Grant: Catalyst Grant: NWHRI: Innovation Fund - Biomedical Discovery Grants

This project will implement artificial intelligence techniques of varying levels of complexity to automate specific portions of the data acquisition and pre-processing pipeline used by a novel, bra-based breast cancer screening technology called the BRAVE system. The BRAVE system generates 3D images of the internal stiffness of a woman's breasts, based on data acquired during short periods (less than 2 minutes) while she wears a specially prepared bra. The BRAVE image reconstruction process itself is a non-linear, physics-based data inversion method that does not rely on artificial intelligence. During an initial pilot study of the system on 22 women, including 7 recently diagnosed with biopsy confirmed cancer, the BRAVE system successfully detected all tumors present within the study population with no false-positives. The automation steps developed during this project will simplify the use of the system so that women can regularly check for the presence of stiff inclusions in their breasts at home. Modern Deep Learning (DL) techniques will be used to streamline the data pipeline for the BRAVE system by performing automated image classification, image segmentation, object detection and feature extraction of the data used to drive the BRAVE image reconstruction algorithm. These DL techniques will convert the raw data captured by the BRAVE data acquisition system into the model-based representation used of by the BRAVE algorithm. These modifications will have the immediate effect of allowing larger populations for ongoing clinical studies of the BRAVE system as well as extension of these studies to multiple centers, rapidly increasing the number of women who will have access to the system. Results from these studies will serve as grounds for approval for the method for at-home breast cancer screening, available even in remote regions, far from existing health services.