A mass detection algorithm for mammogram images

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2014

Öğrenci: MUHAMMED YEŞİLKAYA

Eş Danışman: NEVZAT GÜNERİ GENÇER, GÖZDE AKAR

Özet:

Breast cancer is the most common cancer type encountered among woman in the world and causes many deaths. In order to prevent mastectomies, decrease the probability of return and reduce mortality, early detection of cancer lesion is crucial. Mammography is a frequently used screening technique to detect and diagnose lesions. However, sometimes it is difficult for radiologists to see and diagnose lesions due to low contrast of mammograms. Computer Aided Detection / Diagnosis (CAD / CADx) systems have been developed to help radiologists. In this thesis, we propose a method for classification of mass regions in MLO (Mediolateral oblique) view mammograms. The suspicious regions are first determined by Iris filtering with variable window sizes applied on the breast region without pectoral muscle. Then classification is applied to textural features obtained using Gabor filter applied on these suspicious regions. We reduced false detection ratio nearly 50 percent with a cost of missing 9 percent of true mass regions with classification. For pectoral muscle region determination a novel algorithm is also proposed. This algorithm is based on average derivative calculation and line fitting with least square solution. Our algorithm outperforms other algorithms given in the literature in terms of FP (False positive) pixel percentage and FN (False negative) pixel percentage metrics.