SBIR Phase I: HistoMapr-Breast: Computational diagnostic guides for breast pathologies

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to accelerate the implementation of digital pathology by giving pathologists computational guides to help them achieve higher accuracy and efficiency in making their "calls". Currently biopsy diagnosis is a labor-intensive process that relies upon 19th century microscope technology; pathologists are doctors who use microscopes to make diagnoses. Pathologists examine slides from breast biopsies as part of breast cancer screening programs for more than a million women each year in the US. This is a highly subjective process and there is evidence that even pathologists may disagree with one another, especially with difficult cases. This is potentially large patient safety issue, as some patients may not be optimally triaged. It is currently possible to digitize the slides as whole slide images (WSIs), but there are very limited options for computer assisted pathologist review and this is a major reason that the digital pathology market has not developed as rapidly as most experts predicted. This project is a major advance as it addresses a large clinical need ? augmenting pathologists in their ability to make efficient and accurate diagnoses; and it also has great commercial potential as a disruptive, must-have application in the early digital pathology marketplace. This Small Business Innovation Research (SBIR) Phase I project will address the issue of accuracy and efficiency of pathologist diagnosis in breast core biopsies, and this should validate our approach as an unmet need in the current digital pathology marketplace. There are other groups applying machine learning to pathology, but this project is a unique approach that addresses practical issues related to expert training of the machine, rapidly displaying targeted images to pathologists, and providing a guiding support while keeping the doctors in control. The project will ask pathologists to label WSIs with their diagnoses; this data will be used to train an AI system; and the AI system performance will be studied. Preliminary studies seem to support that a human and AI system working together will be both more efficient and more accurate than a human working alone. If this expanded project validates this, then there is a tremendous commercial opportunity that can also have a positive impact on patients. Not only access to better diagnoses for patients, but also the possibility for any pathologist to perform at a higher level. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.