Artificial Intelligence (AI) has been increasingly employed in healthcare, with new technologies and systems creating better patient outcomes and experiences, improving clinical workflows, and assisting in decision making.
My work in SingHealth AI Office focuses on Research and Development (R&D), commercial partnerships, supporting digital integration and education. The key focus areas for this programme are to develop more AI-powered clinical solutions using different cutting-edge technologies, rapidly bring these technologies to the bedside for clinical implementation and scale them in the global setting.
Over the years, SingHealth has cumulatively secured substantial R&D funding across all its institutions. The use of R&D spans across clinicians, healthcare innovators, nurses, allied health professionals and administrative officers. Like a diagnostic system, AI is able to segment, classify and even make predictions using multi-modal data. For example, we can predict whether a patient requires admission into an intensive care unit, a ward, or be discharged based on their medical information on arrival in an Accident and Emergency (A&E) setting.
Improving AI and data trust in healthcare
Federated learning with the addition of blockchain technology protects the privacy and security of healthcare data. By decentralising healthcare data sharing through blockchain, we can protect the privacy of users and enable data empowerment. Federated learning is used to train other machine learning algorithms by using multiple local datasets without exchanging data. This allows personal data to remain in local sites, reducing possibility of personal data breaches. Integrating blockchain and federated learning with AI can transform Singapore’s existing healthcare ecosystems by optimising clinical workflows, improving trust among stakeholders and shortening existing processes, brought about by automating data verification and transfer.
With SingHealth’s extensive clinical and research capabilities and together with our talented pool of clinicians and healthcare innovators, the AI Health Programme is actively reaching out to industry partners to explore ways to commercialise the numerous novel technologies that we are building within the SingHealth Duke-NUS AMC.
Improving clinical and technical workflow
The creation of clinical and technical sandboxes will expedite the translation of AI technology to clinical implementation in the SingHealth system. Clinical sandboxes will attract AI companies and overseas industry partners to test and use the data sets that we have created and curated. Once the AI algorithms have satisfied clinical performance, we proceed to the implementation stage with the technical sandboxes. The technical sandboxes will shorten the journey and overcome the “innovation valley of death” for innovators. They will have a better understanding of the innovation process, allowing them to overcome the dangers that they are likely to face along the way, and maximise their opportunities for success....
Advancing healthcare AI
The three-year partnership between SingHealth and SGInnovate seeks to advance the development and adoption of AI and other emerging technologies, improve diagnostics and treatment, and enhance the delivery of healthcare and clinical outcomes in Singapore. Its first focus area: Healthcare AI. Through education, we aspire to build a critical mass of current and next generation of AI-literate practitioners within the healthcare space.
SELENA+ is the first home-grown, deep-learning AI system that helps to identify patients with three major types of eye disease: diabetic retinopathy, glaucoma and age-related macular degeneration. Patients with diabetic eye conditions will then undergo a second round of grading by their eye care provider to ensure an accurate diagnosis. The ability of SELENA+ to detect these diseases with greater speed and accuracy results in manpower and cost savings.
We have recently won a $25 million grant from the AI in Health Grand Challenge to look into how we can harness the power of AI to reduce the incidence of progressions of 3H - Hypertension (high blood pressure), Hyperlipidemia (high cholesterol) and Hyperglycemia (diabetes) - within our population. Over the next few years, we will be building comprehensive diabetes registries and a predictive model. This will tremendously reduce morbidity and mortality resulting from 3H.
For innovation to succeed, it is important that we identify long term goals and implement feasible ideas. It is also essential for AI researchers and clinicians to work closely during the clinical design phase.