Insider Brief
- HOPPR announces the launch of Grace, a groundbreaking multi-modal foundation model for medical imaging, in collaboration with AWS and with investment from Health2047.
- Grace enables image-to-image and text-to-image learning across various medical imaging modalities and is available via private beta for development and fine-tuning by developers, radiology PACS, and AI companies.
- HOPPR’s development of Grace on AWS, including plans to use AWS HealthImaging and Amazon Bedrock for data storage and model development, aims to address challenges in medical imaging AI, such as static solutions and lack of integration, and this collaboration is poised to accelerate clinical application development, increase image depth, and potentially improve patient outcomes.
PRESS RELEASE — CHICAGO/November 26, 2023 /PRNewswire/ — Today, HOPPR announces the launch of Grace, a multi-modal foundation model for medical imaging, powered by Amazon Web Services, Inc. (AWS) and available via private beta to developers, radiology PACS, and AI companies for fine tuning and application development. Together with a milestone investment from Health2047, a venture studio founded by the American Medical Association (AMA), this launch marks a significant step forward in HOPPR’s quest to unlock the potential of generative AI in medical imaging.
Grace is a first-of-its-kind B2B foundation model that enables image-to-image and text-to-image learning across all medical imaging modalities, including X-rays, CTs, MRIs, and echocardiograms. Available via an API service, Grace enables application developers to more quickly build meaningful AI solutions that physicians, technicians, and support staff can use to engage interactively with medical images.
With Grace, users can unlock diagnostic, clinical, and operational value from medical imaging data. An organization’s own data can be used to securely fine tune the model for use in applications that allow users to then converse with medical imaging studies about findings, alternative imaging views, suggested surgical interventions, and treatment protocols. The model also supports non-clinical use cases including workflow, billing and coding review, and QA, providing a one-API shop for all the data needed to support the imaging sector.
Grace has been meticulously developed using over a petabyte of permission-based, anonymized medical imaging study data. These studies have been enriched with corresponding reports to ensure robust training for commercial deployment across extensive datasets, spanning both 2D and 3D modalities and inclusive of longitudinal imaging studies through strategic collaborations with key partners like Gradient Health. At scale, Grace will contain approximately five trillion parameters — five times more than current commercial generative models trained on one trillion parameters. Committed to responsible AI practices, Grace has been developed with a privacy-centric approach using healthcare industry-standard quality management systems based on the ISO 13485. In preparation for widespread release, HOPPR is actively engaging with partners such as RadNet and Rad AI to refine its offerings to meet the precise needs of the healthcare sector.
“We are thrilled to launch the beta HOPPR foundation model to trusted PACS vendors and developers to fine tune models and provide feedback to prepare us for commercial expansion in Q1 of 2024. Grace represents a game-changing advance for HOPPR and the broader medical imaging space, which stands to benefit enormously from the transformative potential of AI to improve the efficiency and quality of clinical care,” said Dr. Khan M. Siddiqui, CEO of HOPPR.
HOPPR developed its foundation model exclusively on AWS using Amazon SageMaker, with plans to utilize AWS HealthImaging, Amazon Bedrock, and other services for data storage, inferencing, and model development in the future as it’s scaled. Working together, the companies aim to address key obstacles to optimal AI use in medical imaging:
- Dynamic Integration: Many current AI solutions for medical imaging do not fully meet the needs of medical imaging professionals. They are static and lack integration with broader patient context. HOPPR enables cross-modality comparison, historical and contextual perspective, real-time prompt and recall, and system-wide treatment planning.
- Faster and More Cost-Effective Application Development: Clinical app developers spend 12 to 18 cost-intensive months training and developing models and equally lengthy periods of integration and deployment. By exposing the Foundation Model for fine tuning by clients, the development process can be compressed to about a month.
- Increased Image Depth: Most available AI tools were developed by downsampling images, meaning 99% of the data contained in the medical imaging study is not available in traditional training models. Whereas many current AI solutions require downsampling grey scale to 256 shades, HOPPR sees 65,000 shades of grey. HOPPR has developed proprietary vision transformers for its development of the model.
“Accelerating AI’s clinical and operational value in medical imaging eases burdens for radiologists, providers, and support staff, which could ultimately result in better patient outcomes,” said Dan Sheeran, General Manager of Healthcare and Life Sciences at AWS. “We are excited to work with HOPPR to make fine tuning and deploying foundation models for medical imaging easier and faster — decreasing the time to value from years to months.”
In parallel with this milestone, HOPPR has received a $3 million funding round led by Health2047, a Silicon Valley-based venture studio founded by the American Medical Association. With this investment, HOPPR joins Health2047’s portfolio of startups reshaping healthcare for the future. Other companies in the portfolio include Evidium, Medcurio, Phenomix Sciences, ScholarRx SiteBridge Research and Zing Health.
“Health2047 is proud to support HOPPR’s work to build a powerful data repository for researchers and clinicians,” said Lawrence K. Cohen, CEO of Health2047. “As a physician, imagine if you could chat with imaging studies, asking them for treatment protocols, alternate views, and more. HOPPR’s work to unleash the full potential of AI for medical imaging promises to spur innovation and improve efficiency and outcomes.”
Alongside AWS and early partners RadNet and Rad AI, HOPPR will conduct live demonstrations at the Radiological Society of North America (RSNA) annual conference November 26–29. The demonstrations will highlight the potential for solutions to leverage HOPPR’s Foundation Model across radiology and healthcare. Visit HOPPR at Booth #4059, AWS at Booth #4724, RadNet at Booth #4547, and Rad AI at Booth #4733.
To join the HOPPR AI Beta program, please visit our website and sign up for early access. HOPPR — Contact Us Page — HOPPR
About HOPPR AI: HOPPR is changing medical imaging forever by providing data back to clinical systems that will enable physicians, technicians, and clinical support staff to “converse” with medical imaging studies, transforming medical imaging interactions from static to dynamic. HOPPR has created both medical and administrative use cases that it will unveil with commercial partners at RSNA in November 2023.
About Health2047: Health2047 is a Silicon Valley venture studio powered by the American Medical Association. Health2047 is transforming healthcare at the system level, seeking powerful ideas, industry partners, and entrepreneurs to address systemic transformation in the areas of data, chronic disease, and productivity. Health2047’s deep relationships with both the AMA and its network of strategic partners create a unique force multiplier that helps drive informed, large-scale change in healthcare. For more information, please visit www.health2047.com.
