Establishing a comprehensive “AI-assisted Physician” using generative AI from Google Cloud, including MedLM

Superintendent Der-Yang Cho of CMUH announced a collaboration with Google Cloud to develop a large language model, MedLM, based on Med-PaLM 2. CMUH is the first medical center in Asia to adopt Med-PaLM 2.

China Medical University Hospital (CMUH) in Taiwan announced its collaboration with Google Cloud in mid-December 2023. Using Google Cloud's generative AI technology, including MedLM, which is a large language model built on Med-PaLM 2, CMUH has developed a comprehensive "AI-assisted Physician" system which aims to assist healthcare professionals in disease diagnosis, treatment planning, patient education, and medical research, etc.

Aashima Gupta, Global Director of Healthcare Strategy and Solutions at Google Cloud, elaborated on how generative AI can serve as a crucial ally for healthcare professionals to reduce their workload and improve personal care for patients.

The collaboration between CMUH and Google Cloud will introduce assistive tools designed to support medical professionals in precision cancer treatment. These tools, including “Customized Cancer Treatment Guidelines” and “Cancer Therapy Q&As,” aim to reduce the time spent searching for information and enable the quick capture of accurate data. This can aid professionals in developing cancer treatment plans, offering personalized treatment-related information to patients, and providing responses to patient health education inquiries.

Greg Corrado, a recognized Google AI influencer and a Distinguished Scientist and Senior Research Director at Google Research, emphasized the significance of collaborating with medical centers like CMUH in testing and deploying new generative AI solutions.

CMUH is one of the first university hospitals in Asia to test MedLM. Through its collaboration with Google Cloud, CMUH is dedicated to driving AI innovation in smart healthcare to shape the future of healthcare in Taiwan. MedLM will help facilitate CMUH in the timely access of accurate medical information to establish the best Healthcare AI models for the Chinese-speaking market in Asia.

Prof. Der-Yang Cho, Superintendent of CMUH, pointed out that MedLM is advanced in its ability to provide precise and rapid data analysis for drug guidelines, disease gene sequencing, and medical records, among other areas. Such capabilities can significantly accelerate the discovery of new treatments for difficult-to-treat diseases by medical teams and assist in the rapid development of treatment plans to help patients. “We are excited about this collaboration, which positions smart hospitals in Taiwan at the forefront of providing more precision- and safety-oriented support to healthcare professionals and patients alike,” Prof. Der-Yang Cho said.

Kathy Lee, Managing Director at Google Cloud North Asia, highlighted that Google Cloud’s products and generative AI solutions are tailored specifically for the healthcare industry to boost operational efficiency in healthcare and life science organizations while enhancing patient care outcomes.

CMUH was already recognized as the world's third smart hospital by the US. Healthcare Information and Management Systems Society (HIMSS) in their 2022 Digital Health Indicator (DHI) ranking. Additionally, Newsweek named CMUH as one of the World’s Best Hospitals for 2023 and one of the World’s Best Smart Hospitals for 2024. With its widespread recognition, CMUH is deeply committed to advancing the diversification and precision of smart healthcare services.

Prof. Oscar K. Lee, Vice Superintendent of CMUH, explained that the current collaboration with Google Cloud begins with precision treatment for difficult-to-treat cancers.

“Life sciences problems are fundamentally data problems, and this is where AI can fully unleash its potential,” said Kathy Lee, Managing Director of Google Cloud, North Asia. “Generative AI solutions from Google Cloud can help healthcare and life sciences organizations operate more efficiently and improve patient outcomes. In this remarkable era of gen AI, we are excited to collaborate with medical industry leaders that drive innovation through AI like CMUH, and partner with them to build the future of healthcare in Taiwan.”

Superintendent Der-Yang Cho of CMUH mentioned that Google’s MedLM, will enable the establishment of a comprehensive “AI-assisted Physician”, assisting medical experts in various domains, including disease diagnosis, treatment planning, patient education, and medical education research.

Aashima Gupta, Global Director of Healthcare Strategy and Solutions at Google Cloud, shared that healthcare professionals are committed to delivering the best care for patients, and gen AI can be a crucial ally in reducing their workload. Gupta pointed out that this technology will ease the cognitive burden for staff and improve personal care for patients. Furthermore, it will allow healthcare professionals at CMUH to devote more time to tasks that require human judgment and expertise. She mentioned that ensuring the safety and effectiveness of gen AI is fundamental to these benefits.

Greg Corrado, a Distinguished Scientist and Senior Research Director at Google Research, discussed how MedLM is designed specifically for the healthcare industry and is part of the growing family of gen AI technologies that have the potential to significantly improve the healthcare experience for professionals and patients. Corrado emphasized the importance of collaborating with healthcare institutions like CMUH in testing and implementing new gen AI solutions, underlining it as a critical step towards developing safe and effective AI technologies.

Prof. Der-Yang Cho, Superintendent of CMUH, said that the collaboration with Google will improve the overall clinical effectiveness and promote quality healthcare, as well as foster warmer interactions between doctors and patients.

Prof. Oscar K. Lee, Vice Superintendent of CMUH explained that among global generative AI large language models, Med-PaLM 2 stands out in model performance. It scored 85% on MedQA-USMLE (US Medical Licensing Examination), reaching a level of excellence comparable to professional physicians. This makes it a pioneer among the world's medical AI models in terms of outstanding performance. By training the model to analyze treatment guidelines and care cultures from various regions, MedLM will become even better at addressing the medical and healthcare needs of local populations.”

Currently, the collaborative applications between CMUH and Google Cloud include “Customized Cancer Treatment Guidelines” and “Cancer Therapy Q&As”. The “Customized Cancer Treatment Guidelines” are developed based on the National Comprehensive Cancer Network (NCCN) guidelines and are integrated with MedLM, enabling clinicians to quickly capture accurate information based on the patient's cancer staging, treatment methods, and risk assessments, etc. It can also generate cancer treatment plans and provide personalized treatment recommendations for clinicians to review, edit and use with patients. Focusing on health education for cancer patients, the “Cancer Therapy Q&As” application incorporated the cancer care-related health educational content of CMUH into the MedLM model for training and optimization, resulting in a system with extensive medical knowledge and Q&A capabilities. Such advancement not only diminishes the time healthcare professionals spend searching for information but also provides comprehensive treatment options. Furthermore, it contributes to enhancing overall clinical efficacy, improving patient understanding and engagement, and enhancing the quality of healthcare, thereby fostering warmer doctor-patient interactions and steering us collectively toward a healthier future.

Prof. Cho of CMUH said that Google’s MedLM can achieve precise and rapid data analysis for drug information, disease gene sequencing, and medical records, etc. This will accelerate research advancements for difficult-to-treat diseases by medical teams and assist in the development of precise treatment plans to help patients.

In addition, in terms of new drug development, CMUH is leveraging Google’s unique AI accelerators, TPU (Tensor Processing Unit), to assist in calculations related to protein folding and the development of new drugs. Preliminary tests at CMUH have shown that utilizing these resources can reduce the computation time for related programs by over tenfold.

With its robust research and development capabilities, CMUH’s Artificial Intelligence Center has provided over 20 distinct AI medical services in recent years, assisting doctors in making fast and accurate medical decisions.