Leveraging Supercomputing Facility, Singapore’s National University Health System Develops Large Language Model for Clinicians

The National University Health System (NUHS) in Singapore has developed a large language model called NUHS RUSSELL-GPT, which is aimed at enhancing clinicians’ productivity. This model can quickly summarize patient case notes and generate referral letters for doctors in a matter of seconds. Additionally, NUHS staff can ask questions related to medical conditions and clinical practice guidelines.

The NUHS RUSSELL-GPT model plays a key role in the Patient Trajectory Prediction AI model developed by NUHS. By analyzing historical data, the AI model predicts individual patient healthcare journeys, aiming to improve patient outcomes and optimize healthcare delivery.

NUHS plans to gradually roll out the healthcare language model across its cluster. It also intends to use the model to predict the severity and trajectory of common conditions, such as urinary tract infections.

The development of RUSSELL-GPT is part of NUHS’s larger AI program, which leverages the Prescience supercomputing infrastructure. The recently operational petabyte-scale supercomputing facility, co-developed with the National Supercomputing Centre Singapore, utilizes multiple NVIDIA DGX A100 compute nodes to accommodate large language models.

In addition to RUSSELL-GPT, Prescience has also been used in the development of the SMILE AI project. This project focuses on creating a 3D teeth charting model, called Smart Monitoring and Intelligent Learning for Enhancing oral health (SMILE), which provides digital representations of the condition and position of teeth.

Another project in the works is Secure GPT for Healthcare Professionals, a healthcare language model being developed by Synapxe and Microsoft. This technology, built using the Azure OpenAI Service, aims to generate condensed patient information from doctors’ clinical notes and laboratory reports in the electronic medical record system. It will also track medication changes and usage, improving the efficiency and accuracy of healthcare processes.

Dr. Derrick Au, Group Chief Transformation Officer at NUHS, expressed excitement about the potential impact of NUHS RUSSELL-GPT on healthcare delivery, stating that it has the potential to revolutionize the way healthcare is delivered at NUHS by synthesizing local medical knowledge and reducing administrative work.

Sources:

Source 1: AI-Med

Source 2: NUHS Official Website

Source 3: Microsoft Official Website

Definitions:

• Language Model: A language model is a model that is trained to predict the next word in a sequence of words, given the context of the preceding words.
• Supercomputing Facility: A supercomputing facility is a data center equipped with high-performance computers that are capable of carrying out complex computations at a much higher speed than conventional computers.
• Patient Trajectory Prediction: Patient trajectory prediction refers to the process of using historical data to predict the future health outcomes and healthcare needs of individual patients.
• Generative AI: Generative AI refers to AI systems or models that are capable of generating new and original content, such as images, text, or audio.
• Electronic Medical Record (EMR) System: An electronic medical record system is a digital version of a patient’s medical history, including diagnoses, medications, and treatment plans.