. CNN'in yapısı incelenmiş ve görüntü sınıflama alanında kullanımı açıklanmıştır. Farklı CNN mimarileri ve geliştirme ortamları tanıtılmıştır. Pandemi teşhisinde kullanılan biyomedikal görüntüleme teknikleri kısaca tanıtılmıştır. CNN uygulamalarında kullanılacak görüntü setleri için ön işleme ve veri artrma yöntemleri irdelenmiştir. Oluşturulan CNN modeli açık kaynalardan elde edilen göğüs bilgisayarlı tomografi(BT) görüntüleriyle eğitilmiş ve test edilmiştir. Test sonuçları literatürdeki çalışmalarla karşılaştırılmıştır. Eğitilmiş ağ ile belirli bir örneğin sınıflandırma başarısını yorumlamada bulanık mantık uygulanmıştır.
The new type of coronavirus (Covid-19), first detected in a local wild animal market in China's Hubei province in December 2019, has turned into a pandemic that has affected the whole world in a year. A definitive treatment for the disease has not yet been found. Although the vaccine development and applications have yielded positive results, the epidemic has not yet been brought under control worldwide. The most effective method in stopping the epidemic is the rapid detection of cases and the implementation of the quarantine measures together with the vaccination. The use of Convolutional Neural Netwoks(CNN), which has successful applications in the field of Computer Vison(CV) in recent years, as an auxiliary method for radiologists in the diagnosis of pandemic cases is the subject of this thesis. Within the scope of the study, the Covid-19 diagnostic methods are examined. The structure of the CNN is examined and its usage in image classification is explained. Different CNN architectures and development environments are introduced. Biomedical imaging techniques used in the diagnosis of the pandemics are briefly introduced. Preprocessing and data augmentation methods for image sets to be used in the CNN applications are discussed. The generated CNN model was trained and tested with Computed Tomohraphy(CT) chest images obtained from open sources. Test results were compared with studies in the literature. Fuzzy logic was applied to interpret the classification success of a specific sample with the trained network.
The new type of coronavirus (Covid-19), first detected in a local wild animal market in China's Hubei province in December 2019, has turned into a pandemic that has affected the whole world in a year. A definitive treatment for the disease has not yet been found. Although the vaccine development and applications have yielded positive results, the epidemic has not yet been brought under control worldwide. The most effective method in stopping the epidemic is the rapid detection of cases and the implementation of the quarantine measures together with the vaccination. The use of Convolutional Neural Netwoks(CNN), which has successful applications in the field of Computer Vison(CV) in recent years, as an auxiliary method for radiologists in the diagnosis of pandemic cases is the subject of this thesis. Within the scope of the study, the Covid-19 diagnostic methods are examined. The structure of the CNN is examined and its usage in image classification is explained. Different CNN architectures and development environments are introduced. Biomedical imaging techniques used in the diagnosis of the pandemics are briefly introduced. Preprocessing and data augmentation methods for image sets to be used in the CNN applications are discussed. The generated CNN model was trained and tested with Computed Tomohraphy(CT) chest images obtained from open sources. Test results were compared with studies in the literature. Fuzzy logic was applied to interpret the classification success of a specific sample with the trained network.
Açıklama
Fen Bilimleri Enstitüsü, Elektrik-Elektronik Mühendisliği Ana Bilim Dalı
Anahtar Kelimeler
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrol, Computer Engineering and Computer Science and Control
