Accelerating Healthcare Research with AIDPI

 · The Workshop on AI-based Digital Phenotyping and Interventions (AIDPI), held on May 16, 2024, brought together experts to explore AI's potential to improve healthcare outcomes and reduce costs by enhancing the productivity of research. 

· The lead for the AIDPI Workshop is Professor Robert Morris. It was held at the NRF (National Research Foundation) Singapore and attracted around eighty participants who attended both in person and virtually. The attendees comprised individuals from various local institutions including the National Healthcare Group, NUS, NTU, A*STAR, Singapore-ETH Centre, as well as one international participant from City University of Hong Kong. 

· The Workshop's main presentations were organized into two sessions. The first session delved into topics such as chronic disease, behavior, and coaching, while the second session centered on mental health, monitoring, and recovery. 

· The breakout sessions conducted during the Workshop meticulously scrutinized the intricacies surrounding data sharing, patient engagement strategies, and the seamless integration of AI technologies into clinical settings. The breakout session concluded by urging interdisciplinary collaboration and addressing systemic obstacles to foster healthcare innovation. 

· At the end of the Workshop, a poll was conducted to gauge the perspectives of attendees and participants regarding the potential of AI to enhance healthcare outcomes and reduce costs through heightened research productivity. The consensus resoundingly favored the affirmative stance, underscoring a prevailing sentiment in support of the notion of AI-based Digital Phenotyping and Interventions. 

The main talks in the Workshop were divided into two sessions. Session 1 discussed chronic disease, behavior, and coaching, while the second session focused on mental health, monitoring, and recovery. A summary of the talks is as follows: 

Deep phenotyping provides a detailed analysis of phenotypic abnormalities, which is crucial for developing disease predictors and patient-specific assessments. AI models utilizing multi-modal data, including behavioral and genetic information, were discussed for their potential in predicting disease risks and suggesting personalized interventions. Foundation models aim to integrate diverse data sources, achieve expert-level performance on medical tasks, and maintain human expertise in patient interactions. The presentation concludes by introducing the grand challenge, establishing the foundation for the subsequent progression of the workshop's agenda. 

The EMPOWER project focuses on using AI to create personalized nudges and health advice for chronic disease management. Key features include an AI-driven nudging system, personalized health notifications, and a localized GPT model trained on medical data. Validation through medical trials has shown promising results in improving patient outcomes. 

The core objective of the AMI-HOPE (Acute Myocardial Infarction) initiative is to enhance post-discharge outcomes and minimize expenses by enabling health professionals to actively oversee and offer remote assistance to patients following their hospital discharge. Under this new care model, patients use Health Discovery+ to share vital signs with pharmacists, enabling timely medication adjustments and early advice within a week of discharge.  

The WellFeet platform addresses the gap in digital education for diabetes management by providing a user-friendly application with multimedia content and interactive tools. A feasibility study showed significant improvements in health literacy and self-care behaviors among users, paving the way for future platforms targeting cardiovascular disease prevention. 

The workshop highlighted the need for new intervention pipelines for maternal mental health, focusing on preconception screening to identify women at risk. AI was discussed for its potential to enhance mental health interventions through risk detection, feature analysis, and tailored cognitive behavioral therapy (CBT). 

This talk provides an overview of the state of mental health in Singapore. Additionally, it highlights transformational shifts in healthcare through the deployment of an evidence-based framework, explaining how AI has been successfully implemented to address both serious mental health conditions and mental wellness. 

This talk provides an in-depth look at HOPES (Health Outcomes through Positive Engagement and Self-Empowerment), a platform co-developed by MOHT (MOH Office for Healthcare Transformation) and the Institute of Mental Health (IMH). The project aims to improve the outcomes of patients with mental disorders by leveraging on passive and active digital data collected from smartphones and wearables. One of the goals is to empower users with knowledge of their own health and personalized digital interventions. 

AI's role in accelerating the research process through transfer learning and state-space modelling was emphasized. AI's ability to reuse knowledge from one task to boost performance on related tasks enhances the efficiency of developing and applying medical interventions.