Abstract:
This project is on “Convolutional Neural Network Based Image Classifier for
Breast Cancer Histopathology Images”. Thousands of people around the globe die
every year as result of Breast Cancer (BC). Breast Cancer stages are ranging from
early curable stages to late metastatic stages. An early detection of breast cancer can
save millions of lives each year. Cancer cells can be detected through several ways
like breast MRI scan, Mammogram, breast Ultrasound and Histopathology images.
In our research we have used publicly available breast cancer histopathology image
dataset hosted at http://web.inf.ufpr.br/vri/breast-cancer-database. We have tried to
develop automated malignant and benign breast cancer detection system which can
detect cancer from the histopathology images and thus making it more efficient and
diagnosis more scalable and less prone to error. This type of research can be extended
further to apply on other type of cancer detection.
Description:
Cancer is a significant public health problem in the world today. According to the
IARC (International Agency for Research on Cancer) of the WHO (World Health
Organization), 8.2 million deaths were caused by cancer in 2012 and 27 million of
new cases of this disease are expected before 2030 [1]. In particular, breast cancer
(BC) is one of most common type of cancer among women.
Mortality of BC is very high when compared to other types of cancer. Detection and
diagnosis of BC can be achieved by imaging procedures such as diagnostic
mammograms (x-rays), magnetic resonance imaging, ultrasound (sonography), and
thermography [2]. Imaging for cancer screening has been investigated for more than
four decades [3]. However, biopsy is the only way to diagnose with confidence if
cancer is really present. Among biopsy techniques, the most common are fine needle
aspiration, core needle biopsy, vacuum-assisted and surgical (open) biopsy (SOB) [4].