Early Breast Health Screening Performance Verification based on UWB Imaging and Neural Network

Breast cancer is an abnormal cell locates in the breast tissue. Early detection plays an important role as it helps for long-term survival. The available systems are expensive and need expert to operate it. This paper presents a low-cost and user-friendly breast cancer detection system for end user at home. This system consisted of a pair of home-made antenna, Ultra wide-band (UWB) transceiver to receive signal and a Neural Network (NN) module for signal analysis. Transmitted UWB signals from one antenna were received by the other and saved in PC. Discrete Cosine Transform (DCT) was to use to convert the received analogue signal to discrete values integers 1632 points. Then through feature reduction only four points were achieved from 1632 point values, which were fed into NN module. The system's efficiency is tested on developed breast phantom before real clinical test. Breast phantom is placed in the center while two UWB antennas were placed diagonally opposite side of the breast phantom. K-fold cross validation based feed forward NN is used to train, validate and test the features. The system can screen the breast cancer with 100%, 82.62% and 90.69% using heterogeneous breast phantom and 100%, 89.06% and 83.96% using homogeneous breast phantom for existence, location and size respectively. The proposed breast cancer detection system will be very useful for home user to check breast health regularly.