－ AI in Medicine: Bridging Innovation & Practice －
CPHAI Symposium
HANOVER INN, GRAND BALLROOM
WEDNESDAY, APRIL 3RD
Overview
Thank you for joining us at the Inaugural Dartmouth Symposium on Precision Health & AI, on April 3rd at the Hanover Inn Grand Ballroom. The event showcased Dartmouth's leadership in AI and healthcare, featuring keynotes by Dr. Curtis Langlotz and Dr. Faisal Mahmood, with opening remarks by President Sian Leah Beilock. Dartmouth's AI experts and clinicians shared their insights into transforming healthcare through AI, emphasizing collaborative research and innovation. We appreciate your participation and engagement with pioneers in AI-driven healthcare solutions.
Keynote Presentations
Curtis P. Langlotz, MD, PhD
Stanford University
Dr. Langlotz is Professor of Radiology, Medicine, and Biomedical Data Science, Director of the Center for Artificial Intelligence in Medicine and Imaging (AIMI Center), and Associate Director of the Institute for Human-Centered Artificial Intelligence (HAI) at Stanford University. The AIMI Center comprises over 150 faculty from across Stanford who conduct interdisciplinary machine learning research that optimizes how clinical data are used to promote health. Dr. Langlotz’s laboratory investigates the use of machine learning technologies to help physicians detect disease and eliminate diagnostic errors. He has led many national and international efforts to improve medical imaging, including the RadLex terminology standard and the Medical Imaging and Data Resource Center (MIDRC), an NIH-funded U.S. national imaging research repository.
As Associate Chair for Information Systems and a Medical Informatics Director for Stanford Health Care, he is responsible for the computer technology that supports the Stanford Radiology practice, including 20 million imaging studies that occupy 1.6 petabytes of storage. He currently serves as President of the Radiological Society of North America (RSNA). He has founded three healthcare IT companies, including Montage Healthcare Solutions, which was acquired by Nuance Communications in 2016.
The Future of AI in Medical Imaging
Dr. Curtis P. Langlotz showcased the powerful role of artificial intelligence in transforming medical imaging. He highlighted how AI significantly boosts the accuracy of diagnosing diseases, reducing the typical error rates seen in human interpretations, which range from 3-6%. Dr. Langlotz pointed out that AI is particularly useful in radiology, where it automates routine tasks and improves the detection of subtle medical conditions. He also highlighted the breadth of AI applications, from computer-aided detection to offering new insights into imaging, showcasing examples of AI systems transforming medical practices in real time. Furthermore, he also addressed some critical challenges and limitations of AI in this domain, such as data biases and the need for continuous oversight to ensure AI systems remain effective and fair. The presentation highlighted the current benefits of AI in medical practice and underlined the ongoing efforts to address ethical concerns and enhance collaboration for future breakthroughs.
Faisal Mahmood, PhD
Harvard University
Dr. Mahmood is an Associate Professor of Pathology at Harvard Medical School and the Division of Computational Pathology at the Brigham and Women's Hospital. He received his Ph.D. in Biomedical Imaging from the Okinawa Institute of Science and Technology, Japan, and was a postdoctoral fellow at the Department of Biomedical Engineering at Johns Hopkins University. Dr. Mahmood directs the pathology image analysis research laboratory, and his research interests include image analysis, morphological feature, and biomarker discovery using data fusion and multimodal analysis. Dr. Mahmood is a full member of the Dana-Farber Cancer Institute / Harvard Cancer Center ; an Associate Member of the Broad Institute of Harvard and MIT, and a member of the Harvard Bioinformatics and Integrative Genomics (BIG) faculty.
Multimodal Generative AI for Pathology
Dr. Faisal Mahmood's keynote on "Multimodal Generative AI for Pathology" showcased advancements in computational pathology. He introduced a three-dimensional (3D) computational framework that utilizes deep learning to capture the full complexity of tissue structures, significantly enhancing diagnostic accuracy beyond traditional two-dimensional (2D) tissue analysis. Dr. Mahmood also highlighted the integration of multimodal data, combining histology with genomic data to improve cancer prognosis predictions.
Further, he discussed using weakly supervised learning techniques better to handle the growing volume of digital pathology data, reducing the variability seen with human observers and enhancing diagnostic precision. The highlight of his presentation was PathChat, a vision language-based AI assistant developed using a massive dataset of histology images and pathology captions. PathChat uses a sophisticated vision-language model to provide highly accurate diagnostic support, proven especially effective in educational and clinical decision-making contexts.
Additionally, Dr. Mahmood introduced CONCH, a visual-language foundation model that excels in various pathology tasks through its innovative use of contrastive learning from vast amounts of image-caption pairs. This model sets a new histopathological standard by enabling effective learning from limited labeled data and supporting diverse diagnostic applications without extensive retraining.
Opening Remarks
Sian Leah Beilock, PhD
The President of Dartmouth College
Guest Speakers
Timothy Burdick, MD, MBA, MSc
Associate Professor of The Dartmouth Institute, Geisel School of Medicine, Dartmouth
Associate Professor of Community and Family Medicine, Geisel School of Medicine, Dartmouth
Jessica Sin, MD, PhD
Assistant Professor of Radiology, Geisel School of Medicine, Dartmouth
Jennifer Hong, MD
Assistant Professor of Surgery, Geisel School of Medicine, Dartmouth
Assistant Professor of Molecular and Systems Biology, Geisel School of Medicine, Dartmouth
Louis Vaickus, MD, PhD
Assistant Professor of Pathology and Laboratory Medicine, Geisel School of Medicine, Dartmouth
During the Clinician Panel moderated by Dr. Timothy Burdick with Drs. Jessica Sin, Jennifer Hong, and Louis Vaickus, panelists, explored the transformative impact of AI on diagnostics and care efficiency. They highlighted the vast potential of AI while emphasizing the need for its careful implementation to ensure patient safety and equitable healthcare delivery. They also noted that AI's benefits in healthcare require responsible integration into clinical practices.
Dr. Sin emphasized the critical need to scrutinize AI-driven decisions, pointing out that while AI offers substantial benefits, healthcare providers must remain vigilant about the technology's recommendations. Dr. Burdick highlighted AI’s potential to reduce wait time and extend care to underserved populations, underscoring that embracing these technologies responsibly is crucial for advancing health equity.
Barry Schweitzer, PhD
Associate Director of the Magnuson Center for Entrepreneurship at Dartmouth
Dr. Barry Schweitzer discussed Dartmouth's dedicated efforts to support entrepreneurship as he highlighted the institution's commitment to transforming healthcare innovations into scalable solutions through the programs that nurture entrepreneurial skills among researchers and provide essential resources like funding, mentorship, and a series of educational workshops on topics from cybersecurity to commercialization strategies. Through these efforts, Dartmouth has successfully launched startups, licensed new technologies, and fostered a culture of innovation that connects academia with industry, aiming to improve people's lives with new medical solutions.
Soroush Vosoughi, PhD
Assistant Professor of Computer Science, Dartmouth College
Technical Associate Director, CPHAI
Advocating for a transparent approach, Dr. Vosoughi introduces the AI Alliance, a collaborative effort aiming to democratize AI technology by developing open models and standards, ensuring AI advancements are accessible and beneficial to all. Dr. Vosoughi emphasizes the importance of open innovation in the AI field. He points out the drawbacks of the current trend where big corporations dominate with closed-source tools, limiting intellectual security and broader benefits.
Thomas Thesen, PhD
Associate Professor of Medical Education, Geisel School of Medicine, Dartmouth
Dr. Thomas Thesen talked about harnessing generative AI to transform medical education through his initiative "Conversational Generative AI for Precision Medical Education." This approach integrates advanced AI technologies like large-language models and AI Patient Actors, designed not only to improve diagnostic accuracy for trainees but also to simulate real-life medical interactions, thereby enriching the learning experience for medical students with diverse educational backgrounds and learning styles. Dr. Thesen’s work emphasizes creating personalized and interactive educational environments that prepare students more effectively for their medical careers.
Parth Shah, MD
Assistant Professor of Pathology and Laboratory Medicine, Geisel School of Medicine, Dartmouth
Dr. Parth Shah's talk focused on integrating cutting-edge genomic-based technologies into clinical practice. He detailed the shift from older sequencing methods to newer, more efficient techniques like Massively Parallel Sequencing and Optical Genome Mapping, which help physicians diagnose and treat diseases like leukemia more accurately. Dr. Shah discussed the essential elements of a successful clinical genomics program, which covers a wide range of patient care to sophisticated lab work and data analysis. He also addressed the challenges in expanding these technologies, such as the high costs and the need for advanced data handling systems. He suggested solutions like automated AI data processing methods and tools to manage and analyze genetic information more effectively.