Automated Breast Cancer Diagnosis Based on Neural Network Algorithms

Breast cancer, the most common among the women, is primarily examined visually by starting with initial screening of breast and lymph nodes in armpit, dermoscopic analysis, histopathological tests and a biopsy. In India as per the ratings, 25–35% women are affected with breast cancer in every city in India. Being in a developing country like India, majority of women are dying due to breast cancer and for every 4 min women is affected with the same. Women are suffering a lot due to this curse, masses or micro calcification are the symptoms a woman can have at a previous stage, where microcalcification is mineral which gets into the breast tissue. It’s quite hard to examine due to ultra-fine shapes and uncertain edges. It is being categorized into three main categories normal, malignant and Benign, which can be detected using proposed system. In proposed scheme, input images are being compared with large collection of databases with the replacement of sigmoid activation function. Probabilistic neural network is used to describe nonlinear statement limits which further leads to Bayes optimal and also all the function which bear same properties as well. Any input data or algorithm can be pointed with the four layer neural network. Standard PNN consists of input layer of N nodes, a pattern layer of m nodes, a summation layer of k node and it consists of pseudo layer L nodes, which is used for decision making which is also known as decision layer for producing output.