Akademik Veri Yönetim Sistemi
Tezin Türü: Yüksek Lisans
Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye
Tezin Onay Tarihi: 2016
Tezin Dili: İngilizce
Öğrenci: Mustafa Ümit Öner
Danışman: UĞUR HALICI
Özet:Breast cancer digital histopathology is a new application area of deep learning. Breast cancer was the leading cause of cancer death among women with 15.1% death rate among all cancer deaths in the world in 2012. Insufficient number of pathologists is one of the key factors in that situation. There were 5.7 pathologists per 100.000 people in USA in 2013 and this value was 1.56 in Turkey in 2011. It is possible to increase the number of slide analysis made by the pathologists within the same period by developing deep learning based systems to assist them. In this thesis, a convolutional neural networks based system is introduced. This system accepts the whole slide images of lymph node excisions from breast cancer patients as input and detects and localizes metastasis regions on these images automatically. In this system, performance values of 0.9259 and 0.8669 for slide-based evaluation and 0.5349 and 0.4060 values for the lesion based evaluation are achieved on CAMELYON16 training and test sets, respectively.