Performance Based Analysis of Classification Techniques in Breast Cancer Screening – A Review

Abstract

According to the survey of WHO, in 2020 there are 2.3 million women found with breast cancer and 685,000 deaths in world wide. 81% women get affected with cancer over the age of 50 at the time of detection. Breast cancer is the world’s number 2 cancer and number 1 cancer in India and 66% survival rate in India is very low if compare to 90% in U.S and 90.2% in Australia. However, treatment for this cancer has possibility of 90% or more. So that, it needs to detect the cancer at very early stage to overcome the death rate. In healthcare sector, there are many ways for screening breast cancer like: mammography, sonography, ultrasound and MRI for detection of benign and malignant tumors before symptoms appear. There are some other ongoing experiments exist i.e., PET (positron emission tomography) scans, thermography, ductogram (ducto lavage, ductoscopy) etc. CAD system which are used for classification breast cancer abnormalities, assisting doctor as a second opinion. Now a days DL-CADs (Deep learning CAD) in use, which are better than traditional CADs for complex data analysis. This paper discussed the complete survey of deep learning techniques and data sets which are in use for breast cancer classification. And resulting with challenges/limitation or future work in this area of study.