Industry Day

Jona

Abstract: Artificial intelligence (AI) has immense potential to transform clinical practice by enabling precision medicine, optimizing decision-making and offering large-scale personalization of patient care. However, translating AI from research to real-world clinical applications presents significant challenges, including integration in clinical pathways, regulatory, messaging and appropriate payment structures. In this talk, we discuss key strategies to overcome these barriers to help researchers, clinicians and industry leaders bridge the gap between AI innovation and clinical implementation. As a case study, we will examine how Jona has created a custom LLM and digital twin to analyze complex gut microbiome data and assess interventions. We show how we have been able to apply these strategies to bring powerful new AI technology into clinical practice.

Biography: Leo is internationally recognized for his work to deliver AI in health for over 20 years at pioneering bay area startups (HeartFlow), multinational medical companies (Siemens) and as CEO of Paige.ai. As CEO of Paige, Leo led the company to become an industry leader, internationally launching groundbreaking products and receiving the first-ever FDA approval for an AI product in pathology. Leo’s new company, Jona, has developed a unique AI and digital twin to both decode and shape the gut microbiome to significantly improve health. Leo has authored two books on AI, over 100 peer-reviewed scientific papers and is an inventor on over 300 patents. Winner of the Edison Patent Award, he was inducted as Fellow in the American Institute for Medical and Biological Engineering. Leo earned a Ph.D. in Cognitive and Neural Systems from BU. Leo is CEO in Residence with Breyer Capital and the Founder and CEO of Jona.

Google Health

Abstract: The potential of artificial intelligence (AI) in healthcare has been limited by the breadth of the applications required for transformative impact and high cost of building task-specific models, due to data and compute requirements. Recent work unlocks the potential of new capabilities and applications including zero-shot classification, multimodal classification, semantic search, visual question answering, and automated radiology report generation. To accelerate the adoption of AI in medicine, we created Health AI Developer Foundations (HAI-DEF), a suite of foundational models spanning diverse medical modalities, including computed tomography, pathology, dermatology, chest X-ray, and bioacoustic data. HAI-DEF models, trained on large datasets, can significantly reduce the data and computational resources required for downstream AI development. These models can be used to advance novel research investigations and medical device development by setting a strong baseline in performance that can be further improved upon by researchers and developers. This work provides a critical step towards more accessible, powerful, and versatile AI solutions in healthcare.

Biography: Atilla has over 20 years of experience in healthcare research and innovation. As part of Health AI at Google Research, he developed state-of-the-art lung and breast cancer detection systems, from conception to medical devices and technologies deployed internationally through healthcare partnerships. Currently, his focus is on creating generalizable foundational models to accelerate healthcare discovery and innovation. Prior to Google, Atilla contributed to startups and medical imaging companies, developing novel technologies for radiology and interventional radiology, earning an R&D 100 Award. He has authored over 100 publications and patents and co-authored an ISO standard on AI bias. He holds PhD and MS degrees in Computer Science and Electrical Engineering, specializing in medical imaging, from The Pennsylvania State University. Atilla is passionate about AI’s potential to revolutionize healthcare.

GE HealthCare

Abstract: The rapid evolution of Large Language Models (LLMs) and Foundation Models (FMs) is transforming many industries including the landscape of healthcare AI. Meanwhile, as we move beyond traditional AI paradigms, the emergence of Agentic AI-AI systems that can reason, plan, and autonomously execute complex tasks-is unlocking new frontiers in imaging iunderstanding, workflow automation, and precision diagnostics.

In this talk, we will explore how FMs trained on multimodal medical data are driving breakthroughs in radiology and imaging-based clinical decision support. We will discuss the integration of LLMs with imaging AI to enhance clinical workflows, reduce physician burden, and improve patient outcomes. Additionally, we will examine the role of Agentic AI in orchestrating intelligent automation, enabling more adaptive and personalized healthcare solutions. Through real-world examples and the latest advancements, this talk will also provide insights into the opportunities and challenges of deploying these AI technologies at scale while ensuring safety, transparency, and clinical efficacy.

Biography: Cao (Danica) Xiao is VP of AI at GE Healthcare, leading a global AI science and engineering team focused on generative AI and large language models to enhance clinical and operational efficiencies. Previously, she held leadership roles at Relativity, Amplitude, IQVIA, and IBM Research, driving AI innovations in healthcare. A recognized thought leader, she has published over 160 highly cited papers in top AI/ML venues and co-authored a deep learning for healthcare textbook used in leading CS graduate programs like UIUC, GaTech, and PSU. Danica was named a “Top Chinese Young Scholar in AI” (2022) and “Top Chinese Female Scholar in AI” (2023) by Baidu. She earned her Ph.D. in machine learning from the University of Washington, Seattle, in 2016.

Subtle Medical

Abstract: The rapid evolution of artificial intelligence and generative AI is transforming clinical imaging, offering unprecedented opportunities to enhance diagnostic workflows. In this talk, we present our latest advancements in AI-powered imaging applications, focusing on augmenting radiology data generation: acceleration, enhancement, synthesis, and low-dose imaging. We will discuss how deep learning and generative models are integrated into our imaging platforms to significantly reduce scan times while improving image quality and diagnostic accuracy. Our approach leverages AI to standardize imaging protocols and synthesize missing contrast sequences, enabling superior visualization of anatomical structures and pathology. Furthermore, by optimizing imaging techniques, we have achieved ultra-low-dose imaging solutions that maintain diagnostic integrity while minimizing risks. Clinical validation studies across multiple institutions and clinical adoption in hundreds of sites globally have demonstrated the clinical and operation value and potential for widespread adoption of these technologies, marking a pivotal step towards next-generation medical imaging and improved patient care.

Biography: Dr. Enhao Gong is a researcher and entrepreneur specializing in the application of AI/GenAI in medical imaging. Dr. Gong holds MS and PhD in Electrical Engineering from Stanford University, focusing on deep learning applications for medical image acquisition, reconstruction, enhancement, and quantification.

As the founder and CEO of Subtle Medical, he leads the development of cutting-edge AI technologies that enhance medical imaging acquisition, reconstruction, and quantification. Under his leadership, Subtle Medical has achieved significant milestones, including securing 9 FDA clearances for its AI-powered products and servicing over 700 hospitals and imaging centers worldwide. The company has raised over $50 million from top venture capitalists and has been recognized with prestigious awards such as the NVIDIA Inception AI Award, CB Insights AI-100, Top 50 GenAI, etc. Dr. Gong’s groundbreaking work has been awarded by RSNA, Forbes 30-under-30, Radiology Business 40-under-40, Fortune 40-under-40, and Top 50 Healthcare AI Entrepreneur Award.

United Imaging Intelligence

Abstract: Artificial intelligence (AI) is transforming medical imaging by addressing critical clinical challenges across various applications. Recently, foundation models have emerged as a powerful solution, offering a scalable learning framework that enhances model generalization, enables few-shot learning, and improves real-time imaging applications. In this talk, we will also explore the role of diffusion models as prior distribution foundation models, particularly in accelerating Cardiac MRI (CMR) acquisitions to enhance image quality and ensure more reliable visualization. In interventional cardiology, low-dose imaging often compromises device and vessel visibility, impacting procedural accuracy. To address this, we introduce a contrastive learning-based approach that improves vessel visualization in real-time imaging, aiding precise device placement. Additionally, we present our work on label-efficient data augmentation using video diffusion models for guidewire segmentation, demonstrating how diffusion-based foundation models optimize downstream tasks.

By integrating diffusion models, contrastive learning, and few-shot adaptation, foundation models are driving advancements in AI-driven cardiac imaging. This talk will showcase their impact on CMR, interventional imaging, and real-time decision support.

Biography: Dr. Shanhui Sun is the Senior Director of AI and Medical Imaging at UII America Inc., based in Boston, MA, a subsidiary of Shanghai United Imaging Intelligence Healthcare Co. Ltd. (UII). His research focuses on medical image and video analytics, machine learning, cardiac MRI, breast imaging analysis, and AI-driven imaging, with expertise in MRI reconstruction and image enhancement. His recent work explores generative AI and diffusion models for image restoration, denoising, and data augmentation. He also specializes in AI-driven interventional imaging, including real-time imaging enhancements, device tracking, and surgical navigation optimization. Dr. Sun is actively involved in deploying AI models into clinical workflows, ensuring seamless integration and scalability. He has co-authored over 50 publications and co-invented more than 80 granted patents. Previously, he was a Principal Scientist at CuraCloud and a Staff Scientist at Siemens. He earned his Ph.D. from the University of Iowa in 2012.