AWS and the AI startup Hoppr unveiled a groundbreaking foundational model called Grace at the RSNA 2023 conference in Chicago on Sunday. The primary objective is to broaden the incorporation of generative AI solutions within the medical domain.
Grace, Hoppr’s newest innovation, is a business-to-business (B2B) model crafted to expedite the progress of AI solutions for clinical image analysis among software developers. Alongside the introduction of Grace, Hoppr disclosed a significant investment from Health2047, the venture arm of the American Medical Association.
CEO Khan Siddiqui disclosed that Hoppr, established in Chicago in 2019, has amassed a total funding of $4.1 million to date. Notably, both the company and its latest product pay homage to the pioneering computer scientist Grace Hopper.
According to Hoppr’s official communication, the foundational model aims to enrich “image-to-image and text-to-image learning” across various medical imaging modalities.
Grace’s comprehensive training on an extensive dataset enables it to glean insights from diverse imaging modalities and radiology reports. Through the utilization of embeddings and vectors derived from image data, Grace can effectively establish correlations across different modalities, illustrating how a specific anomaly in an x-ray might appear in a CT scan, Siddiqui elaborated.
The model can extract valuable medical, scientific, and operational insights from health imaging data. For example, Grace can prompt users regarding the necessity for additional imaging at the point of care and prepopulate images with results for clinician verification post-assessment.
Siddiqui further elaborated that the model can assist clinical practitioners in treatment planning by facilitating communication via medical imaging studies. This includes inquiring about results, exploring alternative viewpoints, receiving suggestions for medical interventions, and devising treatment strategies.
Developers now have access to Grace through an API service, empowering them to develop solutions that streamline interactions with health images for radiologists, technicians, and other healthcare professionals.
Siddiqui highlighted that Grace accelerates the development of AI solutions by enabling developers to refine their models within weeks, a stark contrast to the conventional year-long process of supervised learning that entails manual image annotation.
Hoppr harnessed Amazon SageMaker, AWS’s machine learning platform, to construct Grace entirely on AWS infrastructure. Siddiqui commended AWS for its state-of-the-art technology, robust data storage capabilities, and scalability, affirming that the cloud platform seamlessly aligns with their needs.
He underscored Hoppr’s unparalleled proficiency in medical imaging, positioning the company as a frontrunner in leveraging AI to advance healthcare practices.
