MACHINE-LEARNING-ENABLED RECOVERY OF PRIOR INFORMATION FROM EXPERIMENTAL BREAST MICROWAVE IMAGING DATA

Abstract

|We demonstrate the recovery of simple geometric and permittivity information of breast models in an experimental microwave breast imaging system using a synthetically trained machine learning work(cid:13)ow. The recovered information consists of simple models of adipose and (cid:12)broglandular regions. The machine learning model is trained on a labelled synthetic dataset constructed over a range of possible adipose and (cid:12)broglandular regions and the trained neural network predicts the geometry and average permittivty of the adipose and (cid:12)broglandular regions from calibrated experimental data. The proposed work(cid:13)ow is tested on two different experimental models of the human breast. The (cid:12)rst model is comprised of two simple, symmetric phantoms representing the adipose and (cid:12)broglandular regions of the breast that match the model used to train the neural network. The second, more realistic model replaces the symmetric (cid:12)broglandular phantom with an irregularly shaped, MRI-derived (cid:12)broglandular phantom. We demonstrate the ability of the machine learning work(cid:13)ow to accurately recover geometry and complex valued average permittivity of the (cid:12)broglandular region for the simple case, and to predict a symmetric convex hull that is a reasonable approximation to the proportions of the MRI-derived (cid:12)broglandular phantom.