We compare the performance accuracy of a slot-based antenna and a Vivaldi antenna for breast tumor detection using machine learning (ML) algorithms jointly with microwave imaging (MWI) processing. MWI is known for having low resolution. Therefore, we here study the conjoint use of ML and MWI, in order to enable accurate detection of breast tumors and evaluate how the probing antenna affects the overall system performance. To this end, we perform measurements in the frequency range of 2-6 GHz on anthropomorphic breasts of different volumes and shapes, where we placed two types of tumors. The slot-based antenna provides better imaging results (i.e. good detection of the tumor), but the accuracy of ML techniques is only 60%. Concerning the Vivaldi antenna, the images present clutter, but the accuracy of ML techniques is as high as 85%. These results show that ML and MWI can be complementary to each other.

--

Artificial inteligence,Breast cancer diagnosis,Broadband antennas,Data fusion,K-nearest neighbour,Linear discriminant analysis,Machine learning,Principal component analysis,Support vector machines