BPIPO and DCB Collaborate with NVIDIA to Host AI Forum, Focusing on Cross-Domain Integration for Precision Health

The Biotechnology and Pharmaceutical Promotions Office (BPIPO) and the Development Center for Biotechnology (DCB) successfully held the “Generative AI in Smart Biomedical and Drug Discovery Technology Seminar” in collaboration with international tech giant NVIDIA. Experts from NVIDIA, along with various partner companies and organizations, gathered at the Nangang Exhibition Center in Taipei. The discussions centered on the application of generative AI in precision smart healthcare and Taiwan’s development opportunities in this global wave.

NVIDIA Provides Powerful Computing Resources to Drive Diverse Innovations in AI Healthcare

In his welcoming remarks, Director Michael Huang of BPIPO stated that AI can significantly enhance medical imaging, smart biomedicine, drug discovery and precision medicine. Coincidentally, NVIDIA has established Taiwan’s largest AI supercomputer, Taipei-1, in Kaohsiung this year. This supercomputer provides robust computing resources for companies or research institutions needing to develop AI models. Additionally, with the vast data from the National Health Insurance database, Taiwan’s AI smart biomedicine is poised to reach new heights.  Dr. Ettikan Kandasamy Karuppiah from NVIDIA emphasized that generative AI is at the forefront of healthcare innovation, contributing substantially to healthcare decision-making, drug discovery and development, medical imaging, and genomics. Furthermore, NVIDIA’s cloud-native microservices (NVIDIA Inference Microservice, NIM) simplify the deployment and application of generative AI models in various research and clinical scenarios, ultimately improving medical outcomes for patients worldwide.

Insights into precision health industry trends; Experts Share Successful Examples of AI Tools in Practice

Dr. Ying-Ja Chen, Solutions Architect for Biomedicine at NVIDIA, highlighted that with the global wave of generative AI, digital biology, digital surgery, and digital health will see tremendous advancements. For example, in digital surgery, AI models can analyze digitized medical images to detect and adjust surgical instruments in real-time, enhancing surgical precision. Looking ahead, there may even be a possibility of medical equipment and surgical robotics becoming commonplace. In the realm of digital health, AI integration can reduce the workload of healthcare personnel, allowing them to focus on patient interactions. This improvement can enhance the doctor-patient relationship and increase healthcare outcomes. Large language models (LLMs) can assist doctors in documenting medical histories and patient conditions while providing educational information tailored to patient needs. Through the NeMo Retriever NIM microservice, AI models can integrate and retrieve vast healthcare data, such as the National Health Insurance database, driving interactions between virtual caregivers and patients to address their inquiries. In fact, NVIDIA can offer diverse solutions for various research and healthcare institutions, including the generative AI-driven platform BioNeMo for drug development optimization and validation., the open-source medical imaging AI software MONAI, the computation platform Holoscan for medical devices, and the LLM training platform NeMo.

NVIDIA’s Solutions Architect for Biomedicine in the Asia-Pacific region, Ken Ying-Kai Liao, centered his presentation around two cutting-edge technologies: MONAI and Holoscan. He explained their critical roles in AI digital surgery. MONAI enables researchers and clinicians to develop and train high-performance AI models and deploy them to hospital IT systems, medical devices, and any cloud infrastructure, accelerating medical imaging analysis for more accurate diagnoses. Additionally, NVIDIA integrates MONAI with federated learning technology, allowing different healthcare institutions to collaboratively train AI models without sharing patient data, thus addressing both model performance and privacy protection. The Holoscan edge AI processing platform connects to hospitals’ smart medical devices, reducing latency in image operations and enabling AI models to process and analyze imaging and sensor data in real time. This capability minimizes misjudgment risks and enhances surgical accuracy and safety. Moreover, Holoscan includes multiple software development kits (SDKs) that support multimodal sensor data processing, allowing developers to create new AI medical devices and bring them to market.

The Seminar Featured Experts from NVIDIA Discussing Generative AI Applications in Smart Biomedicine, Along with Partner Companies Sharing Successful Examples of AI Tools in Practice

The seminar invited NVIDIA experts to explain the applications of generative AI in smart biomedicine, alongside numerous partner companies and public and private entities that shared successful implementations of AI tools. General Manager Xu Junpeng (許駿鵬) of Findings Tech and Product Manager Li Zonglin (李宗霖) of YUAN discussed how they utilized NVIDIA’s powerful AI resources to develop innovative medical products. Fan Ding Technology leveraged the MONAI software to analyze patients’ liver MRI images, thereby reducing false positives and false negatives during image interpretation. This improvement enables doctors to accurately identify patients with fatty liver and intervene early to control the condition. On the other hand, Congtai Technology integrated its low-latency imaging technology with Holoscan data processing and the MONAI imaging analysis platform to create an AI-driven medical endoscopy system. This advancement enhances the precision and efficiency of endoscopic surgeries. Furthermore, combining ultrasound devices with AI technology assists doctors in evaluating patients’ liver health and detecting liver tumors, significantly reducing time and costs compared to traditional methods.

Product Manager Hua Zhizhong (花志忠) of Advantech discussed the flexible deployment of on-premises large language models. He pointed out that generative AI can assist in generating medical reports, compiling voice records, managing nurse handovers, and improving doctor-patient interactions. This assistance boosts healthcare workers’ efficiency, streamlines hospital management processes, and reduces medical disputes. Regarding the deployment of generative AI and LLM applications in smart healthcare, Advantech proposed a “hybrid cloud architecture” that combines the strengths of public and private clouds. The public cloud offers rapid computing capabilities and flexible deployment, making initial testing more cost-effective. Conversely, private clouds are suitable for large-scale LLMs and scenarios requiring low latency and data confidentiality, such as data processing within healthcare institutions. This combination not only provides flexible deployment models for generative AI but also allows dynamic adjustment of cloud and on-premises resource allocation according to different needs, thus avoiding unnecessary cybersecurity risks. Additionally, Manager Hua mentioned Advantech’s achievements in developing autonomous mobile robots (AMR) equipped with NVIDIA AI tools, including glass bottle inspection devices and hospital cleaning robots.

The NeMo Platform Contains Powerful Features to Train LLMs, Meeting Diverse Research and Healthcare Scene Needs

Dr. William Wong (黃威仁), Senior Developer Relations Manager at NVIDIA, shared how to train LLMs with the NeMo platform and develop clinical-grade apps for healthcare applications. He first introduced NVIDIA’s NIM microservices, which operate at least three times faster than the native Llama 3.1 model. This service includes AI training and inference frameworks, as well as assessment and protective tools that ensure the output of newly trained LLMs. These tools evaluate the quality of medical LLM outputs and prevent the model from disclosing patient information or generating irrelevant data. Moreover, NIM is user-friendly; developers can download it directly and deploy it in their own development environments. Dr. Huang also mentioned NVIDIA’s initiative to create a sovereign AI model in Taiwan, training it with traditional Chinese data to help the model grasp Taiwan’s cultural and industrial knowledge while responding fluently in traditional Chinese.

Following Dr. Huang’s presentation, the conference invited three other partner organizations working with NVIDIA to share their applications of generative AI tools in the healthcare industry. Dr. Hu Xiangwei (胡翔崴) from the Industrial Technology Research Institute introduced the collaborative outcomes achieved with NVIDIA. By integrating Jetson, NeMo, and other accelerated software with edge computing platforms, the team successfully developed various multimodal language model applications suitable for the healthcare sector. Specific innovations include AI-assisted drug development, telehealth for chronic kidney disease, personalized health education, long-term care voice dialogue companion robots with 12 built-in emotional modules, and 3D imaging and generation tools to assist dentists with surgical planning and training. Dr. Lin Jingheng (林敬恒), Assistant Researcher at Chang Gung Memorial Hospital, explained how the Chang Gung team leveraged the NeMo platform to train autonomous LLMs based on the hospital’s accumulated care SOPs. This system has evolved to its second generation, providing text, image, and table responses, while incorporating instant messaging functionality to effectively support knowledge retrieval during nursing consultations. This solution addresses gaps in professional knowledge and experience due to the loss of senior nurses.

AI-Assisted Pancreatic Cancer Detection System Gains International Recognition; NVIDIA Inception Program Promotes Startup Development

Professor Wang Weizhong (王偉仲), co-founder of PanCAD.ai, introduced the world’s first AI-assisted pancreatic cancer detection system, PANCREASaver®. This system targets two significant challenges in pancreatic cancer: “the difficulty of detection with the naked eye leading to low survival rates” and “frequent misdiagnosis or missed diagnoses.” By combining machine learning and deep learning techniques, the system can automatically analyze CT images to identify pancreatic cancer and locate tumors while generating detection reports. Verified by national datasets, the system achieves nearly 90% accuracy in detecting pancreatic cancer and can handle cases where distinguishing between benign and malignant is challenging. It even identifies certain unique types of pancreatic tumors. Currently, the system has obtained medical device certification from Taiwan’s Ministry of Health and Welfare and breakthrough medical device designation from the FDA in the United States. Additionally, it has received the prestigious Margulis Award in Radiology. The system is expected to be launched at National Taiwan University Hospital in December. Furthermore, PanCAD.ai has joined NVIDIA’s Inception Program, which supports innovative startups and assists them in integrating AI technologies to develop healthcare applications. Professor Wang emphasized that through collaboration with NVIDIA and the program’s resources, the company has significantly accelerated its development.

Taiwan’s Unique Strengths in Healthcare and ICT Industries Set Stage for Promising AI-Driven Future

As one of the co-organizers of the AI forum, GeneOnline had the privilege of interviewing several experts and speakers in attendance to learn their perspectives on Taiwan’s growth potential within the generative AI wave. General Manager Junpeng Xu of FindingsTech emphasized that Taiwan boasts top talent in both healthcare and technology fields, paving the way for exceptional development through cross-disciplinary collaboration. Dr. Ying-Chia Chen of NVIDIA pointed out that with its numerous medical centers and concentrated patient population, Taiwan serves as an ideal testing ground for medical AI technologies. She also encouraged ICT industry players to leverage these advantages by adopting a more open mindset toward collaboration with healthcare professionals and utilizing AI to promote advancements in biomedical fields.

Dr. Jing-Heng Lin of Chang Gung Memorial Hospital noted that Taiwan’s healthcare environment still faces multiple pain points that need addressing, with front-line healthcare workers experiencing significant stress. However, he believes that “crisis is opportunity,” with each challenge representing a potential testing ground for emerging AI technologies. Advantech Manager Zhizhong Hua stated that Taiwan’s healthcare industry attracts top talent and ranks among world leaders in applied research. Through generative AI, Taiwan can explore more applications, add value, and further strengthen its healthcare industry. Product Manager Zonglin Li of YUAN shared that with NVIDIA’s Taipei-1 now operational, major Taiwanese companies and research institutions are committing resources to AI training and research, positioning Taiwan for a strong presence in the global market. Dr. Xiang-Wei Hu of the Industrial Technology Research Institute (ITRI) remarked that Taiwan holds clear advantages in both hardware and data resources. With multiple NVIDIA supply chains established locally, customers in Taiwan can readily access chips. Combined with Taiwan’s extensive clinical data, leading Asia in this regard, there is significant support for training new AI models.

Zhongyou Yeh, Manager at the National Health Research Institutes (NHRI), highlighted the organization’s recent partnerships with ASUS and NVIDIA to establish the AI Biomedical Information Innovation Research and Application Service Center, offering hardware resources and an AI management development platform. NHRI has also collaborated with Taiwan Web Service (Taiwan AI Cloud) to build Taiwan’s first AI supercomputer dedicated to biomedicine. Patrick Cheng, Manager at Bamboodd, expressed confidence in Taiwan’s strong growth prospects within the global AI landscape. Taking Bamboodd as an example, they have applied generative AI in voice analysis, enabling them to identify users’ emotional states and successfully develop the world’s first Chinese-language emotion analysis app, aimed at mental health applications. Asia Pathogenomics (APG)  COO Ting-Huan Li stated that generative AI not only has potential for image analysis and drug design but could also offer guidance to doctors, supporting clinical diagnosis and disease prevention. CTO Yulun Guo pointed out that Taiwan has robust capabilities in both software and hardware, but a strong infrastructure must be supported by a steady pipeline of professional talent to achieve meaningful industry-wide advancements.

Finally, Chairman Shiing-Jer Twu of the Development Center for Biotechnology stressed, “Generative AI should be used to inspire greater creativity and productivity.” For Taiwan, the foundation of its healthcare and public health systems is particularly strong, with its national health insurance system leading globally. Related databases play a critical role in supporting the development of generative AI applications. However, he cautioned that Taiwan should not only focus on “industrializing AI” but also prioritize “AI-driven industrial transformation” for long-term impact, harnessing generative AI to enhance industry quality.

With strong promotion from BPIPO and DCB, coupled with NVIDIA’s diversified AI software and hardware solutions and services, generative AI is poised to provide critical support in driving the development of Taiwan’s smart biomedical and precision health sectors. Looking ahead, Taiwan’s biotech companies, medical institutions, and academic research units can leverage healthcare databases and clinical resources, taking advantage of Taiwan’s ICT industry strengths to align with global AI innovations. This will enable AI-driven industrial transformation, enhance healthcare service quality, and improve industry competitiveness. With the launch of the NVIDIA Taipei-1 supercomputer continuously upgrading computing resources, Taiwan is set to become an ideal testing and application ground for generative AI in healthcare, benefiting the general population and patients, while also leading Taiwan’s biomedical industry into a promising new chapter.

Author

GeneOnline