SINGAPORE - A new national supercomputer to speed up healthcare research using large and complex sets of patient data is slated to launch this year.
One of the first projects for the supercomputer will be to help design an algorithm analysing clinical data and heart and retinal scans for heart patients so that the most urgent cases can be prioritised for emergency treatment. The machine is being developed by the National Supercomputing Centre (NSCC) and SingHealth.
While it will be located at the healthcare cluster's Singapore General Hospital (SGH), it will be open to all healthcare institutions to use.
Agreements were signed between NSCC, SingHealth and American chipmaker Nvidia for the supercomputer on Tuesday (March 1) at the Supercomputing Asia 2022 event at Suntec Singapore.
Under the agreements, Nvidia will provide access to its software tools and artificial intelligence (AI) models for healthcare research.
The supercomputer is expected to take the AI used by clinicians to make sense of diseases and escalate it. Where standard computers might take days to train the AI using a set of patient data, the supercomputer could do it in hours.
Said Mr Bernard Tan, NSCC's director for strategy, planning and engagement: "We expect the supercomputer to be up to 100 times the speed of a standard computer."
Having the supercomputer in-house also saves time as data does not have to be sent to a third party high-performance computer to be analysed, a process that could take hours to days. To protect patients' privacy, their data will be anonymised so no names and NRIC numbers will be associated with the information.
Beyond healthcare, supercomputers like those from NSCC's have been used here in other research areas, including climate change, weather monitoring, urban planning, and materials research.
The new supercomputer will also be used to train the AI tool used in Apollo, a national AI project to analyse scans of heart arteries to determine if a patient has cardiovascular disease.
While a radiologist could take several hours to interpret a scan, an analysis could be completed using Apollo in 10 minutes. This would free up time for clinicians to focus on other important work such as caring for patients.
Apollo is a collaboration between SingHealth's National Heart Centre Singapore, the Duke-NUS Medical School, Agency for Science, Technology and Research (A*Star), National University Hospital and Tan Tock Seng Hospital.
The supercomputer can also train other AI models, such as the one being used for a project to identify "neoantigens" - new proteins with mutations found on cancer cells.
By figuring out what these proteins are, researchers can develop vaccines to stimulate the production of antibodies that can target them and kill the cancer cells.
Such precise targeting reduces the side effects of the cancer therapy, unlike conventional cancer treatments that are less precise and may harm non-cancerous cells.
SingHealth's National Cancer Centre Singapore is working on this AI project to develop treatments that can prevent relapses in liver cancer.
But in the future, the technology could be used to develop vaccines against diseases such as influenza and those caused by coronaviruses like Covid-19, making it an important tool in preparing for future pandemic threats.
Speaking at the Supercomputing Asia event on Tuesday, Senior Minister of State Communications and Information Janil Puthucheary said that Singapore's supercomputing infrastructure has helped to support the country's Covid-19 research efforts.
For example, researchers here developed a simulation which more accurately modelled the spread of droplets when a person with Covid-19 coughs in Singapore's tropical environment.
The scientists had worked with NSCC to feed past data on the coronavirus into a supercomputer to produce the simulation.
It helped the authorities plan safe management measures to limit the spread of Covid-19.
The cost of the SingHealth-NSCC supercomputer was not immediately available. But a previous supercomputer called Aspire 2a under NSCC cost $40 million to build.
The new supercomputer comes after a similar announcement in December last year that NSCC and National University Health System were building a supercomputer to advance research in Covid-19 and other areas of healthcare.
This included using the supercomputer to train AI to predict the path a patient's condition could take in the future. The machine is expected to be ready by the middle of this year.