Differential Breast Microwave Radar Imaging: Diagnostic Performance Evaluation and Comparison to Single-Breast Diagnosis

Microwave breast imaging (MBI) has been explored as a breast cancer diagnostic tool, but despite the proliferation of clinical evaluation of MBI systems, challenges remain before the modality is ready for clinical use. Suppression of the dominant skin responses, which obfuscate the interior tissues, is a major challenge. A possible solution to this challenge is differential imaging, rather than single-breast images, where only differences between the left/right breasts are displayed. This work presents the first estimates of diagnostic sensitivity and specificity obtained in MBI using a differential imaging approach. The delay-and-sum (DAS), delay-multiply-and-sum (DMAS), and optimization-based radar reconstruction (ORR) algorithms were used. Only the ORR method achieved better-than-random diagnostic performance when the tumour-detection criteria were defined using the signal-to-clutter ratio and localization error. This work also identified the potential of using the maximum image intensity as a diagnostic criterion. When this maximum-response-based metric was used, the diagnostic performance of all reconstruction methods improved. The ORR algorithm was the best performing method and achieved an area under the curve of the receiver operating characteristic curve of 84.6%.