Breast microwave sensing (BMS) is an emerging modality that has the potential to be used as a breast cancer screening technique but challenges remain before the modality is suitable for clinical use. Improvements to image-based and machine learning tumor-detection methods are required. This thesis presents novel improvements in image reconstruction and machine learning methods. This work presents the development of the largest open-access experimental dataset in the published BMS literature to-date, the University of Manitoba Breast Microwave Imaging Dataset (UM-BMID). The impact of the inverse chirp z-transform (ICZT) on radar-based image reconstruction was compared to that of the standard inverse discrete Fourier transform using a subset of this dataset. The ICZT was found to reduce image artifacts, improve image contrast, and increase tumor-detection in reconstructions. A novel reconstruction method, the iterative delay-and-sum (itDAS) beamformer, was compared to two literature standard approaches. The novel method improved image contrast by as much as 249% on average and allows for the implementation of correction factors to improve the radar signal model used in the literature standard algorithms. Three correction factors were examined and modeling the output pulse of the BMS system significantly increased the contrast of itDAS reconstructions. The diagnostic capability of machine learning methods in BMS was investigated using UM-BMID. The area under the curve of the receiver operating characteristic curve of a convolutional neural network was estimated to be between (76 ± 3)% and (91 ± 3)%, where the upper estimate is obtained when the testing set is constrained to consist of phantoms with breast volumes that are within the breast volume bounds of the training set and when the tumor is located at the same vertical position as the antennas. This thesis has set the stage for future large-scale analyses in BMS through the development of the first and largest open-access dataset in the published literature and through the promising results obtained with the application of machine learning methods and the novel itDAS beamformer.