The broader impact of this Small Business Innovation Research (SBIR) Phase I project is to extract new valuable information from medical images, accelerate image interpretation by radiologists, and improve patient outcomes. The process of identifying key features in images, known as ``labeling'', is the key to improved diagnosis and management of certain conditions. The innovations proposed here will substantially lower the cost of labeled datasets, enabling access for developers of artificial intelligence (AI) algorithms and improving the use of AI in health care. This Small Business Innovation Research (SBIR) Phase I project will apply machine learning algorithms to develop a system for assisting in manual labeling of medical tomographic images. The proposed research will result in an adaptive system architecture that evolves to accelerate labeling and increase the volume of labeled data. Moreover, the research will increase labeling accuracy at the edges of anatomical structures. For instance, surgical resections for cancer treatment requires accurate labeling of the edges of abnormal tissue to ensure clean margins and minimal recurrence. Similarly, radiation therapy planning requires accurate labeling of the edges of organs at risk for safety and favorable outcomes. Due to its clinical importance, accurate manual labeling of ambiguities and sophisticated shapes is highly time-consuming. The proposed approach is differentiated from current methods by the inclusion of an additional subsystem for increasing the accuracy of edge labeling.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.
